Infrared vehicle identification method based on combination of target segmentation and target matching
By using a combination of target segmentation and matching in infrared images, morphological filtering and SURF algorithms are used to solve the problem of infrared long-distance vehicle target recognition, and efficient and accurate vehicle target detection is achieved.
Patent Information
- Application Number
- CN202410280798.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-12
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art is difficult to identify when detecting vehicle targets from a long distance due to weak, small, rotation and scaling of the targets. The traditional matching algorithm has a large amount of calculation and is easily disturbed by the ground objects background, and the recognition effect is poor.
Using a combination of target segmentation and target matching, the noise and background are removed through an improved morphological filtering algorithm, local binary mode and target polarity are calculated, candidate targets are screened, and the final match is used to determine the vehicle target.
By removing low-frequency backgrounds and using local binary modes and polar feature screening, the calculation amount is significantly reduced, the recognition accuracy and real-time nature of infrared vehicle targets are improved, and the problem of difficulty in long-distance recognition is solved.
Smart Images

Figure CN118968104A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of photoelectric guidance, and in particular relates to an infrared vehicle recognition method based on the combination of target segmentation and target matching. Background Art
[0002] Infrared target detection requires long-distance target identification. The imaging characteristics of infrared targets such as people and cars at a long distance on the detector are small area, unclear outline, and no texture information. At the same time, infrared targets are rotated and scaled during the identification process, which makes identification extremely difficult.
[0003] Traditional infrared vehicle target recognition requires the target to be close and have a clear outline. When the target is far away, the target image is a circular spot, which blends into the background and the recognition effect is poor. When matching methods such as SIFT and SURF are used directly, there are many interference sources from the background and the matching is very likely to fail. At the same time, such matching algorithms need to calculate feature points in the entire frame, which requires a lot of calculations. Summary of the invention
[0004] The purpose of the present invention is to provide an infrared vehicle recognition method based on the combination of target segmentation and target matching to solve the technical problem that the existing technology cannot recognize the target due to the weak, small, rotation and scaling of the target when detecting vehicle targets at a long distance by infrared.
[0005] To achieve the above object, the present invention adopts the following technical solution:
[0006] An infrared vehicle recognition method based on the combination of target segmentation and target matching comprises the following steps: removing the noise of the infrared image, suppressing the background of the infrared image by using an improved morphological filtering algorithm, and segmenting and extracting candidate targets; calculating the local binary pattern of the candidate target to obtain the binary pattern map of the candidate target, and screening out the candidate targets that meet the matching requirements by comparing with the binary pattern map of a template target; calculating the target polarity of the selected candidate targets that meet the matching requirements, and comparing with the target polarity of the template target, and screening out the candidate targets with the same polarity; using the SURF matching algorithm for the screened candidate targets with the same polarity, matching with the template target, and determining the vehicle target.
[0007] Preferably, removing the noise of the infrared image includes: filtering the infrared image using an improved median filtering algorithm to remove the noise of the infrared image.
[0008] Preferably, an improved morphological filtering algorithm is used to suppress the background of the infrared image and segment and extract the candidate target, including: using an extremum filter to find the maximum and minimum grayscale values to estimate the image background, and performing background cancellation to obtain the candidate target.
[0009] Preferably, the local binary pattern of the above-mentioned candidate target is calculated to obtain a binary pattern map of the above-mentioned candidate target, including: taking the above-mentioned candidate target as the center, obtaining the grayscale mean of the adjacent area; subtracting the grayscale mean of the above-mentioned adjacent area from the grayscale value of the area where the above-mentioned candidate target is located, to obtain the binary pattern map of the above-mentioned candidate target, wherein the above-mentioned binary pattern map indicates the relationship between the grayscale mean of each adjacent area and the grayscale value of the area where the above-mentioned candidate target is located.
[0010] Preferably, candidate targets that meet the comparison requirements are screened out by comparing with the binary pattern graph of the template target, including: comparing the binary pattern graph of the candidate target and the binary pattern graph of the template target; and screening out candidate targets that meet the quantity requirement of the same numerical value.
