Infrared target tracking method and system based on global radial gradient features

Abstract

This invention discloses a method and system for infrared target tracking based on global radial gradient features. This method calculates normalized radial gradient features in multiple directions to highlight distinct contour areas within infrared targets. This method, using a global feature description method weighted by multiple directional gradients, can adapt to the lack of internal texture information and target size variations in infrared images, accurately representing the contour appearance of infrared targets. The proposed infrared target tracking method boasts globality, high precision, and scale adaptability, and is applicable to target tracking tasks in both infrared and visible light images.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an infrared target tracking method and system based on global radial gradient features. Background Art

[0002] Existing target detection and tracking systems primarily include infrared, visible light, and radar. Radar detection requires the active emission of electromagnetic waves, which can easily expose the system and be affected by electromagnetic interference, rendering it inoperable. Furthermore, the stealth technology used by low-altitude targets reduces the radar's scattering area, significantly reducing its detection efficiency. Both infrared and visible light are passive optical imaging systems. However, visible light imaging is significantly affected by weather factors and is not suitable for imaging applications in rainy, cloudy, foggy, or nighttime conditions. Infrared imaging, with its advantages of strong smoke penetration and all-weather operation, is widely used in target monitoring and early warning systems, security surveillance, and other areas within both the military and civilian sectors.

[0003] Infrared target tracking is the technology for detecting the position and size of targets in real-time within infrared image sequences. Compared to visible light images, infrared images are visually blurred, with less discernible texture information within the target and a relatively distinct contour between the target and the background. Existing infrared target tracking methods suffer from a large number of invalid regions with small gradients. Therefore, it is necessary to design feature description methods tailored to the imaging characteristics of targets in infrared images, emphasizing the description of infrared target contours while downplaying the representation of textureless regions within the target. Furthermore, the size variation of infrared targets has always been a challenging issue in the tracking field. The Kernel Correlation Filter (KCF) algorithm cannot use target size variation, while the scale-adaptive Discriminative Scale Space Tracker (DSST) tracking method uses a discrete scale space to estimate target size variation. However, discrete scale estimation methods have limited computational accuracy. Summary of the Invention

[0004] The present invention provides an infrared target tracking method and system based on global radial gradient features, which are used to overcome the defects of the prior art such as a large number of invalid areas with small gradients and low precision.

[0005] To achieve the above object, the present invention proposes an infrared target tracking method based on global radial gradient features, comprising:

[0006] Acquire infrared image sequence set;

[0007] Determine the target search area and the radial length of the WRG feature in the infrared image at the next moment according to the target frame of the infrared image at the current moment in the infrared image sequence;

[0008] Selecting four directions within the target search area and calculating the gradient maps of the four directions within the target search area using a differential algorithm;

[0009] According to the gradient maps in the four directions, the gradient maps in the eight sub-directions are expanded. Within the radial length of the WRG feature, the positions of the extreme values of the pixel gradients in the eight sub-directions are marked. The gradient maps in the eight sub-directions within the target search area are normalized. The normalized gradient maps are combined and combined with the grayscale values of the pixels to obtain the WRG feature.

[0010] According to the WRG features, the KCF tracking algorithm is used to calculate the position of the target in the infrared image at the next moment;

[0011] Update the target size according to the position of the extreme pixel gradient in the 8 sub-directions;

[0012] Based on the updated target size, the target search area is redefined and the tracker is updated.

[0013] To achieve the above object, the present invention further proposes an infrared target tracking system based on global radial gradient features, comprising:

[0014] An image acquisition module, used for acquiring a set of infrared image sequences;

[0015] The infrared target tracking module is used to determine the target search area and the radial length of the WRG feature in the infrared image at the next moment based on the target frame of the infrared image at the current moment in the infrared image sequence set; select 4 directions in the target search area, and use the differential algorithm to calculate the gradient map of the 4 directions in the target search area; based on the gradient map of the 4 directions, expand the gradient map of 8 sub-directions, mark the positions of the pixel gradient extreme values in the 8 sub-directions within the radial length of the WRG feature, and normalize the gradient maps in the 8 sub-directions in the target search area; combine the normalized gradient maps and combine them with the grayscale value of the pixel to obtain the WRG feature; based on the WRG feature, use the KCF tracking algorithm to calculate the position of the target in the infrared image at the next moment; update the target size based on the positions of the pixel gradient extreme values in the 8 sub-directions; based on the updated target size, redefine the target search area and update the tracker.

