JP7323235B2 - Image tracking device, image tracking method, and program - Google Patents

Description

The present invention relates to an image tracking device and an image tracking method for tracking an object such as a person on an image from a camera, and further to a program for realizing these.

In recent years, surveillance systems have added the function of tracking a specific person on the video from a surveillance camera. For example, Patent Literature 1 discloses a surveillance system with such a person tracking function added. The surveillance system disclosed in Patent Literature 1 extracts a feature amount of an area expected to be a person for each frame output from a surveillance camera in chronological order, and extracts the extracted feature amount between successive frames. Tracking a person being tracked by determining identity. According to the monitoring system disclosed in Patent Literature 1, the behavior of a specific person can be monitored, so the administrator can easily identify the suspicious person.

Non-Patent Document 1 discloses a surveillance system that tracks a specific person using a pre-learned neural network. In the monitoring system disclosed in Non-Patent Document 1, human images are used as learning data for neural network learning. Therefore, if the monitoring system disclosed in Non-Patent Document 1 is used, tracking accuracy can be improved more than the monitoring system disclosed in Patent Document 1.

JP 2009-098774 A

Zhedong Zheng, Xiaodong Yang, Zhiding Yu, Liang Zheng, Yi Yang, Jan Kautz1, ``Joint Discriminative and Generative Learning for Person Re-identification'', 22 May 2019, CVPR2019

By the way, it is generally difficult to increase the frame rate of surveillance cameras used in surveillance systems, and the frame rate is set to about 3 fps to 5 fps. This is because if the frame rate is increased, the size of the image data to be transmitted increases, exceeding the capacity of the communication line. Therefore, surveillance systems need to track people under low frame rate conditions.

However, in such a situation where the frame rate is low, the movement distance of a person between frames becomes large, so the monitoring system disclosed in Patent Document 1 suffers from a problem of reduced tracking accuracy. On the other hand, in the monitoring system disclosed in Non-Patent Document 1, it is considered that deterioration in tracking accuracy is suppressed compared to the monitoring system disclosed in Patent Document 1, but this monitoring system requires a large amount of learning in advance. There is a problem that it is not possible to track a person who has not been machine-learned using data.

One example of the object of the present invention is to solve the above problems, and to track an object on the video from the camera without the need for prior machine learning using learning data and without being affected by the frame rate. , an image tracking device, an image tracking method, and a program .

To achieve the above object, an image tracking device according to one aspect of the present invention includes:
an image data acquisition unit that acquires image data output in units of frames in chronological order from a camera;
an object detection unit that detects an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation unit that obtains, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination unit that determines identity with the object detected in the output image data;
an object tracking unit that tracks the detected object based on the identity determination result;
characterized by comprising

In order to achieve the above object, an image tracking method according to one aspect of the present invention includes:
an image data acquisition step of acquiring image data output frame by frame in chronological order from a camera;
an object detection step of detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation step of obtaining, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination step of determining identity with the object detected in the output image data;
an object tracking step of tracking the detected object based on the identity determination result;
characterized by having

Furthermore, in order to achieve the above object, the program in one aspect of the present invention is
to the computer,
an image data acquisition step of acquiring image data output frame by frame in chronological order from a camera;
an object detection step of detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation step of obtaining, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination step of determining identity with the object detected in the output image data;
an object tracking step of tracking the detected object based on the identity determination result;
The program that causes the to run .

As described above, according to the present invention, an object can be tracked on an image from a camera without prior machine learning using learning data and without being affected by the frame rate.

FIG. 1 is a configuration diagram showing a schematic configuration of an image tracking device according to an embodiment of the present invention. FIG. 2 is a configuration diagram specifically showing the configuration of the image tracking device according to the embodiment of the present invention. FIG. 3 is a diagram showing an example of image data acquired in the embodiment and an example of pairs set in the embodiment. FIG. 4 is a flow chart showing the operation of the image tracking device according to the embodiment of the invention. FIG. 5 is a block diagram showing an example of a computer that realizes the image tracking device according to the embodiment of the invention.

