JPS5932743B2 - correlation tracking device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION This invention converts an image signal obtained from an imaging device such as a TV camera into an analog-to-digital converter, and performs correlation calculation processing on the digital image signal. This invention relates to a correlation tracking device that tracks a target.

Conventionally, in this type of device, in order to reduce the amount of data to be processed, image information having a two-dimensional distribution is compressed into one-dimensional images in two directions parallel and perpendicular to the scanning line, and these compressed one-dimensional images The density of the signal was further compressed into binary values, and correlation calculation processing was performed on the binary signals.

FIG. 1 shows the relationship between the screen, the reference window, and the target to be tracked.

In the figure, a reference window 3, a horizontal correlation window 4, and a vertical correlation window 5 are set surrounding a target (or scene) 2 in a screen 1. The signal of the pixel initially surrounded by the reference window 3 is used as a reference signal, and the cross-correlation function between this signal and the signal of the pixel currently surrounded by each of the horizontal correlation window 4 and the vertical correlation window 5 is determined. . Then, by detecting the maximum points of these cross-correlation functions and finding the maximum correlation position, the position of the target 2 is determined, and by setting a new reference window 3 at this position, the target 2 tracking is performed. FIG. 2 shows the configuration of a conventional correlation tracking device, in which an analog video signal obtained from an imaging device 6 such as a TV camera is converted into a digital video signal by an analog-to-digital converter 7, and a reference window is set. The signal is supplied to the circuit 8 and further to the correlation signal extraction circuit 9. Correlation signal extraction circuit 9 extracts video signals of pixels surrounded by horizontal correlation window 4 and vertical correlation window 5, and outputs horizontal image signal 10 and vertical image signal 11, respectively. Both of these image signals are supplied to adder circuits 12A and 12B, respectively, and one-dimensional compressed signals 13A and 1
3B, which is further supplied to the binarization circuits 14A, 14B and converted into a binary signal, and then the correlation calculation circuits 16A, 14B.
16B. In the correlation calculation circuits 16A and 16B, a binary signal compressed from pixels surrounded by an initially set reference window is recorded in a fixed memory as a reference signal, and this reference signal and each horizontal and vertical correlation window are recorded. The cross-correlation function between the compressed binary signal and the pixels surrounded by is determined. The correlation calculation result is sent to the correlation function value comparison circuit 17A.
, 17B, and by detecting the maximum point of the correlation function, the position coordinates of the image that most closely matches the image surrounded by the initially set reference window are determined, and the new position of the target is detected. These position coordinates are fed back to the reference window setting circuit 8 to set a reference window at the new position of the target, and at the same time are used as an error signal for the servo system that controls the attitude of the imaging device 6, so that the reference window is always set at the predetermined target or The scene is controlled so that it is centered on the screen. In the above circuit configuration, since an image signal having multilevel density information is compressed into a binary signal, it is not only cumbersome to change the threshold value for binarization depending on the situation, but also Due to the significant lack of concentration information, a clear maximum value of the correlation function could not be found, making tracking difficult.

In order to eliminate these drawbacks, the present invention aims to provide a correlation tracking device that performs correlation calculation processing on a digital image signal as it is multi-valued using a parallel calculation circuit, thereby making it possible to stably track a target or scene. It is something. Hereinafter, a detailed explanation will be given step by step. Since the principles of tracking in the horizontal and vertical directions are the same, tracking in the horizontal direction will be explained below. An n-bit one-dimensional compressed signal created by compressing the pixels surrounded by the initial reference window and the horizontal correlation window is used as the reference signal {R(1)11≦i≦M} and the correlated signal {S(
J) 11≦j≦N}. Here R(1), S(j)
is a binary value representing the density of a pixel in n bits, and M,
N is the number of pixels in the horizontal direction of the reference window and the horizontal correlation window, respectively, and M<N. In {S(j)}, {R(1)
}, the product correlation function Cp(τ) shown in equation (1) or the difference correlation function CD(τ) shown in equation (2) is determined. {However, τ is the shift amount (
unit pixel), and τ=0, 1, 2,...
It is N-M. }Maximum value of Cp(τ) or CD(τ
) is the value of τ that gives the minimum value of τ. If we find {S(j)
lτo≦j≦τo+M−1} is {R(1)11≦i≦M
}, so the center position of this pixel group is j−τ. Define a new horizontal center position of the reference window at +M/2. In the vertical direction, a new vertical center position of the reference window can be determined in the same manner as described above.

By performing the n-bit multilevel correlation calculation as described above, it becomes possible to track the target using the reference window.