[0016] To achieve the above object, the present invention further proposes a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0017] To achieve the above object, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] The infrared target tracking method based on global radial gradient features proposed in this paper calculates normalized radial gradient features in multiple directions to highlight the contour areas of infrared targets with distinct features. This method, using a global feature description method weighted by multiple directional gradients, can adapt to the lack of internal texture information and target size changes in infrared images, accurately representing the appearance of infrared target contours. The proposed infrared target tracking method boasts advantages such as globality, high precision, and scale adaptability, and is applicable to target tracking tasks in both infrared and visible light images. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0021] Figure 1 Flowchart of the infrared target tracking method based on global radial gradient features provided by the present invention;

[0022] Figure 2 This is a schematic diagram of WRG feature calculation for infrared target tracking in the infrared target tracking method based on global radial gradient features provided by the present invention.

[0023] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0025] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in this field can implement it. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0026] The present invention proposes an infrared target tracking method based on global radial gradient features, such as Figure 1Shown, including:

[0027] 101: Acquire an infrared image sequence set.

[0028] 102: Determine a target search area and a radial length of a WRG feature in an infrared image at a next moment according to a target frame of the infrared image at a current moment in the infrared image sequence set.

[0029] 103: Select four directions in the target search area and use the difference algorithm to calculate the gradient maps of the four directions in the target search area.

[0030] 104: Based on the gradient maps in four directions, eight sub-direction gradient maps are expanded. Within the radial length of the WRG feature, the locations of the pixel gradient extremes in the eight sub-directions are marked. The gradient maps in the eight sub-directions within the target search area are normalized. The normalized gradient maps are combined and combined with the pixel grayscale values to obtain the WRG feature. 105: Based on the WRG feature, the KCF tracking algorithm is used to calculate the target position in the infrared image at the next moment.

[0031] 106: Update the target size according to the position of the extreme pixel gradient in the 8 sub-directions.

[0032] 107: Based on the updated target size, redefine the target search area and update the tracker.

[0033] In one embodiment, for step 102, determining the target search area and the radial length of the WRG feature in the infrared image at the next moment according to the target frame of the infrared image at the current moment in the infrared image sequence includes:

[0034] Assume that the center position of the target frame of the infrared image at the current moment is (x t ,y t ), size (w t ,h t ), then the target search area in the infrared image at the next moment is

[0035]

[0036] Where padding represents the expansion factor of the search window relative to the target size, and round is the rounding function;

[0037] According to the target frame of the infrared image at the current moment (x t ,y t ,w t ,h t ) and the size of the tracker (w f ,h f), zoom the target search area in the infrared image at the next moment,

[0038]

[0039] Where, Represents the target search area after scaling, resize() is the image size scaling function, and various interpolation methods such as nearest neighbor, bilinear, and bicubic can be used. Indicates the image size scaling ratio,

[0040] Calculate the radial length L of the WRG feature,

[0041]

[0042] In the next embodiment, for step 102, four directions are selected within the target search area, and gradient maps of the four directions within the target search area are calculated using a differential algorithm, including:

[0043] Select four directions in the target search area: horizontal, vertical, 45-degree diagonal, and 135-degree diagonal; select the differential operator in the four directions: horizontal d H , vertical d V , 45 degree diagonal d 45 and 135 degrees diagonal d 135 ,

[0044]

[0045] Using the differential operators in the four directions to perform differential processing on the target search area, to obtain gradient maps in the four directions;

[0046]

[0047] Where, Represents the target search area after scaling, and * represents the filtering operation.

[0048] In another embodiment, for step 103, based on the gradient maps in the four directions, gradient maps in eight sub-directions are expanded, and within the radial length of the WRG feature, the positions of the extreme values of the pixel gradients in the eight sub-directions are marked, and the gradient maps in the eight sub-directions within the target search area are normalized, including:

[0049] Taking the intersection of the four directions in the target search area as the center point, expand the four directions into eight sub-directions;

[0050] Count the maximum and minimum values of the pixel gradients in the 8 sub-directions and mark the corresponding positions; taking the pixel at position i as an example, its maximum value in direction o′ and the corresponding position are:

[0051]

[0052] Where, represents the maximum pixel value in direction o′; g i+no′ Represents the pixel value in direction o′; Indicates the position of the maximum pixel value; o′ represents 8 different sub-directions, namely: (1,0), (1,-1), (0,-1), (-1,-1), (-1,0), (-1,1), (0,1), (1,1);

[0053] Similarly, the minimum value of the pixel at position i in direction o′ is With the corresponding position

[0054] According to the maximum and minimum values of the pixel gradients in the 8 sub-directions, the gradient maps in the 8 sub-directions are normalized:

[0055]

[0056] In the next embodiment, for step 103, the normalized gradient map is combined and combined with the grayscale value of the pixel to obtain the WRG feature, such as Figure 2 Shown, including:

[0057] Set the dimension value of the WRG feature to 8·dim+1, sample the normalized gradient values in each direction, and obtain the sampled gradient values.