(Embodiment)
An image tracking device, an image tracking method, and a program according to embodiments will be described below with reference to FIGS. 1 to 5. FIG.

[Device configuration]
First, the schematic configuration of the image tracking device according to the embodiment will be described with reference to FIG. FIG. 1 is a configuration diagram showing a schematic configuration of an image tracking device according to an embodiment of the present invention.

An image tracking device 100 in the embodiment shown in FIG. 1 is a device that tracks an object such as a person on an image from a camera. As shown in FIG. 1 , the image tracking device 100 includes an image data acquisition unit 101 , an object detection unit 102 , a variance calculation unit 103 , an identity determination unit 104 and an object tracking unit 105 .

In this configuration, the image data acquisition unit 101 acquires image data output from the camera in units of frames in time series. The object detection unit 102 detects an object for each image data.

First, for each image data, for each object detected in the image data, the variance calculation unit 103 finds a pair of the object and any object detected in the image data output immediately after the image data. set. Subsequently, for each pair of image data, the variance calculation unit 103 calculates, in the feature space, the first intra-class variance whose class is the pair, and the second intra-class variance whose class is the object other than the pair in the feature space. Find the variance and between-class variance.

The identity determination unit 104 uses the first intra-class variance, the second intra-class variance, and the inter-class variance obtained for each pair to determine the object detected in each image data and the image data. The identity of the object detected in the image data output immediately after is determined. The object tracking unit 105 tracks the detected object based on the identity determination result.

Thus, in the embodiment, object identity is determined by intra-class variance between successive frames. Therefore, according to the embodiment, an object can be accurately tracked on the image from the camera without being affected by the frame rate. Also, in embodiments, there is no need for prior machine learning using learning data when tracking an object.

Next, the configuration and functions of the image tracking device 100 according to the embodiment will be specifically described with reference to FIGS. 2 and 3. FIG. FIG. 2 is a configuration diagram specifically showing the configuration of the image tracking device according to the embodiment of the present invention. FIG. 3 is a diagram showing an example of image data acquired in the embodiment and an example of pairs set in the embodiment.

As shown in FIG. 2, in the embodiment, image tracking device 100 is connected to camera 200 and display device 201 and functions as an image monitoring system. Furthermore, as shown in FIG. 2, the image tracking device 100 includes an image data acquisition unit 101, an object detection unit 102, a variance calculation unit 103, an identity determination unit 104, and an object tracking unit 105. A storage unit 106 and a monitoring unit 107 are further provided.

The camera 200 is a surveillance camera, and outputs image data obtained by photographing a surveillance area in units of frames in chronological order. Image data output from the camera 200 is stored in the image storage unit 106 .

In the example of FIG. 3, image data 10 and image data 20 output in units of frames are shown in chronological order. Each image data is image data of a monitored area. Arrows in FIG. 3 indicate time series, and each image data is output from the camera 200 in the order of image data 10 and image data 20 .

In the embodiment, the image data acquisition unit 101 acquires image data for a set period to be processed at once from the image storage unit 106 in chronological order. In the example of FIG. 2, the image data acquisition unit 101 acquires the image data 10 and the image data 20 from the image storage unit 106 in order.

In the embodiment, the object detection unit 102 scans an image of each image data output from the camera 200, identifies an area having a feature value indicating a person, and detects the identified area as an object (person). do. Specifically, the object detection unit 102 detects a person as an object using existing object detection algorithms such as YOLO and SSD. As a feature quantity representing a person, for example, there is a deep feature obtained by performing machine learning using an image of a person as training data. Other specific feature amounts include HoG (Histograms of Oriented Gradients) feature amounts, color histograms, and the like.

In the example of FIG. 3 , the object detection unit 102 detects objects 11 , 12 and 13 from the image data 10 and detects objects 21 , 22 , 23 and 24 from the image data 20 .

