METHOD OF TRACKING OBJECTS Russian patent published in 2023 - IPC G06T7/269 

FIELD: tracking terrestrial objects.

SUBSTANCE: methods for tracking terrestrial objects using a digitized television image signal. Operator points the window M at the tracking object on the screen S of the television image. The zone of analysis Z of the television image is set, the dimensions of which are less than or equal to the dimensions of the screen S, but greater than the dimensions of the window M with the tracking object. The coordinates of the centre of the window M with the tracking object are taken as the coordinates of the centre of the analysis zone Z. The digitized television image is pre-processed by means of low-frequency filtering. At each point of the zone of analysis Z, the correlation function F(α,β) of the brightness of the image points in the window M with the tracking object and the brightness of the points of the analysed image in the windows N, the dimensions of which coincide with the dimensions of the window M, is calculated. The correlation function F(α,β) is calculated either as a negative correlation function or as a regular correlation function. The resulting correlation function F(α,β) is processed by a differential operator, obtaining a gradient image. After that, the sum of the brightness gradients G_e of the image points in the window M with the tracking object is calculated. By multiplying this sum of brightness gradients G by a coefficient k less than 1, the threshold value P is obtained. At all points of the gradient image located in the zone of analysis Z, the sums of brightness gradients G_i,j are calculated for windows N of the analysed images, the sizes of which coincide with the size of the window M with the tracking object. The coordinates (i_n, j_n) of the centres of windows N are stored, in which the sums of the brightness gradients G_i,j are greater than the threshold P. After that, in the window M with the tracking object and in each window N of the analysed image, the coordinates of the centre of which coincide with the stored coordinates (i_n, j_n), criteria A_n, B_n, C_n and criterion K_n=A_n ⋅B_n ⋅C_n are calculated, where A_n is the criterion for the difference between the sum of the brightness of the image points in the window N of the analysed image from the sum of the brightness of the image points in the window M with the tracking object; B_n is the criterion for the difference between the sum of brightness gradients G_i,j of points of the gradient image in window N of the analysed image and the sum of brightness gradients G_e of points of the gradient image in window M with the tracking object, C_n is the criterion for the distance of the centre of window N of the analysed image from the centre of the zone of analysis Z. For the coordinates of the centre of the new window with the tracking object the coordinates of the centre of the window, in which the value of the criterion K_n has the highest value, are taken. Centre of the zone of analysis Z is aligned with the centre of the new window with the tracking object. After that, the image is updated in the window with the tracking object by means of exponential filtering. Then processing of the next frame of the television image is commenced.

EFFECT: increase in reliability of object tracking, a decrease in probability of its loss.

10 cl, 3 dwg