[0011] Preferably, calculating the target polarity of the selected candidate target that meets the comparison requirements includes: calculating the magnitude relationship between the grayscale mean of the region where the candidate target is located and the grayscale mean of the adjacent region to obtain the polarity of the candidate target.
[0012] Preferably, screening out candidate targets with the same polarity includes: screening out candidate targets with the same polarity as the above-mentioned template target.
[0013] Preferably, the selected candidate targets with the same polarity are matched with the template target using the SURF matching algorithm, including: calculating the feature points of each candidate target respectively, and performing feature matching on each feature point with the template target respectively using the SURF matching algorithm.
[0014] Preferably, determining the vehicle target includes: sorting the candidate targets according to the number of matched feature points; and determining the candidate target with the largest number of matched feature points as the vehicle target.
[0015] An infrared vehicle recognition system based on the combination of target segmentation and target matching comprises a segmentation and extraction module, which is used to suppress the background of the infrared image by using an improved morphological filtering algorithm while removing the noise of the infrared image, and to segment and extract candidate targets; a primary screening module, which is used to calculate the local binary mode of the candidate target, obtain the binary mode map of the candidate target, and screen out the candidate targets that meet the matching requirements by comparing with the binary mode map of the template target; a secondary screening module, which is used to calculate the target polarity of the selected candidate targets that meet the matching requirements, and compare with the target polarity of the template target, and screen out the candidate targets with the same polarity; and a matching module, which is used to match the screened candidate targets with the same polarity with the template target by using the SURF matching algorithm, and determine the vehicle target.
[0016] A computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the infrared vehicle recognition method based on the combination of target segmentation and target matching when running.
[0017] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the infrared vehicle recognition method based on the combination of target segmentation and target matching through the computer program.
[0018] In the present invention, by removing the noise of the infrared image, using an improved morphological filtering algorithm to suppress the background of the infrared image, segmenting and extracting the candidate target; calculating the local binary pattern of the candidate target to obtain the binary pattern map of the candidate target, and screening the candidate target that meets the comparison requirements by comparing it with the binary pattern map of the template target; calculating the target polarity of the selected candidate target that meets the comparison requirements, and comparing the target polarity of the template target to screen out the candidate target with the same polarity; using the SURF matching algorithm for the screened candidate target with the same polarity to match it with the template target, and determining the vehicle target, thereby The background is removed and the two features of local binary pattern and polarity are used for coarse step screening. The number of remaining candidate targets is very small compared to the whole image. The improved SURF algorithm can quickly and accurately identify vehicle targets. The segmentation method is used to find all candidate targets including vehicle targets. Most false targets are eliminated through local binary patterns and target polarity. The SURF algorithm is used for matching in the remaining candidate target areas, which greatly reduces the amount of calculation. It solves the technical problem that the existing technology cannot identify the target when detecting vehicle targets at a long distance in infrared due to weak, small, rotated and scaled targets, and achieves the technical effect of improving the accuracy of target detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a flow chart of an infrared vehicle recognition method based on the combination of target segmentation and target matching in an embodiment of the present invention;
[0020] Figure 2 It is a flow chart of an infrared vehicle recognition method based on the combination of target segmentation and target matching in an embodiment of the present invention;
[0021] Figure 3 is a schematic diagram of the process of improved median filtering in an embodiment of the present invention;
[0022] Figure 4 is a flow chart of a background suppression algorithm in an embodiment of the present invention;
[0023] Figure 5 is a schematic diagram of a local binary pattern in an embodiment of the present invention;
[0024] Figure 6 is a schematic diagram of target polarity in an embodiment of the present invention;
[0025] Figure 7 is a schematic diagram of a flow chart of an improved SURF algorithm in an embodiment of the present invention;
[0026] Figure 8 It is a structural schematic diagram of an infrared vehicle recognition system based on the combination of target segmentation and target matching in an embodiment of the present invention. DETAILED DESCRIPTION
[0027] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims. It should be noted that the drawings are all in very simplified form and are not in precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.