In the example of FIG. 3, for example, for the image data 10, the variance calculation unit 103 calculates the values of the objects 21 to 24 detected in the next frame of the image data 20 for each of the objects 11, 12, and 13. Pair with either. Specifically, the variance calculation unit 103 sets the following pairs for the object 11 of the image data 10, for example. The pairs set in this way show the hypothesis that they are the same object.
(Object 11, Object 21)
(Object 11, Object 22)
(Object 11, Object 23)
(object 11, object 24)

Subsequently, the variance calculation unit 103 sets an N-dimensional feature space according to the number N of dimensions of the feature amount used for object detection (N: any natural number). In the example of FIG. 3, a pair class combining the object 11 and any of the objects 21 to 24 described above, and a class of objects other than the pair, are shown in the feature space.

Then, in the feature space, using the feature amount of each object, the variance calculation unit 103 calculates, for each pair, the first intra-class variance in which the paired objects are in the same class, and the objects that do not form the pair are in the same class. A second within-class variance that Variance calculator 103 also obtains the inter-class variance from the first intra-class variance and the second intra-class variance for each pair.

Specifically, let X _i be a set of patterns (object feature values) belonging to class w _i , n i be the number of patterns included in Xi, m _i be the average vector, n be the total number of patterns, and n _be the total number of patterns. Let m be a vector. In addition, the class of paired objects is set to i=1, and the class of objects other than pairs is set to i=2. In this case, the first intra-class variance σ ² _w1 , the second intra-class variance σ ² _w2 , and the inter-class variance σ ² _B for one object in certain image data are calculated by the following equations 1 to 3. be done.

In the embodiment, the identity determination unit 104 first uses the first intra-class variance, the second intra-class variance, and the inter-class variance for each object detected in the image data for each image data. Then, an index of identity is calculated for each pair containing the object. In the embodiments, an indicator of identity includes Fisher's ratio J _σ . The Fisher ratio J _σ for one object in certain image data is calculated by Equation 4 below.

Subsequently, the identity determination unit 104 selects a pair in which the value of the calculated Fisher ratio is equal to or greater than the threshold value for each image data, and an object detected in the image data, which constitutes the selected pair, and An object detected in image data output immediately after that image data is determined to be the same.

In the embodiment, the object tracking unit 105 tracks the object in the subsequent frame in the pair determined to be identical by the identity determination unit 104, specifically, in the pair whose Fisher ratio value is equal to or greater than the threshold. Set a flag for each object of interest and tracked. If there are a plurality of pairs whose Fisher ratios are equal to or greater than the threshold, the object tracking unit 105 selects the pair with the highest Fisher ratio as a tracking target.

In the example of FIG. 3, the object tracking unit 105 sets the pair of the object 11 detected in the image data 10 and the object 21 detected in the image data 20 as the tracking target. Further, the object tracking unit 105 can determine that an object forming a pair whose Fisher ratio is less than the threshold has newly entered the frame of the camera 200 .

The monitoring unit 107 extracts the image data stored in the image storage unit 106 in chronological order and displays it on the display screen of the display device 201 . In addition, the object tracked by the object tracking unit 105 is represented on the display screen by visual expression indicating that it is being tracked, for example, addition of a symbol, display of a frame, change of color, and the like. The monitoring unit 107 can also output an alarm when a tracked object enters a specific area.

[Device operation]
Next, operation of the image tracking device 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a flow chart showing the operation of the image tracking device according to the embodiment of the invention. 1 to 3 will be referred to as necessary in the following description. Also, in the embodiment, the image tracking method is implemented by operating the image tracking device 100 . Therefore, the description of the image tracking method in the embodiment is replaced with the description of the operation of the image tracking device 100 below.

As shown in FIG. 4, first, the image data acquisition unit 101 acquires image data for a set period stored in the image storage unit 106 (step A1). The set period is set to, for example, 5 to 10 seconds.

