JP7745451B2 - Signal processing device, signal processing method, and signal processing program - Google Patents

Description

This disclosure relates to a signal processing device, a signal processing method, and a signal processing program.

Active sensors output sound waves or radio waves, receive reflected sound waves or radio waves, and analyze the received signals. One analysis method involves visualizing (imaging) the detection results. In this case, the received information is processed to create an image. For example, Patent Document 1 describes a method for performing 3D imaging of synthetic aperture radar using compressed sensing.

Patent No. 6456312

Compressed sensing processing imposes a high computational load, and it can take a long time to obtain imaging results, especially when obtaining high-resolution images with small pixel sizes. Therefore, if detection results are required in real time, it will be necessary to wait for the imaging processing.

The present disclosure has been made in consideration of the above, and aims to provide a signal processing device, a signal processing method, and a signal processing program that can improve processing speed while maintaining resolution.

The signal processing device according to the present disclosure includes a data receiving unit that acquires observation data received at a plurality of measurement positions in a first direction when an emitted wave spreading in a second direction perpendicular to the first direction is transmitted from the plurality of measurement positions, arranges the observation data in a matrix in the first and second directions, and generates a received signal matrix in which the signal strength is a position value; a range processing unit that identifies a range in the second direction in the received signal matrix, generates an observation vector constructed based on the first-direction components of the identified range, and sets a sparse vector constructed of the first-direction components of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is greater than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix; a restoration processing unit that performs restoration processing in parallel for a plurality of ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a synthesis processing unit that synthesizes the sparse vectors calculated by the restoration processing to generate the imaging matrix.

The signal processing method disclosed herein includes a data receiving step of acquiring observation data received at a plurality of measurement positions in a first direction when an emitted wave spreading in a second direction perpendicular to the first direction is transmitted from the measurement positions, and arranging the observation data in a matrix in the first and second directions to generate a received signal matrix in which the signal strength represents a position value; a range processing step of identifying a range in the second direction in the received signal matrix, generating an observation vector constructed based on the first-direction components of the identified range, and setting a sparse vector constructed from the first-direction components of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is greater than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix; a restoration processing step of performing restoration processing in parallel for a plurality of ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a combination processing step of combining the sparse vectors calculated by the restoration processing to generate the imaging matrix.

The signal processing program disclosed herein causes a computer to execute the following steps: a data reception step of acquiring observation data received at multiple measurement positions in a first direction when an emitted wave spreading in a second direction perpendicular to the first direction is transmitted from the measurement positions, matrixing the observation data in the first and second directions, and generating a received signal matrix in which signal strength represents a position value; a range processing step of identifying a range in the second direction in the received signal matrix, generating an observation vector constructed based on the first-direction components of the identified range, and setting a sparse vector constructed from the first-direction components of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is greater than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix; a restoration processing step of performing restoration processing in parallel for multiple ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a combination processing step of combining the sparse vectors calculated by the restoration processing to generate the imaging matrix.

This disclosure makes it possible to improve processing speed while maintaining resolution.

FIG. 1 is a schematic diagram showing an example of a measurement system including a signal processing device according to this embodiment. FIG. 2 is a schematic diagram illustrating the measurement process of the measurement system according to this embodiment. FIG. 3 is an explanatory diagram for explaining an example of signal processing. FIG. 4 is a flowchart showing an example of processing in the signal processing device. FIG. 5 is a flowchart showing an example of processing in the signal processing device. FIG. 6 is an explanatory diagram for explaining another example of signal processing. FIG. 7 is a flowchart showing another example of processing in the signal processing device. FIG. 8 is a flowchart showing another example of the processing in the signal processing device. FIG. 9 is a schematic diagram illustrating another measurement process of the measurement system according to this embodiment. FIG. 10 is a flowchart showing another example of processing in the signal processing device. FIG. 11 is a flowchart showing another example of processing in the signal processing device. FIG. 12 is a schematic diagram showing a measurement system including a signal processing device according to a modified example. FIG. 13 is a diagram showing a measurement system according to a modified example.

Embodiments of the present disclosure will be described below with reference to the drawings. Note that the present invention is not limited to these embodiments. Furthermore, the components in the following embodiments include those that are easily replaceable by those skilled in the art, or those that are substantially identical. In the following description, an example will be given in which the first direction is the azimuth direction and the second direction perpendicular to the first direction is the range direction.