[0028] It should be noted that, in order to clearly explain the content of the present invention, the present invention specifically cites multiple embodiments to further illustrate different implementations of the present invention, wherein the multiple embodiments are enumerated rather than exhaustive. In addition, for the sake of brevity of explanation, the contents mentioned in the previous embodiments are often omitted in the subsequent embodiments. Therefore, the contents not mentioned in the subsequent embodiments can refer to the previous embodiments accordingly.
[0029] Example 1
[0030] An infrared vehicle recognition method based on the combination of target segmentation and target matching, such as Figure 1 As shown, the method includes:
[0031] S102, while removing the noise of the infrared image, using an improved morphological filtering algorithm to suppress the background of the infrared image, and segment and extract the candidate target;
[0032] S104, calculating the local binary pattern of the candidate target to obtain a binary pattern map of the candidate target, and screening out the candidate targets that meet the comparison requirements by comparing with the binary pattern map of the template target;
[0033] S106, calculating the target polarity of the selected candidate targets that meet the comparison requirements, and comparing the target polarity of the template target to screen out candidate targets with the same polarity;
[0034] S108, using the SURF matching algorithm to match the selected candidate targets with the same polarity with the template target to determine the vehicle target.
[0035] As an optional implementation, removing noise from the infrared image includes: using an improved median filtering algorithm to filter the infrared image to remove noise from the infrared image.
[0036] As an optional implementation, an improved morphological filtering algorithm is used to suppress the background of infrared images and segment and extract candidate targets, including: using an extremum filter to find the maximum and minimum grayscale values to estimate the image background, and performing background cancellation to obtain candidate targets.
[0037] As an optional implementation, the local binary pattern of the candidate target is calculated to obtain a binary pattern map of the candidate target, including: taking the candidate target as the center, obtaining the grayscale mean of the adjacent area; subtracting the grayscale mean of the adjacent area from the grayscale value of the area where the candidate target is located to obtain a binary pattern map of the candidate target, wherein the binary pattern map indicates the relationship between the grayscale mean of each adjacent area and the grayscale value of the area where the candidate target is located.
[0038] As an optional implementation, candidate targets that meet the comparison requirements are screened out by comparing with the binary pattern graph of the template target, including: comparing the binary pattern graph of the candidate target and the binary pattern graph of the template target; screening out candidate targets that meet the quantity requirements of the same numerical value.
[0039] As an optional implementation, the target polarity is calculated for the selected candidate target that meets the comparison requirements, including: calculating the magnitude relationship between the grayscale mean of the region where the candidate target is located and the grayscale mean of the adjacent region to obtain the polarity of the candidate target.
[0040] As an optional implementation manner, screening out candidate targets with the same polarity includes: screening out candidate targets with the same polarity as the template target.
[0041] As an optional implementation, the selected candidate targets with the same polarity are matched with the template target using the SURF matching algorithm, including: calculating the feature points of each candidate target respectively, and using the SURF matching algorithm to perform feature matching on each feature point with the template target respectively.
[0042] As an optional implementation manner, determining the vehicle target includes: sorting the candidate targets according to the number of matched feature points; and determining the candidate target with the largest number of matched feature points as the vehicle target.
[0043] In an embodiment of the present application, by removing the low-frequency background and coarse-step screening through two features of local binary pattern and polarity, the remaining candidate targets are very small compared to the whole image. The improved SURF algorithm can quickly and accurately identify vehicle targets, use a segmentation method to find all candidate targets including vehicle targets, eliminate most false targets through local binary patterns and target polarity, and use the SURF algorithm for matching in the remaining candidate target areas, which greatly reduces the amount of calculation, solves the technical problem of the prior art that the target cannot be identified when detecting vehicle targets at long distances by infrared due to weak, small, rotated and scaled targets, and achieves the technical effect of improving the accuracy of target detection.