[0058]

[0059] Where G i,o′ is the sampling gradient value; f(·) is the sampling function, which can be used for summation, maximum value, interpolation and other operations; is the sampling rate, L is the radial length;

[0060] The sampled gradient value G i,o′ and the gray value of the pixel I i Combined, the WRG features of the pixels are composed,

[0061] wrg i ={G i,(1,0) ,G i,(1,-1) ,G i,(0,-1) ,G i,(-1,-1) ,G i,(-1,0) ,G i,(-1,1) ,G i,(0,1) ,G i,(1,1) ,I i}

[0062] Where wrg iis the WRG feature of pixel i.

[0063] The gradient value refers to the pixel value in the gradient map.

[0064] Through the above method, the WRG features of all pixels 8·dim+1 in the search box are obtained.

[0065] In one embodiment, for step 105, based on the WRG features, the KCF algorithm (JF Henriques, R. Caseiro, P. Martins and J. Batista, High-Speed Tracking with Kernelized Correlation Filters, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 3, pp. 583-596, 2015.) is used to calculate the position (x t+1 ,y t+1 ).

[0066] In another embodiment, in step 106, updating the target size according to the positions of the extreme values of the pixel gradients in the eight sub-directions includes:

[0067] According to the center position pixel of the target in the infrared image at the current moment and the next moment, the size change rate of the target in the infrared image at the next moment is calculated using the position of the extreme value of the pixel gradient in the 8 sub-directions.

[0068]

[0069] Where, ρ o′ Indicates the rate of change of size in direction o′;

[0070] Remove the dimensional change rate ρ with a value less than zero o′ , select the median value among the remaining dimensional change rates

[0071] Set the threshold and remove the median value The size change rate ρ between the two values is greater than the threshold o′ ;

[0072] Calculate the remaining size change rate ρ o′ The mean of , get the new target size,

[0073]

[0074] Where ρ represents the new target size; N is the remaining ρ o′ number.

[0075] In one embodiment, in step 107, re-determining the target search area and updating the tracker based on the updated target size includes:

[0076] According to the updated target size and the infrared target size requirements, the target search area in the infrared image at the next moment is scaled to the same size as the tracker;

[0077] Calculate the WGR features of the target search area in the infrared image at the next moment after the update, and use the tracking algorithm in KCF to update the tracker.

[0078] The present invention also proposes an infrared target tracking system based on global radial gradient features, comprising:

[0079] An image acquisition module, used for acquiring a set of infrared image sequences;

[0080] The infrared target tracking module is used to determine the target search area and the radial length of the WRG feature in the infrared image at the next moment based on the target frame of the infrared image at the current moment in the infrared image sequence set; select 4 directions in the target search area, and use the differential algorithm to calculate the gradient map of the 4 directions in the target search area; based on the gradient map of the 4 directions, expand the gradient map of 8 sub-directions, mark the positions of the pixel gradient extreme values in the 8 sub-directions within the radial length of the WRG feature, and normalize the gradient maps in the 8 sub-directions in the target search area; combine the normalized gradient maps and combine them with the grayscale value of the pixel to obtain the WRG feature; based on the WRG feature, use the KCF tracking algorithm to calculate the position of the target in the infrared image at the next moment; update the target size based on the positions of the pixel gradient extreme values in the 8 sub-directions; based on the updated target size, redefine the target search area and update the tracker.

[0081] The present invention further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0082] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above-mentioned method when executed by a processor.

[0083] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made by using the contents of the present invention description and drawings under the inventive concept of the present invention, or direct / indirect application in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. An infrared target tracking method based on global radial gradient features, characterized in that: include: Acquire infrared image sequence set; Determine the target search area and the radial length of the WRG feature in the infrared image at the next moment according to the target frame of the infrared image at the current moment in the infrared image sequence; Selecting four directions within the target search area and calculating the gradient maps of the four directions within the target search area using a differential algorithm; According to the gradient maps in the four directions, the gradient maps in the eight sub-directions are expanded. Within the radial length of the WRG feature, the positions of the extreme values of the pixel gradients in the eight sub-directions are marked. The gradient maps in the eight sub-directions within the target search area are normalized. The normalized gradient maps are combined and combined with the grayscale values of the pixels to obtain the WRG feature. According to the WRG features, the KCF tracking algorithm is used to calculate the position of the target in the infrared image at the next moment; Update the target size according to the position of the extreme pixel gradient in the 8 sub-directions; Based on the updated target size, the target search area is redefined and the tracker is updated; The normalized gradient maps are combined with the grayscale values of the pixels to obtain the WRG features, including: Set the dimension value of the WRG feature to , the normalized gradient values in each direction are sampled to obtain the sampled gradient values, Where, is the sampling gradient value; is the sampling function; is the sampling rate, is the radial length; The sampled gradient value The grayscale value of the pixel Combined, the WRG features of the pixels are composed, Where, Pixels WRG characteristics.