Next, for each image data acquired in step A1, the object detection unit 102 identifies an area having a feature amount indicating a person on the image, and detects the identified area as an object (person) (step A2 ).

Next, the variance calculation unit 103 selects the image data output by the camera at the earliest time from among the image data acquired in step A1 and not yet selected in step A3. (Step A3).

Next, in the image data selected in step A3, the variance calculation unit 103 pairs each object detected in step A2 with each object detected in the image data of the next frame (step A4).

Next, variance calculator 103 selects one of the pairs set in step A4 (step A5). Then, the variance calculation unit 103 uses the feature amount of each object to calculate the first intra-class variance, the second intra-class variance, and the inter-class variance for the pair selected in step A5 (step A6).

Next, variance calculation section 103 determines whether or not there is a pair that has not yet been selected (step A7). As a result of the determination in step A7, if there is a pair that has not been selected yet, the variance calculation unit 103 executes step A5 again.

On the other hand, as a result of the determination in step A7, if there are no pairs that have not been selected yet, the identity determination unit 104 calculates the first intra-class variance calculated in step A6 for each pair, and the second Using the within-class variance and the between-class variance of , an index indicating identity is calculated. The identity determination unit 104 then determines identity for each pair (step A8).

Next, identity determination unit 104 determines whether or not there is image data that has not been selected in step A3 among the image data acquired in step A1 (step A9).

As a result of the determination in step A9, if there is image data that has not been selected in step A3 among the image data acquired in step A1, the variance calculation unit 103 executes step A3 again.

On the other hand, if the result of determination in step A9 is that there is no image data that has not been selected in step A3 among the image data acquired in step A1, the object tracking unit 105 determines identity in step A8. In the pair determined to be high, the object in the later frame is set as the tracking target (step A10).

Next, the monitoring unit 107 retrieves the image data stored in the image storage unit 106 in chronological order and displays it on the display screen of the display device 201. Then, an image expression indicating that the object is being tracked is performed (step A11).

The above steps A1 to A11 are repeatedly executed as long as the monitoring area is monitored.

[Effects of Embodiment]
As described above, in the embodiment, identity of an object is determined by intra-class variance between successive frames. Therefore, according to the embodiment, even when the frame rate drops, it is possible to suppress the decrease in tracking accuracy, and furthermore, it is possible to track an object that has not undergone machine learning using a large amount of learning data in advance. .

[program]
The program in the embodiment may be any program that causes a computer to execute steps A1 to A11 shown in FIG. By installing this program in a computer and executing it, the image tracking device 100 and the image tracking method in this embodiment can be realized. In this case, the computer processor functions as an image data acquisition unit 101, an object detection unit 102, a variance calculation unit 103, an identity determination unit 104, an object tracking unit 105, and a monitoring unit 107, and performs processing.

In the embodiment, the image storage unit 106 may be realized by storing the data files constituting these in a storage device such as a hard disk provided in the computer, or may be realized by storing the data files in a storage device of another computer. It may be realized by

Moreover, as a computer, a smart phone and a tablet-type terminal device are mentioned other than general-purpose PC. Furthermore, it may be a computer built into a camera such as a surveillance camera.

Also, the programs in the embodiments may be executed by a computer system constructed by a plurality of computers. In this case, for example, each computer functions as one of the image data acquisition unit 101, the object detection unit 102, the variance calculation unit 103, the identity determination unit 104, the object tracking unit 105, and the monitoring unit 107. Also good.

[Physical configuration]
Here, a computer that realizes the image tracking device 100 by executing the program according to the embodiment will be described with reference to FIG. FIG. 5 is a block diagram showing an example of a computer that realizes the image tracking device according to the embodiment of the invention.

As shown in FIG. 5, the computer 110 includes a CPU (Central Processing Unit) 111, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader/writer 116, and a communication interface 117. and These units are connected to each other via a bus 121 so as to be able to communicate with each other.