[0044] The above infrared vehicle recognition application based on target segmentation and target matching is not limited to the following: Figure 2 As shown, the following steps are included:
[0045] a. Improved median filtering image denoising: Improved median filtering is performed on the image to be identified to remove the salt and pepper noise introduced by the detector imaging;
[0046] b. Background suppression target extraction: Use the improved morphological filtering algorithm to extract all candidate targets, remove all low-frequency backgrounds in the image, and obtain all suspected vehicle targets;
[0047] c. Eliminate fake local binary pattern targets at one time: Calculate the local binary patterns of all suspected vehicle targets, calculate the local binary patterns of the binding targets, and compare them with the local binary patterns of all suspected vehicle targets, and eliminate the suspected targets that do not meet the pairing requirements;
[0048] d. Secondary rejection of target polarity: For the suspected targets remaining from the rough screening, the target polarity of the suspected targets is calculated and compared with the target polarity of the bound targets to eliminate the suspected targets with opposite polarity;
[0049] e. SURF matching target determination: For the suspected targets remaining after the secondary screening, use the SURF matching algorithm to match the suspected targets and intercept the vehicle targets.
[0050] That is, by combining target segmentation and target matching, median filtering is performed on each frame of the input image to obtain an image with noise or bad pixels removed; background suppression target extraction is performed on the image without noise interference to obtain all high-frequency candidate targets; all candidate targets are compared with the bound template using the local binary pattern for the first round of screening to remove candidate targets that do not satisfy the local binary pattern; for the candidate targets remaining after the first round of screening, the polarity of the candidate targets is compared with the bound template to remove candidate targets with opposite polarities; finally, the SURF algorithm is used to match the target after the second screening with the template to confirm the final target.
[0051] Specifically, the infrared vehicle recognition method based on the combination of target segmentation and target matching comprises the following steps when applied and implemented:
[0052] In the first step, the infrared image is filtered using an improved median filter algorithm to remove infrared image noise.
[0053] Median filtering has a good effect on Gaussian noise and bad pixels of infrared images. The traditional median filtering algorithm generally uses a 3*3 or 5*5 sliding kernel as a window, but the algorithm is time-consuming and has poor real-time performance. In order to improve the real-time performance of the algorithm, the median filter window is improved, and a 1*5 filter window is used. At the same time, the sorting algorithm is optimized and an iterative sorting algorithm is used, which greatly improves the efficiency and real-time performance of the algorithm. The process of the improved median filtering algorithm is not limited to the following: Figure 3 As shown, select the first 5 pixels of the i-th row, sort the 5 data, write the pixel value of the 3rd position in the sorting into the 3rd pixel value, and compare the template to the right by one position. Determine whether it has reached the rightmost end? If it is determined that it has not reached the rightmost end, return to sort the 5 data. If it is determined that it has reached the rightmost end, determine whether it is the last row? If it is determined to be the last row, then end. If it is determined that it is not the last row, move down one row i=i+1, and execute the selection of the first 5 pixels.
[0054] In the second step, an improved morphological filtering algorithm is used to suppress the background and segment the candidate targets.
[0055] When the target is extremely small, it is imaged as a speckled target. Using image filtering to suppress the background can significantly improve the signal-to-noise ratio of the target, making it easier to extract. The traditional morphological filtering algorithm is very computationally intensive. In order to improve the efficiency of the algorithm, an extreme value filter is used. The method of obtaining the maximum and minimum grayscale values of the structural operator region is used to replace the expansion and corrosion operations, estimate the image background, and then cancel the background to obtain an image that retains the target. The process of using the improved morphological filtering algorithm to suppress the background and segment the candidate target is not limited to the following. Figure 4 As shown, the denoised image is first eroded, then the eroded image is expanded, the denoised image is subtracted, and histogram statistics are performed. Then, the candidate target is obtained through threshold segmentation.
[0056] The third step is to obtain the local binary pattern of the candidate region.