FIG. 3 shows an embodiment of an arithmetic circuit configuration for performing multilevel correlation calculation according to the present invention, taking the horizontal direction as an example. In the figure, a one-dimensional compressed signal created from pixels surrounded by the initial reference window among the n-bit one-dimensional compressed signals 13A is selected by the selection circuit 18, stored in the storage circuit 19, becomes the reference signal, and is used as the current signal. A one-dimensional compressed signal created from pixels surrounded by the horizontal correlation window is selected by the selection circuit 18 and stored in the storage circuit 20 to become a correlated signal. 1 drive circuits 21 and 22 are connected to memory circuits 19 and 20, respectively.
Read the address and the contents of the (1+τ) address.

After these are temporarily stored in the storage circuits 23 and 24, respectively, in equation (1), the multi-bit reference signal and the correlated signal are used for the purpose of detecting the maximum value of the cross-correlation function as the maximum correlation point. An arithmetic circuit 2 that calculates the absolute value of the difference between the multi-bit reference signal and the correlated signal in order to calculate the multiplication of
5. All the parallel calculation outputs of the calculation circuit 25 calculated over all values of i are added together in the addition circuit 26, and Cp(τ) or CD(
τ) is required. This addition result 27 is supplied to a comparison circuit 28, and Cp(τ) or CD at various τ values is
The (τ) values are compared and the τ value that gives each of these maximum or minimum values is determined. The coordinate conversion circuit 29 calculates the horizontal coordinates of the next field reference window from this τ value. By using a similar configuration in the vertical direction, the vertical coordinates of the next field reference window are calculated.

In the above embodiment, a case has been described in which correlation calculation is performed by compressing image information having a two-dimensional distribution into one-dimensional images in mutually orthogonal directions. It can also be applied when performing calculations. As described above, in the correlation tracking device according to the present invention, by fixedly setting the dynamic range of the device to a value calculated from the expected maximum value of the input signal strength, it is possible to Not only is it not necessary to reset the threshold value depending on the situation, but also the density information of the pixels is effectively used, so it is possible to stably track a predetermined target or scene.

[Brief explanation of the drawings] Figure 1 is a layout diagram showing the relationship between the screen, reference window and the target to be tracked, Figure 2 is a configuration diagram showing the configuration of a conventional correlation tracking device, and Figure 3 is a diagram showing the relationship between the screen, reference window and the target to be tracked. FIG. 2 is a configuration diagram showing an example of the configuration of an arithmetic circuit for performing multilevel correlation calculation in the horizontal direction according to the invention. 1...Screen, 2...Goal, 3...
...Reference window, 4...Horizontal correlation window, 5...
・Vertical correlation window, 6... Imaging device, 7...
・Analog-digital converter, 8... Reference window setting circuit, 9... Correlation signal extraction circuit, 10...
...Horizontal image signal, 11... Vertical image signal, 12A, 12B... Addition circuit, 13A, 13
B... One-dimensional compressed signal, 14A, 14B...
... Binarization circuit, 15A, 15B ... Binary signal, 16A, 16B ... Correlation calculation circuit, 17A
, 17B...Correlation function value comparison circuit, 18...
...Selection circuit, 19,20...Memory circuit, 2
1, 22... Drive circuit, 23, 24...
・Memory circuit, 25... Arithmetic circuit, 26...
...Addition circuit, 27...Addition result, 28...
... Comparison circuit, 29 ... Coordinate conversion circuit.

Claims (1)

[Claims] 1. A digital image obtained by subdividing an image supplied from an imaging device into pixels in the horizontal scanning direction and vertical scanning direction, and converting the density of each pixel from analog to digital using multiple bits. A digital image in which an area including a desired target or scene within the field of view of the imaging device is stored in advance as a reference signal, and a digital image that is newly sequentially supplied from the imaging device as a correlated signal and digitized. In a correlation tracking device that calculates a correlation between A horizontal correlation window and a vertical correlation window, which are rectangular areas, are set for each scanning direction and vertical scanning direction, respectively, and the densities of pixels within the horizontal correlation window and vertical correlation window are added to obtain multi-bit density information. A storage circuit 1 obtains a one-dimensional signal having multi-bit density information of a reference window, which is a specific area including a desired target or scene, as a reference signal R(i).
9, and a one-dimensional signal having multi-bit density information obtained for an image supplied from the imaging device at a time different from the time stored in the storage circuit 19 is converted into a correlated signal S(
i+τ) in the memory circuit 20, and each memory circuit 19
, 20 and the correlated signal S
The product correlation function C_P(τ) or the difference correlation function C_D(τ) with (i + τ) is calculated by the calculation circuit as follows ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ ▲ There are mathematical formulas, chemical formulas, tables, etc. In the case of horizontal correlation calculation, M is the number of pixels in the horizontal scanning direction of the reference window, τ is the shift amount (unit pixel), and τ = 0, 1, 2, ..., N - M, N is the horizontal correlation In the case of vertical correlation calculation, M indicates the number of pixels in the vertical scanning direction of the reference window, and N indicates the number of pixels in the vertical direction of the vertical correlation window. A correlation tracking device characterized in that it performs calculations according to the following.