FIG. 1 is a schematic diagram showing an example of a measurement system 10 equipped with a signal processing device according to this embodiment. As shown in FIG. 1, the measurement system 10 includes an ultrasonic sensor 12, a signal processing device 14, a control unit 16, and a memory unit 18.

The ultrasonic sensor 12 is an active sensor that outputs sound waves for detection and receives reflections of the output sound waves to perform detection. The ultrasonic sensor 12 comprises a transmitter 32 and a receiver 34. The transmitter 32 outputs ultrasonic waves. In this embodiment, the transmitter 32 outputs a pulse signal. The receiver 34 receives the reflected waves output from the transmitter 32, reflected off something, and reaching the ultrasonic sensor 12.

In the ultrasonic sensor 12, the transmitter 32 and receiver 34 may be the same unit, or the transmitter 32 and receiver 34 may be arranged in an array to transmit and receive in all directions. In this embodiment, ultrasonic waves are used for detection, but radio waves, etc. may also be used. When using highly directional inspection waves as a sensor, the direction in which the inspection waves are emitted may be moved, i.e., swept, to detect the target area.

The signal processing device 14 is connected to the ultrasonic sensor 12, for example, via a pulser receiver 28. The signal processing device 14 processes the signals received by the ultrasonic sensor 12 to detect the surroundings. The signal processing device 14 has a first calculation unit 22, a second calculation unit 24, and a memory unit 26. The first calculation unit 22 is, for example, a CPU (Central Processing Unit). The second calculation unit 24 is capable of performing more parallel calculations than the first calculation unit 22, and is capable of performing dozens of parallel calculations. In this embodiment, the second calculation unit 24 is, for example, a GPU (Graphics Processing Unit). The memory unit 26 stores various information such as the calculation contents and programs of the first calculation unit 22 and the second calculation unit 24. The memory unit 26 includes at least one of a main memory unit such as RAM (Random Access Memory) or ROM (Read Only Memory), and an external memory unit such as an HDD (Hard Disk Drive).

In this embodiment, the storage unit 26 stores a signal processing program that causes a computer to execute the following steps: a data reception step of acquiring observation data received at measurement positions when transmitting waves that spread in the range direction perpendicular to the azimuth direction from multiple measurement positions in the azimuth direction, organizing the observation data into matrices in the azimuth and range directions, and generating a reception signal matrix in which signal strength represents a position value; a range processing step of identifying a range in the range direction in the reception signal matrix, generating an observation vector based on the azimuth direction components of the identified range, and setting a sparse vector composed of the azimuth direction components of the range in the range direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the azimuth direction is greater than that of the reception signal matrix and the number of divisions in the range direction is the same as that of the reception signal matrix; a restoration processing step of performing restoration processing in parallel for multiple ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a combination processing step of combining the sparse vectors calculated by the restoration processing to generate an imaging matrix.

The first calculation unit 22 includes a data receiving unit 42, a range processing unit 44, and a synthesis processing unit 48. The second calculation unit 24 includes a restoration processing unit 46.

The data receiving unit 42 acquires information on the received signals acquired by the ultrasonic sensor 12, performs demodulation, integration processing, etc., and in the first embodiment creates a received signal matrix that organizes the measurement results in the azimuth direction and the range direction within the range of the synthetic aperture length L _A. The data receiving unit 42 may perform demodulation processing, beamforming processing, pulse compression processing, range curvature correction, etc. on the received signals.

The range processing unit 44 selects an area to be processed from the matrix of pixels that make up the acoustic image (imaging image) created by processing the received signal matrix. The matrix of pixels that make up the imaging image is an observation matrix in which the number of pixels in the azimuth and range directions is preset by user specification, etc. The imaging image is created by processing the information of each pixel in the observation matrix based on the information in the received signal matrix. Here, the range to be processed is the pixel that is located at the same position in the range direction.