[0057] Since the target to be identified is small and has a very low resolution, the recognition probability is extremely low if the texture information and edge information of the target are used alone. In order to improve the recognition probability, the neighborhood information of the candidate target is fully utilized, and the relative relationship between the candidate target and the neighborhood is used to improve the recognition probability.
[0058] After finding the candidate target area, the grayscale mean of the candidate target area and its 8 neighborhoods are calculated respectively, and the grayscale mean of the 8 neighborhoods is subtracted from the central target area. If it is greater than, it is set to 1, and if it is less than, it is set to 0. In this way, the grayscale mean distribution relationship diagram of the target and the target neighborhood is obtained. The binary pattern diagram of the grayscale mean is not limited to the following Figure 5 As shown in the figure, the larger circle in the middle represents the candidate target area, white indicates 0, and black indicates 1. The grayscale distribution relationship diagram has rotation invariance and scale invariance. The binary pattern map of the binding template target is compared with the binary pattern map of the candidate area. If 5 or more than 5 of the 8 neighborhood areas are correctly compared, the candidate target is retained; otherwise, the candidate target is eliminated.
[0059] Step 4: Polarity comparison between candidate targets and binding templates
[0060] When the grayscale mean of the candidate target area is greater than the grayscale mean of the neighborhood, it is considered to be positive polarity; otherwise, it is considered to be negative polarity. Figure 6 As shown. The polarity of the candidate target is compared with the binding template. If the polarity is consistent, the candidate target is retained. If the polarity is opposite, the candidate target is eliminated. Polarity matching is simple to calculate and some candidate targets with opposite polarity are eliminated.
[0061] The fifth step is to improve the SURF algorithm to match and confirm the target.
[0062] The candidate target areas that have undergone two rounds of screening are matched with the binding templates using SURF. First, the feature points of the binding template and the candidate area are calculated respectively to obtain two groups of feature points. All feature points are described using 64-dimensional descriptors. In order to reduce the amount of calculation, the first 32-dimensional descriptors with larger module values are selected to describe the feature points. The 32-dimensional descriptors of the two groups of feature points are used to perform pairwise matching and calculate the Euclidean distance of the descriptors. If the Euclidean distance is less than 1, the match is considered successful, otherwise, the match fails. Specifically, the process of using SURF to match the binding templates is not limited to the following. Figure 7 As shown in the figure, based on SURF, the binding template is matched with the candidate area, and the matching points are counted to determine whether the matching points are greater than 4. If the matching points are less than or equal to 4, the target is eliminated. If the matching points are greater than 4, the number of matching points in the area is recorded, and the number of matching points of all candidate areas is obtained. The candidate area with the largest number of matching points is the recognized target, and the recognized target position information is output.
[0063] The number of correct matching points in all candidate regions is sorted, and the region with the largest number of matching points is the identified vehicle target. Since the low-frequency background has been removed and the local binary pattern and polarity features have been coarsely screened, the remaining candidate targets are fewer in number than the entire image. The improved SURF algorithm can quickly and accurately identify vehicle targets.
[0064] In an embodiment of the present application, an infrared vehicle recognition method based on a combination of target segmentation and target matching uses a segmentation method to find all high-frequency targets including vehicle targets; most false targets are eliminated through local binary patterns and target polarity; and in the remaining target areas, a SURF algorithm is used for matching, which greatly reduces the amount of calculation and improves detection accuracy.
[0065] By using target segmentation, all suspected targets are extracted, and pseudo targets are eliminated using local binary patterns and polarity correlation; the SURF matching algorithm is used to achieve fast and robust detection and recognition of infrared vehicle targets. This effectively solves the problems of small imaging area, unclear contours, no texture, rotation, and scaling of long-distance infrared vehicle targets, and improves the real-time and accuracy of matching and recognition.