2. The infrared target tracking method according to claim 1, wherein: Determining a target search area and a radial length of a WRG feature in an infrared image at a next moment according to a target frame of the infrared image at a current moment in the infrared image sequence includes: Assume that the center position of the target frame of the infrared image at the current moment is , size is , then the target search area in the infrared image at the next moment is , Where, Indicates the expansion factor of the search window relative to the target size; According to the target frame of the infrared image at the current moment and the size of the tracker , zoom the target search area in the infrared image at the next moment, Where, represents the target search area after scaling, is the image size scaling function, Indicates the image size scaling ratio, ; Calculate the radial length of the WRG feature , For the rounding function: 。 3. The infrared target tracking method according to claim 1, wherein: Four directions are selected within the target search area, and gradient maps of the four directions within the target search area are calculated using a differential algorithm, including: Select four directions in the target search area: horizontal, vertical, 45-degree diagonal and 135-degree diagonal; select the differential operators in the four directions: horizontal , vertical , 45 degree diagonal and 135 degrees diagonally , ， ， ， ； Using the differential operators in the four directions to perform differential processing on the target search area, to obtain gradient maps in four directions; Where, represents the target search area after scaling, Represents a filtering operation.

4. The infrared target tracking method according to claim 1, wherein: Based on the gradient maps in the four directions, the gradient maps in the eight sub-directions are expanded. Within the radial length of the WRG feature, the positions of the extreme values of the pixel gradients in the eight sub-directions are marked. The gradient maps in the eight sub-directions within the target search area are normalized, including: Taking the intersection of the four directions in the target search area as the center point, expand the four directions into eight sub-directions; Count the maximum and minimum values of pixel gradients in the 8 sub-directions and mark the corresponding positions; take the position as For example, the pixel in the direction The maximum value and the corresponding position on are: Where, Indicates the direction The maximum pixel value on ; Indicates the direction The pixel value on ; Indicates the position of the maximum pixel value; Indicates 8 different sub-directions, namely: , , , , , , , ; Similarly, the position is The pixels in the direction Minimum value on With the corresponding position ; According to the maximum and minimum values of the pixel gradients in the 8 sub-directions, the gradient maps in the 8 sub-directions are normalized: 。 5. The infrared target tracking method according to claim 1, wherein: Update the target size based on the position of the pixel gradient extreme value in the 8 sub-directions, including: According to the center position pixel of the target in the infrared image at the current moment and the next moment, the size change rate of the target in the infrared image at the next moment is calculated using the position of the extreme value of the pixel gradient in the 8 sub-directions. Where, Indicates the direction Upper dimensional change rate; Remove dimensional changes with values less than zero , select the median value among the remaining dimensional change rates ; Set the threshold and remove the median value The size change rate between the two is greater than the threshold ; Calculate the remaining size change rate The mean of , get the new target size, Where, Indicates the new target size; For the remaining number.

6. The infrared target tracking method according to claim 1, wherein: Based on the updated target size, the target search area is redefined and the tracker is updated, including: According to the updated target size and the infrared target size requirements, the target search area in the infrared image at the next moment is scaled to the same size as the tracker; Calculate the WGR features of the target search area in the infrared image at the next moment after the update, and use the tracking algorithm in KCF to update the tracker.

7. An infrared target tracking system based on global radial gradient features, characterized in that: include: An image acquisition module, used for acquiring a set of infrared image sequences; The infrared target tracking module is used to determine the target search area and the radial length of the WRG feature in the infrared image at the next moment according to the target frame of the infrared image at the current moment in the infrared image sequence set; select four directions in the target search area, and use the differential algorithm to calculate the gradient map of the four directions in the target search area; according to the gradient map of the four directions, expand the gradient map of eight sub-directions, mark the positions of the pixel gradient extreme values in the eight sub-directions within the radial length of the WRG feature, and normalize the gradient maps in the eight sub-directions in the target search area; combine the normalized gradient maps and combine them with the grayscale value of the pixels to obtain the WRG feature; Based on the WRG features, the KCF tracking algorithm is used to calculate the position of the target in the infrared image at the next moment; the target size is updated according to the position of the pixel gradient extreme value in the eight sub-directions; based on the updated target size, the target search area is re-determined and the tracker is updated; The normalized gradient maps are combined with the grayscale values of the pixels to obtain the WRG features, including: Set the dimension value of the WRG feature to , the normalized gradient values in each direction are sampled to obtain the sampled gradient values, Where, is the sampling gradient value; is the sampling function; is the sampling rate, is the radial length; The sampled gradient value The grayscale value of the pixel Combined, the WRG features of the pixels are composed, Where, Pixels WRG characteristics.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images