The restoration processing unit 46 performs processing based on the range selected by the range processing unit 44 and a known observation matrix, and calculates the range components. The restoration processing unit 46 uses, for example, ADMM (Alternating Direction Method of Multipliers). The restoration processing unit 46 restores the data in the range of the observation matrix using the information on the received signal and the range of the observation matrix. In this embodiment, the restoration processing unit 46 performs restoration processing for multiple ranges in parallel using the second calculation unit 24, which is composed of a GPU.

The synthesis processing unit 48 accumulates information for each pixel in the range of the imaging image acquired by the restoration processing unit 46 and creates an image of the synthetic aperture length.

The control unit 16 and memory unit 18 are mechanisms that control the movement of the ultrasonic sensor 12. The control unit 16 controls the movement of the ultrasonic sensor 12 by reading and executing a program (software) from the memory unit 18. The memory unit 18 stores various information such as the calculation contents and programs of the control unit 16. The memory unit 18 may also store the processing results detected by the signal processing device 14, i.e., the results of the exploration.

FIG. 2 is a schematic diagram illustrating the measurement process of the measurement system according to this embodiment. As shown in FIG. 2, the measurement system 10 is configured with a plurality of ultrasonic sensors 12 arranged side by side in the azimuth direction x. FIG. 2 illustrates a case in which there are M ultrasonic sensors 12, and the respective ultrasonic sensors 12 are identified as #1, #2, ..., #i, ..., #M. In the example shown in FIG. 2, the measurement target R of the measurement system 10 is metal, underwater, or the like. In the measurement system 10, a transmission signal ξ(t) is transmitted from one ultrasonic sensor 12 to the inside of the measurement target R, and the signal reflected by a reflection source l inside the measurement target R is received by the ultrasonic sensor 12 itself as a reception signal ζ _i (t).

In the example shown in Figure 2, the measurement system 10 uses a so-called linear scan measurement method to detect a reflection source l within the measurement target R, which is surrounded by the azimuth direction and the range direction perpendicular to the azimuth direction. In Figure 2, the search area is shown on one surface, but the search range is a three-dimensional area surrounded by the area reached by the transmitted wave.

When the linear scan measurement method is performed using the above-described measurement principle, the signal processing device 14 receives a received signal ζ _i (t), which is a real signal, from the ultrasonic sensor 12. The signal processing device 14 down-converts the real received signal ζ _i (t) to obtain a complex baseband signal.

In the measurement system 10, the received signal (real number) at the i-th measurement position is expressed as follows:
It can be written as:

Furthermore, the round-trip time τ _il′ (i=1, 2, ..., M, l′=1, 2, ..., L′) of the signal to and from the reflection source l is calculated using the position [X _i 0] ^T of the ultrasonic sensor 12 and the position [x _l′ r _l′ ] ^T of the reflection source as follows:
In the above [Equation 2], L' is the true value of the number of reflection sources.

The complex baseband signal obtained by down-converting this received signal is
It can be expressed as follows.

The received signal at each measurement position is expressed as a vector shown in [Equation 4] below.
To summarise,
The vector a _l′ represents the phase difference between the measurement positions.

Figure 3 is an explanatory diagram illustrating an example of signal processing. The left side of Figure 3 shows an example of a received signal matrix 130. In the received signal matrix 130, the horizontal direction is the azimuth direction (measurement position) and the vertical direction is the range direction (time). The right side of Figure 3 shows an imaging matrix 132 estimated based on the received signal matrix 130. In the imaging matrix 132, the horizontal direction is the azimuth direction and the vertical direction is the range direction. In the imaging matrix 132, each section represents one pixel of the acoustic image.

Here, assuming that a reflection source l such as a metal defect or target exists at the position of pixel l=1, 2, . . . , L (L _x L _y ) in the acoustic image shown on the right side of FIG. 3, the received signal model is as follows:
can be rewritten as:

The number of pixels L of the acoustic image (the number of pixels _Lx in the azimuth direction and the number of pixels _Ly in the range direction) is a parameter that can be arbitrarily set by the user. Therefore, it is unrelated to the true value L' of the number of reflection sources. Assuming the sparsity of the acoustic image, the signal _sl (t) representing the reflection intensity at the position of pixel l is approximately 0.