[0066] Example 2
[0067] An infrared vehicle recognition system based on the combination of target segmentation and target matching, such as Figure 8 As shown, the system includes:
[0068] The segmentation and extraction module 802 is used to suppress the background of the infrared image by using an improved morphological filtering algorithm and to segment and extract candidate targets while removing the noise of the infrared image;
[0069] A primary screening module 804 is used to calculate the local binary pattern of the candidate target, obtain the binary pattern map of the candidate target, and screen out the candidate targets that meet the comparison requirements by comparing with the binary pattern map of the template target;
[0070] A secondary screening module 806 is used to calculate the target polarity of the selected candidate targets that meet the comparison requirements, and compare the target polarity of the template target to screen out candidate targets with the same polarity;
[0071] The matching module 808 is used to use the SURF matching algorithm to match the selected candidate targets with the same polarity with the template target to determine the vehicle target.
[0072] Optionally, the segmentation and extraction module 802 removes noise from the infrared image, including: using an improved median filtering algorithm to filter the infrared image to remove noise from the infrared image.
[0073] Optionally, the segmentation and extraction module 802 uses an improved morphological filtering algorithm to suppress the background of the infrared image and segment and extract candidate targets, including: using an extremum filter to find the maximum and minimum grayscale values to estimate the image background, and performing background cancellation to obtain candidate targets.
[0074] Optionally, the above-mentioned primary screening module 804 calculates the local binary pattern of the candidate target to obtain a binary pattern map of the candidate target, including: taking the candidate target as the center, obtaining the grayscale mean of the adjacent area; subtracting the grayscale mean of the adjacent area from the grayscale value of the area where the candidate target is located, to obtain the binary pattern map of the candidate target, wherein the binary pattern map indicates the relationship between the grayscale mean of each adjacent area and the grayscale value of the area where the candidate target is located.
[0075] Optionally, the primary screening module 804 screens out candidate targets that meet the comparison requirements by comparing with the binary pattern graph of the template target, including: comparing the binary pattern graph of the candidate target with the binary pattern graph of the template target; and screening out candidate targets that meet the quantity requirements of the same numerical value.
[0076] Optionally, the secondary screening module 806 calculates the target polarity for the selected candidate targets that meet the comparison requirements, including: calculating the magnitude relationship between the grayscale mean of the region where the candidate target is located and the grayscale mean of the adjacent region to obtain the polarity of the candidate target.
[0077] Optionally, the secondary screening module 806 screens out candidate targets with the same polarity, including: screening out candidate targets with the same polarity as the template target.
[0078] Optionally, the matching module 808 uses the SURF matching algorithm to match the selected candidate targets with the same polarity with the template target, including: calculating the feature points of each candidate target respectively, and using the SURF matching algorithm to perform feature matching on each feature point with the template target respectively.
[0079] Optionally, the matching module 808 determines the vehicle target, including: sorting the candidate targets according to the number of matched feature points; and determining the candidate target with the largest number of matched feature points as the vehicle target.
[0080] In the embodiment of the present application, by removing the noise of the infrared image, using an improved morphological filtering algorithm to suppress the background of the infrared image, segmenting and extracting the candidate target; calculating the local binary pattern of the candidate target to obtain the binary pattern map of the candidate target, and by comparing it with the binary pattern map of the template target, screening out the candidate targets that meet the comparison requirements; calculating the target polarity of the selected candidate targets that meet the comparison requirements, and comparing the target polarity of the template target, screening out the candidate targets with the same polarity; using the SURF matching algorithm for the screened candidate targets with the same polarity, matching them with the template target, and determining the vehicle target, thereby The low-frequency background is removed, and the two features of local binary pattern and polarity are used for coarse-step screening. The number of remaining candidate targets is very small compared to the whole image. The improved SURF algorithm can quickly and accurately identify vehicle targets. The segmentation method is used to find all candidate targets including vehicle targets. Most false targets are eliminated through local binary patterns and target polarity. The SURF algorithm is used for matching in the remaining candidate target areas, which greatly reduces the amount of calculation. It solves the technical problem that the existing technology cannot identify targets when detecting vehicle targets at long distances in infrared due to weak, small targets, rotation and scaling, and achieves the technical effect of improving target detection accuracy.