Further, the computer 110 may include a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to the CPU 111 or instead of the CPU 111 . In this aspect, a GPU or FPGA can execute the programs in the embodiments.

The CPU 111 develops the program in the embodiment, which is composed of a code group stored in the storage device 113, in the main memory 112 and executes each code in a predetermined order to perform various operations. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory).

Also, the program in the embodiment is provided in a state stored in computer-readable recording medium 120 . It should be noted that the program in this embodiment may be distributed on the Internet connected via communication interface 117 .

Further, as a specific example of the storage device 113, in addition to a hard disk drive, a semiconductor storage device such as a flash memory can be cited. Input interface 114 mediates data transmission between CPU 111 and input devices 118 such as a keyboard and mouse. The display controller 115 is connected to the display device 119 and controls display on the display device 119 .

Data reader/writer 116 mediates data transmission between CPU 111 and recording medium 120 , reads programs from recording medium 120 , and writes processing results in computer 110 to recording medium 120 . Communication interface 117 mediates data transmission between CPU 111 and other computers.

Specific examples of the recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital); magnetic recording media such as flexible disks; An optical recording medium such as a ROM (Compact Disk Read Only Memory) can be used.

Note that the image tracking device 100 according to the present embodiment can also be realized by using hardware corresponding to each part instead of a computer in which a program is installed. Furthermore, the image tracking device 100 may be partly implemented by a program and the rest by hardware.

Part or all of the above-described embodiments can be expressed by (Appendix 1) to (Appendix 9) described below, but are not limited to the following descriptions.

(Appendix 1)
an image data acquisition unit that acquires image data output in units of frames in chronological order from a camera;
an object detection unit that detects an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation unit that obtains, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination unit that determines identity with the object detected in the output image data;
an object tracking unit that tracks the detected object based on the identity determination result;
An image tracking device characterized by comprising:

(Appendix 2)
1. The image tracking device of Claim 1,
The identity determination unit
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object calculating an index indicating identity, further selecting a pair in which the value of the calculated index is equal to or greater than a threshold, and forming the selected pair, the object detected in the image data and the image data Determining that the object detected in the image data output immediately after is the same,
An image tracking device characterized by:

(Appendix 3)
The image tracking device according to Appendix 2,
The identity determination unit
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculating the Fisher ratio as the indicator of identity,
An image tracking device characterized by:

(Appendix 4)
an image data acquisition step of acquiring image data output frame by frame in chronological order from a camera;
an object detection step of detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation step of obtaining, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination step of determining identity with the object detected in the output image data;
an object tracking step of tracking the detected object based on the identity determination result;
An image tracking method characterized by comprising:

(Appendix 5)
The image tracking method according to Appendix 4,
In the identity determination step,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object calculating an index indicating identity, further selecting a pair in which the value of the calculated index is equal to or greater than a threshold, and forming the selected pair, the object detected in the image data and the image data Determining that the object detected in the image data output immediately after is the same,
An image tracking method characterized by:

(Appendix 6)
The image tracking method according to Appendix 5,
In the identity determination step,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculating the Fisher ratio as the indicator of identity,
An image tracking method characterized by:

(Appendix 7)
to the computer,
an image data acquisition step of acquiring image data output frame by frame in chronological order from a camera;
an object detection step of detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For each pair, a variance calculation step of obtaining, in a feature space, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; an identity determination step of determining identity with the object detected in the output image data;
an object tracking step of tracking the detected object based on the identity determination result;
The program that causes the to run .

(Appendix 8)
The program according to Supplementary Note 7,
In the identity determination step,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object calculating an index indicating identity, further selecting a pair in which the value of the calculated index is equal to or greater than a threshold, and forming the selected pair, the object detected in the image data and the image data Determining that the object detected in the image data output immediately after is the same,
A program characterized by

(Appendix 9)
The program according to Appendix 8,
In the identity determination step,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculating the Fisher ratio as the indicator of identity,
A program characterized by

Although the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

This application claims priority based on Japanese Patent Application No. 2020-54679 filed on March 25, 2020, and the entire disclosure thereof is incorporated herein.