For example, when the number of pixels is 1000 x 1000, L = ¹⁰⁶ , and restoring an acoustic image all at once requires a significant amount of processing time. Therefore, in this embodiment, restoration is performed row by row in the range direction. The range r of the restoration target can be expressed as r = 0.5tν using the time t in the range direction and the sound speed ν, so it can be seen that there is a one-to-one correspondence between time and range. The coefficient 0.5 is a value used to obtain the required time for a one-way trip between the measurement position and the pixel position.

Here, the observation vector obtained by extracting only the waveform of a specific range r from the received signal η(t) is
Far away.

The reflection intensity at range r corresponding to the observation vector is expressed as a sparse vector:
Far away.

In this case, the correspondence between the observation vector and the sparse vector is
where l(r) represents the set of pixel numbers in the range r.

Also, A(r) is
and is a horizontally long matrix (M<<Lx) consisting of l(r) that satisfies the following [Equation 11].

The sparse vector s(r) can be estimated by applying a known compressed sensing restoration algorithm (e.g., the Alternating Direction Method of Multipliers) to the above [Equation 9]. The pixel region 134a of the matrix 130 shown on the left side of Figure 3 represents the observation vector η(r) at range r. The pixel region 136a of the matrix 132 shown on the right side of Figure 3 represents the sparse vector s(r) corresponding to the observation vector η(r) at range r.

In this way, restoration processing is performed using the observation vector η(r) for each range r and the observation matrix A(r) as representative values of the received signal matrix η(t). In this case, restoration processing can be performed in parallel for multiple ranges r. In this embodiment, the restoration processing unit 46 of the second calculation unit 24, which is composed of a GPU, performs restoration processing in parallel for multiple ranges r.

Next, an example of processing in the signal processing device 14 will be described using Figures 4 and 5. Figures 4 and 5 are flowcharts showing an example of processing in the signal processing device 14.

As shown in FIG. 4, first, before the measurement by the ultrasonic sensor 12, an observation matrix
is set (step S101) and stored in the storage unit 26 (step S102). The observation matrix does not depend on the measurement results of the ultrasonic sensor 12. Therefore, the observation matrix can be set offline before performing imaging. In the observation matrix shown in [Equation 12], the pixel size in the range direction is set to _wy .

The process shown in Figure 5 is the processing flow when the ADMM (Alternating Direction Method of Multipliers) is used as the restoration algorithm.

In the signal processing device 14, the data receiving unit 42 receives the signal (real number) from the ultrasonic sensor 12.
and down-converts the received signal to generate a received signal matrix using complex baseband signals (step S201).

The range processing unit 44 generates observation vectors by allocating data for each range r from the received signal matrix based on, for example, the number of pixels set by the user, and sets sparse vectors (step S202). The range processing unit 44 transfers the observation vector and sparse vector data to the second calculation unit 24 (step S203). In the second calculation unit 24, the restoration processing unit 46 performs restoration processing in parallel for multiple ranges r based on the transferred observation vector and sparse vector data and the observation matrix data stored in the memory unit 26 (step S204). The restoration processing unit 46 transfers the restored sparse vector data to the first calculation unit 22 (step S205). In the first calculation unit 22, the synthesis processing unit 48 generates an imaging matrix (imaging image) based on the restored sparse vector data (step S206), and then ends the processing.

As described above, the signal processing device 14 according to this embodiment includes a data receiving unit 42 that acquires observation data received at measurement positions when transmitted waves spreading in the range direction perpendicular to the azimuth direction are transmitted from multiple measurement positions in the azimuth direction, arranges the observation data in a matrix in the azimuth direction and the range direction, and generates a received signal matrix in which signal strength represents a position value; a range processing unit 44 that identifies the range direction range in the received signal matrix, generates an observation vector constructed based on the azimuth direction components of the identified range, and sets a sparse vector constructed from the azimuth direction components of the range direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the azimuth direction is greater than that of the received signal matrix and the number of divisions in the range direction is the same as that of the received signal matrix; a restoration processing unit 46 that performs restoration processing in parallel for multiple ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a synthesis processing unit 48 that synthesizes the sparse vectors calculated by the restoration processing to generate an imaging matrix.