[0081] Example 3
[0082] Another aspect of the embodiments of the present invention further provides an electronic device for implementing the above infrared vehicle identification method based on the combination of target segmentation and target matching, and the electronic device is not limited to being a terminal device or a server in a radar system. The electronic device is not limited to including a memory and a processor, the memory stores a computer program, and the processor is configured to execute the steps in any of the above method embodiments through the computer program.
[0083] Example 4
[0084] Another aspect of the embodiments of the present invention provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in various optional embodiments of the infrared vehicle recognition method based on the combination of target segmentation and target matching. The computer program is configured to execute the steps of any of the above method embodiments when it is run.
Claims
1. An infrared vehicle recognition method based on the combination of target segmentation and target matching, characterized in that: include: Under the condition of removing the noise of the infrared image, using the improved morphological filtering algorithm to suppress the background of the infrared image, and segment and extract the candidate target; Calculating the local binary pattern of the candidate target to obtain the binary pattern map of the candidate target, and screening out the candidate targets that meet the comparison requirements by comparing with the binary pattern map of the template target; Calculating the target polarity of the selected candidate targets that meet the comparison requirements, and comparing the target polarity of the template target to screen out candidate targets with the same polarity; The SURF matching algorithm is used to match the selected candidate targets with the same polarity with the template target to determine the vehicle target.
2. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: Remove noise from infrared images, including: The infrared image is filtered using the improved median filtering algorithm to remove the infrared image noise.
3. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: The infrared image is subjected to background suppression by using an improved morphological filtering algorithm, and candidate targets are segmented and extracted, including: The maximum and minimum grayscale values are calculated using the extreme value filter to estimate the image background, and the background is cancelled to obtain the candidate target.
4. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: Calculating the local binary pattern of the candidate target to obtain a binary pattern map of the candidate target includes: Taking the candidate target as the center, obtaining the grayscale mean of the adjacent area; Subtract the grayscale mean of the adjacent regions from the grayscale value of the region where the candidate target is located to obtain a binary pattern map of the candidate target, wherein the binary pattern map indicates the size relationship between the grayscale mean of each adjacent region and the grayscale value of the region where the candidate target is located.
5. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 4, characterized in that: By comparing the binary pattern with the template target, candidate targets that meet the comparison requirements are screened out, including: Comparing the binary pattern image of the candidate target with the binary pattern image of the template target; Filter out candidate targets that meet the quantity requirement of the same value.
6. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: Calculate the target polarity for the selected candidate targets that meet the comparison requirements, including: The magnitude relationship between the grayscale mean of the region where the candidate target is located and the grayscale mean of the adjacent region is calculated to obtain the polarity of the candidate target.
7. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: Screen out candidate targets with the same polarity, including: Candidate targets whose polarity is consistent with the template target are screened out.
8. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 1, characterized in that: The selected candidate targets with the same polarity are matched with the template target using the SURF matching algorithm, including: The feature points of each candidate target are calculated respectively, and the SURF matching algorithm is used to perform feature matching between each feature point and the template target respectively.
9. The infrared vehicle recognition method based on the combination of target segmentation and target matching as claimed in claim 8, characterized in that: Determine vehicle objectives, including: Sorting the candidate targets according to the number of matched feature points; The candidate target with the largest number of matched feature points is determined as the vehicle target.
10. An infrared vehicle recognition system based on target segmentation and target matching, characterized in that: include: A segmentation and extraction module is used to suppress the background of the infrared image by using an improved morphological filtering algorithm and to segment and extract candidate targets while removing noise from the infrared image; A primary screening module is used to calculate the local binary pattern of the candidate target, obtain the binary pattern map of the candidate target, and screen out the candidate targets that meet the comparison requirements by comparing with the binary pattern map of the template target; A secondary screening module, used to calculate the target polarity of the selected candidate targets that meet the comparison requirements, and compare the target polarity of the template target to screen out candidate targets with the same polarity; The matching module is used to use the SURF matching algorithm to match the selected candidate targets with the same polarity with the template target to determine the vehicle target.