As described above, according to the present invention, an object can be tracked on an image from a camera without prior machine learning using learning data and without being affected by the frame rate. INDUSTRIAL APPLICABILITY The present invention is useful for a monitoring system that monitors human intrusion based on video from a monitoring camera.

REFERENCE SIGNS LIST 10 image data 11, 12, 13 object 20 image data 21, 22, 23, 24 object 100 image tracking device 101 image data acquisition unit 102 object detection unit 103 variance calculation unit 104 identity determination unit 105 object tracking unit 106 image storage unit 107 monitoring unit 110 computer 111 CPU
112 main memory 113 storage device 114 input interface 115 display controller 116 data reader/writer 117 communication interface 118 input device 119 display device 120 recording medium 121 bus 200 camera 201 display device

Claims (9)

image data acquisition means for acquiring image data output frame by frame in time series from a camera;
an object detection means for detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; Variance calculation means for obtaining, for each pair, a first intra-class variance having the pair as a class, a second intra-class variance having the object other than the pair as a class, and an inter-class variance in the feature space. ,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; identity determination means for determining identity with the object detected in the output image data;
object tracking means for tracking the detected object based on the identity determination result;
An image tracking device characterized by comprising: The image tracking device of claim 1, comprising:
The identity determination means is
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object calculating an index indicating identity, further selecting a pair in which the value of the calculated index is equal to or greater than a threshold, and forming the selected pair, the object detected in the image data and the image data Determining that the object detected in the image data output immediately after is the same,
An image tracking device characterized by: The image tracking device of claim 2, wherein
The identity determination means is
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculating the Fisher ratio as the indicator of identity,
An image tracking device characterized by: Acquire the image data output frame by frame along the time series from the camera,
detecting an object for each image data;
For each of the image data, for each of the objects detected in the image data, a pair of the object and any of the objects detected in the image data output immediately after the image data is set; For a pair, in the feature space, obtain a first intra-class variance with the pair as a class, a second intra-class variance with the object other than the pair as a class, and an inter-class variance,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; determining the identity of the object detected in the output image data;
tracking the detected object based on the identity determination result;
An image tracking method characterized by: The image tracking method of claim 4,
In determining the identity,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object calculating an index indicating identity, further selecting a pair in which the value of the calculated index is equal to or greater than a threshold, and forming the selected pair, the object detected in the image data and the image data Determining that the object detected in the image data output immediately after is the same,
An image tracking method characterized by: An image tracking method according to claim 5, wherein
In determining the identity,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculating the Fisher ratio as the indicator of identity,
An image tracking method characterized by: to the computer,
Get the image data output frame by frame along the time series from the camera,
detecting an object for each image data;
for each of the image data, for each of the objects detected in the image data, setting a pair of the object and any of the objects detected in the image data output immediately after the image data; For a pair, obtain a first intra-class variance with the pair as a class, a second intra-class variance with the object other than the pair as a class, and an inter-class variance in the feature space,
using the first within-class variance, the second within-class variance, and the between-class variance determined for each pair, the object detected in the image data and immediately after the image data; determining the identity of the object detected in the output image data;
performing tracking processing of the detected object based on the identity determination result;
program . The program according to claim 7,
In determining the identity ,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculate an index indicating identity, further select a pair whose calculated index value is equal to or greater than a threshold value, and configure the selected pair, the object detected in the image data, and the image data determining that the object detected in the image data output immediately after is the same;
A program characterized by The program according to claim 8,
In determining the identity ,
For each of the image data, for each of the objects detected in the image data, using the first within-class variance, the second within-class variance, and the between-class variance, in each pair containing the object Calculate the Fisher ratio as the indicator of identity,
A program characterized by

Images