The signal processing method according to this embodiment also includes a data reception step of acquiring observation data received at measurement positions when transmitting waves that spread in the range direction perpendicular to the azimuth direction from multiple measurement positions in the azimuth direction, organizing the observation data into matrices in the azimuth and range directions, and generating a received signal matrix in which signal strength represents a position value; a range processing step of identifying a range in the range direction in the received signal matrix, generating an observation vector based on the azimuth direction components of the identified range, and setting a sparse vector composed of the azimuth direction components of the range in the range direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the azimuth direction is greater than that of the received signal matrix and the number of divisions in the range direction is the same as that of the received signal matrix; a restoration processing step of performing restoration processing in parallel for multiple ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a combination processing step of combining the sparse vectors calculated by the restoration processing to generate an imaging matrix.

The signal processing program according to this embodiment also causes a computer to execute the following steps: a data reception step of acquiring observation data received at measurement positions when transmitting waves that spread in the range direction perpendicular to the azimuth direction from multiple measurement positions in the azimuth direction, organizing the observation data into matrices in the azimuth and range directions, and generating a received signal matrix with signal strength as a position value; a range processing step of identifying a range in the range direction in the received signal matrix, generating an observation vector based on the azimuth direction components of the identified range, and setting a sparse vector composed of the azimuth direction components of the range in the range direction corresponding to the observation vector in an imaging matrix with a larger number of divisions in the azimuth direction than the received signal matrix and the same number of divisions in the range direction as the received signal matrix; a restoration processing step of performing restoration processing in parallel for multiple ranges to calculate the components of the sparse vector based on the observation vector and the sparse vector; and a combination processing step of combining the sparse vectors calculated by the restoration processing to generate an imaging matrix.

With this configuration, by setting ranges for the matrix of an imaging image and performing compressed sensing restoration on a range-by-range basis, processing can be performed more quickly and accurately than when the entire image is restored at once. In this case, by performing restoration processing on multiple ranges in parallel, processing speed can be improved compared to when composite processing is performed serially for each range. This allows processing speed to be improved while maintaining resolution.

In the signal processing device 14 according to this embodiment, the data receiving unit 42, range processing unit 44, and synthesis processing unit 48 are provided in the first calculation unit 22, and the restoration processing unit 46 is provided in the second calculation unit 24, which is capable of performing more parallel calculations than the first calculation unit 22. Therefore, the restoration processing unit 46 can efficiently perform parallel processing for multiple ranges.

In the signal processing device 14 according to this embodiment, the first calculation unit 22 is a CPU, and the second calculation unit 24 is a GPU. Therefore, the second calculation unit 24, which is configured as a GPU, can efficiently perform parallel processing for multiple ranges.

In the signal processing device 14 according to this embodiment, the range processing unit 44 sets the number of divisions in at least one of the azimuth direction and the range direction based on information input from the outside. Therefore, it is possible to increase the allowable number of divisions that can be set by, for example, a user.

In the signal processing device 14 according to this embodiment, the range processing unit 44 sets a range corresponding to one row in the range direction of the imaging matrix. This reduces the amount of processing per range, thereby improving processing speed.

In the signal processing device 14 according to this embodiment, the observation data is data received at the measurement position in response to an emitted wave transmitted at the measurement position. Therefore, when using the so-called linear scan measurement method, it is possible to improve processing speed while maintaining resolution.

Figure 6 is an explanatory diagram illustrating another example of signal processing. As shown in Figure 6, the range processing unit 44 can set a range r corresponding to multiple rows in the range direction of the imaging matrix. The number of rows included in range r can be set, for example, by the user. Below, we will explain the case where the number of rows included in range r is K. Note that when K = 1, the situation is the same as in the above-mentioned embodiment.

In this case, the observation vector is
can be expanded as follows.

where:
Regarding each,
is.

Next, another example of processing in the signal processing device 14 will be described using Figures 7 and 8. Figures 7 and 8 are flowcharts showing another example of processing in the signal processing device 14. As shown in Figure 7, before measurements are performed using the ultrasonic sensor 12, an observation matrix is set in advance (step S301), and an observation matrix with K rows, which aggregates multiple rows (hereinafter referred to as K rows), is obtained (step S302). Then, the obtained observation matrix with K rows is stored in the memory unit 26 (step S303).

As shown in FIG. 8, in the signal processing device 14, the data receiving unit 42 acquires a received signal (real number) from the ultrasonic sensor 12 and down-converts the acquired received signal to generate a received signal matrix using complex baseband signals (step S401).

The range processing unit 44 generates observation vectors by allocating data for each range r from the received signal matrix, for example, based on the number of pixels set by the user, and sets sparse vectors (step S402). Based on the observation vectors and sparse vectors, the range processing unit 44 generates observation vectors and sparse vectors that aggregate K rows of data (step S403), and transfers the observation vectors and sparse vectors to the second calculation unit 24 (step S404).

In the second calculation unit 24, the restoration processing unit 46 performs restoration processing in parallel for multiple ranges r based on the transferred observation vector and sparse vector data, i.e., the observation vector and sparse vector data in which K rows of data for one range r are aggregated, and the observation matrix data for K rows stored in the memory unit 26 (step S405). The restoration processing unit 46 transfers the restored sparse vector data to the first calculation unit 22 (step S406). In the first calculation unit 22, the synthesis processing unit 48 generates an imaging matrix (imaging image) based on the restored sparse vector data (step S407), and ends the processing.

In this way, in the signal processing device 14 according to this embodiment, the range processing unit 44 sets ranges corresponding to multiple rows in the range direction of the imaging matrix. With this configuration, the correlation between pixel values of multiple rows aggregated in one range can be better reflected compared to when one range contains data for one row, thereby improving restoration accuracy.

FIG. 9 is a schematic diagram illustrating another measurement process of the measurement system according to this embodiment. As shown in FIG. 9 , the measurement system 10 is similar to the above embodiment in that multiple ultrasonic sensors 12 are arranged side by side in the azimuth direction x. In the measurement system 10, when a transmission signal is transmitted from one ultrasonic sensor 12 (#i, i = 1, 2, ..., M _T ) into the interior of the measurement object R, the signal reflected by a reflection source l inside the measurement object R is received as a reception signal by multiple ultrasonic sensors 12 (#j, j = 1, 2, ..., M _R ). In other words, in the example shown in FIG. 9 , the measurement system 10 detects the reflection source l inside the measurement object R, which is surrounded by the azimuth direction and the range direction perpendicular to the azimuth direction, using a so-called full matrix capture (FMC: synonymous with multistatic) measurement method.

The following describes a case where a signal transmitted from one ultrasonic sensor 12 is received by all ultrasonic sensors 12. In this case,
This becomes:

Here, the received signal η ⁱ (t) received by each ultrasonic sensor 12 is
In this case,
The correspondence between the sparse vector and the observation vector is
It is expressed as:

where:
indicates the phase difference when a transmission signal transmitted from the ultrasonic sensor 12 as a transmitter is reflected at the position of each pixel at range r and arrives at each ultrasonic sensor 12, and
The column vectors that make up this matrix are
It can be written as:

Here, the time τ _ilj represents the time required for a round trip from when a transmission signal is transmitted from the ultrasonic sensor 12 that is the transmission source, to when the signal is reflected from pixel #1 and received by each ultrasonic sensor 12. If the time required for the outbound journey is τ _il and the time required for the return journey is τ _lj , then
It can be written as:

In this way, in the FMC measurement method, the dimension of the received signal matrix is ^M2 , which is larger than that of the linear scan measurement method. For example, if the number of ultrasonic sensors 12 is 64, the dimension of the received signal matrix is ⁶⁴² = 4096.

In order to handle such a huge matrix, in this embodiment,
Multiply both sides of the observation vector by a known horizontal matrix, denoted by
This becomes:

So,
The vector indicated by is considered as a new observation vector,
Let us consider the matrix shown as a new observation matrix.

In this case, the observation vector is
In this way, even if the observation vector and observation matrix change, the sparse vector s(r) to be estimated is not affected. Therefore, the restoration processing unit 46 can perform restoration processing in parallel for multiple ranges r using the new observation vector shown in [Equation 29] and the new observation matrix shown in [Equation 30].

When multiple rows of data are aggregated in one range, the observation vector of range r shown in [Equation 31] is
can be expanded as follows.

where:
The observation vector and observation matrix shown in [Equation 32] and [Equation 33] have the same sizes as the observation vector and observation matrix shown in [Equation 18]. Therefore, even in the FMC measurement method, restoration processing can be performed with the same amount of calculation and processing time as in the linear scan measurement method.

Next, another example of processing in the signal processing device 14 will be described using Figures 10 and 11. Figures 10 and 11 are flowcharts showing another example of processing in the signal processing device 14. As shown in Figure 10, before performing measurement using the ultrasonic sensor 12, a widthwise matrix Φ (see [Equation 27]) is set in advance (step S501). The set widthwise matrix Φ is stored in the memory unit 26. An observation matrix is also set (step S502), and a new observation matrix is set by multiplying the set observation matrix by the widthwise matrix Φ (step S503). Furthermore, when multiple rows of data are aggregated in one range r, an observation matrix (shown in Equation 33) with K rows that aggregates multiple rows (hereinafter referred to as K rows) is obtained (step S504). When one range r contains one row of data, step S504 does not need to be performed. Thereafter, the set widthwise matrix Φ and the new observation matrix are stored in the memory unit 26 (step S505).

As shown in FIG. 11, in the signal processing device 14, the data receiving unit 42 acquires a received signal (real number) from the ultrasonic sensor 12 and down-converts the acquired received signal to generate a received signal matrix using complex baseband signals (step S601).

The range processing unit 44 generates an observation vector by allocating data for each range r from the received signal matrix, for example, based on the number of pixels set by the user, and sets a sparse vector (step S602). The range processing unit 44 calculates a new observation vector by multiplying the observation vector by the horizontally long matrix Φ stored in the memory unit 26 (step S603). Furthermore, when multiple rows of data are aggregated into one range r, an observation vector is generated by aggregating K rows of data from the calculated new observation vector, and a sparse vector is similarly generated by aggregating K rows of data (step S604). If one range r contains only one row of data, step S604 does not need to be performed. The range processing unit 44 transfers the observation vector and sparse vector to the second calculation unit 24 (step S605).

In the second calculation unit 24, the restoration processing unit 46 performs restoration processing in parallel for multiple ranges r based on the transferred observation vectors and sparse vectors and the observation matrix stored in the storage unit 26 (step S606). The restoration processing unit 46 transfers the restored sparse vector data to the first calculation unit 22 (step S607). In the first calculation unit 22, the synthesis processing unit 48 generates an imaging matrix (imaging image) based on the restored sparse vector data (step S608), and the processing ends.

In this way, in the signal processing device 14 according to this embodiment, the observation data is data received at all measurement positions for an emitted wave transmitted at one measurement position. With this configuration, the restoration process can be performed with the same amount of calculation and processing time as with the linear scan measurement method, even with the FMC measurement method.

The technical scope of the present invention is not limited to the above-described embodiment, and modifications can be made as appropriate without departing from the spirit of the present invention. For example, the above-described embodiment was described using an example configuration in which an ultrasonic sensor 12 is disposed at each of multiple measurement positions, but the present invention is not limited to this.

Figure 12 is a schematic diagram showing a measurement system 10A according to a modified example. As shown in Figure 12, the measurement system 10A includes a drive unit 20 in addition to the components of the measurement system 10 (see Figure 1) described above. The remaining components of the measurement system 10A are the same as those of the measurement system 10 described above. The drive unit 20 functions as a power source for moving the ultrasonic sensor 12. The specific configuration of the drive unit 20 depends on the operating mode of the ultrasonic sensor 12. For example, if the driving body is a vehicle that travels on land, the drive unit 20 includes multiple wheels and a prime mover that drives some or all of the multiple wheels. If the ultrasonic sensor 12 is a vehicle that travels through the air, the drive unit is a propeller and a drive source that drives the propeller. If the ultrasonic sensor 12 is a vehicle that travels underwater, the drive unit is a screw and a prime mover that drives the screw. The specific configuration of the drive unit 20 illustrated here is merely an example and is not limited to this. The drive unit 20 only needs to function as a power source that enables the ultrasonic sensor 12 to travel.

Figure 13 shows a measurement system 10A according to a modified example. The measurement system 10A shown in Figure 12 is configured to perform surrounding detection processing while moving the ultrasonic sensor 12 in the azimuth direction x using the drive unit 20. The ultrasonic sensor 12 moves in the azimuth direction x, transmitting a transmission signal to the measurement object R at each of positions #1, #2, ..., #i, ..., #M, and receiving a signal reflected by a reflection source l of the measurement object R as a received signal. The ultrasonic sensor 12 emits a transmission signal with a spread width of angle θ, and the emitted range is the search range. This configuration allows the number of ultrasonic sensors 12 to be reduced. Note that multiple ultrasonic sensors 12 may be moved in the azimuth direction x to perform multistatic measurement.

10, 10A Measurement system 12 Ultrasonic sensor 14 Signal processing device 16 Control unit 18, 26 Memory unit 20 Drive unit 22 First calculation unit 24 Second calculation unit 28 Pulsar receiver 32 Transmitter 34 Receiving unit 42 Data receiving unit 44 Range processing unit 46 Restoration processing unit 48 Synthesis processing unit 130 Received signal matrix 132 Imaging matrix 134a, 136a Pixel area

Claims (10)

a data receiving unit that acquires observation data received at a plurality of measurement positions when an emission wave spreading in a second direction perpendicular to the first direction is emitted from the measurement positions, arranges the observation data in a matrix in the first direction and the second direction, and generates a received signal matrix in which the signal strength is a position value;
a range processing unit that identifies a range in the second direction in the received signal matrix, generates an observation vector configured based on components in the first direction of the identified range, and sets a sparse vector configured of components in the first direction of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is larger than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix;
a restoration processing unit that performs restoration processing for calculating components of the sparse vector based on the observation vector and the sparse vector in parallel for a plurality of the ranges;
a synthesis processing unit that synthesizes the sparse vectors calculated by the restoration processing to generate the imaging matrix. the data receiving unit, the range processing unit, and the synthesis processing unit are provided in a first calculation unit,
The signal processing device according to claim 1 , wherein the restoration processing unit is provided in a second calculation unit that can execute more parallel calculations than the first calculation unit. the first calculation unit is a CPU,
The signal processing device according to claim 2 , wherein the second processing unit is a GPU. The signal processing device according to claim 1 , wherein the range processing unit sets the number of divisions in at least one of the first direction and the second direction based on information input from an external device. The signal processing device according to claim 1 , wherein the range processing unit sets the range corresponding to one row in the imaging matrix in the second direction. The signal processing device according to claim 1 , wherein the range processing unit sets the ranges corresponding to a plurality of rows in the imaging matrix in the second direction. The signal processing device according to claim 1 , wherein the observation data is data received at the measurement position in response to the transmitted wave transmitted at the measurement position. The signal processing device according to claim 1 , wherein the observation data is data received at a plurality of the measurement positions in response to the transmitted wave transmitted at one of the measurement positions. a data receiving step of acquiring observation data received at a plurality of measurement positions in a first direction when an emission wave spreading in a second direction perpendicular to the first direction is transmitted from the measurement positions, and forming a matrix of the observation data in the first direction and the second direction to generate a received signal matrix in which the signal strength is a position value;
a range processing step of specifying a range in the second direction in the received signal matrix, generating an observation vector configured based on the components in the first direction of the specified range, and setting a sparse vector configured of the components in the first direction of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is larger than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix;
a restoration processing step of performing restoration processing for calculating components of the sparse vector based on the observation vector and the sparse vector in parallel for a plurality of the ranges;
a synthesis processing step of synthesizing the sparse vectors calculated by the restoration processing to generate the imaging matrix. a data receiving step of acquiring observation data received at a plurality of measurement positions in a first direction when an emission wave spreading in a second direction perpendicular to the first direction is transmitted from the measurement positions, and forming a matrix of the observation data in the first direction and the second direction to generate a received signal matrix in which the signal strength is a position value;
a range processing step of specifying a range in the second direction in the received signal matrix, generating an observation vector configured based on the components in the first direction of the specified range, and setting a sparse vector configured of the components in the first direction of the range in the second direction corresponding to the observation vector in an imaging matrix in which the number of divisions in the first direction is larger than that of the received signal matrix and the number of divisions in the second direction is the same as that of the received signal matrix;
a restoration processing step of performing restoration processing for calculating components of the sparse vector based on the observation vector and the sparse vector in parallel for a plurality of the ranges;
a synthesis processing step of synthesizing the sparse vectors calculated by the restoration processing to generate the imaging matrix.