A small target detection method and detection system based on infrared sequence images

Abstract

The present invention provides a small target detection method and a small target detection system based on infrared sequence images. The detection method comprises: performing background suppression on the infrared image to be detected based on the gray value of each pixel in the infrared image to be detected; performing threshold segmentation on the infrared image to be detected after background suppression, and extracting the region of interest therein, wherein the region of interest is a region that may contain a target; inputting the region of interest into a trained support vector machine, and detecting and classifying the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples of infrared small targets containing generated tracks.

Description

Technical Field

[0001] The present invention relates to the field of photoelectric detection, and more specifically, to a small target detection method and a detection system based on infrared sequence images. Background Art

[0002] Infrared early warning systems have been widely used in modern optoelectronic warfare due to their good ability to resist electronic interference, passively detect targets and work around the clock, and are responsible for early warning of incoming aircraft or missiles. The difficulty of infrared early warning lies in the detection of infrared weak targets under complex backgrounds. Infrared weak targets refer to the weak images of long-distance aircraft or missiles with few pixels in the image, which are easily submerged by complex backgrounds or noise and difficult to detect and identify.

[0003] At present, there are many ways to improve the target detection distance and detection probability under the condition of certain detector performance. The existing mainstream small target detection method is to detect first and then track, which is easy to implement and has high algorithm execution efficiency. Another mainstream method is based on statistical learning. First, a sample set of infrared small targets and background edges is constructed, and then the parameters of the optimal classification hyperplane are obtained through the classifier. Finally, the classification hyperplane is used to identify infrared small targets and remove background edges during the detection of infrared small targets. Summary of the invention

[0004] The present invention aims at solving the technical problems in the prior art and provides a small target detection method and a small target detection system based on infrared sequence images.

[0005] According to a first aspect of the present invention, a method for detecting small targets based on infrared sequence images is provided, comprising:

[0006] Based on the gray value of each pixel in the infrared image to be detected, background suppression is performed on the infrared image to be detected;

[0007] Performing threshold segmentation on the infrared image to be detected after background suppression, and extracting the region of interest therein, wherein the region of interest is a region that may contain the target;

[0008] The region of interest is input into a trained support vector machine to detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples containing infrared dim small targets with generated tracks.

[0009] Based on the above technical solution, the present invention can also make the following improvements.

[0010] Optionally, the performing background suppression on the infrared image to be detected based on the grayscale value of each pixel in the infrared image to be detected includes:

[0011] For any pixel in the infrared image to be detected, a filtering window is selected around it;

[0012] The grayscale value of each pixel in the filter window is arranged in order according to size, and the grayscale value median of the arrangement result is used to replace the grayscale value of any pixel;

[0013] Each pixel in the infrared image to be detected is traversed, and the gray value of each pixel is changed to complete the background suppression of the infrared image to be detected.

[0014] Optionally, performing threshold segmentation on the infrared image to be detected after background suppression to extract the region of interest therein includes:

[0015] Segmenting the infrared image to be detected after background suppression based on a preset size window to obtain a plurality of infrared image regions, wherein each infrared image region contains a plurality of pixels;

[0016] Perform a traversal operation on each infrared image area, use h(t) to represent the number of pixels corresponding to each gray value t, subtract adjacent h(t)s from each other, and obtain g(t)=h(t)-h(t-1); at this time, if g(t-2)+g(t-1)+g(t+1)+g(t+2)≥0, then g(t)=1; if g(t-2)+g(t-1)+g(t+1)+g(t+2)<0, then g(t)=-1, and g(t) is obtained. At this time, traverse g(t), add the number of consecutive positive 1s and negative 1s from negative to positive, and the turning point from negative to positive with the largest sum is the first set threshold T; traverse the pixel values ​​inside each infrared image area and compare them with the threshold T. If it is greater than T, the mark value of the area is 1; if it is less than T, the mark value is 0, wherein the mark value 1 indicates that the infrared image area may contain a target, and the mark value 0 indicates that the infrared image area does not contain a target;

[0017] Performing image segmentation on the infrared image region with a mark value of 1, and calculating the coordinate value of each candidate target region in the infrared image region;

[0018] Based on the coordinate value of each candidate target region, the region of interest of the infrared image region is segmented to extract the regions of interest, wherein each region of interest contains a candidate target.

[0019] Optionally, the step of performing image segmentation on the infrared image region with a mark value of 1 and calculating the coordinate value of each candidate target region in the infrared image region further includes:

[0020] For the infrared image regions with a mark value of 1, the probability of containing a target in each infrared image region is calculated, and when the probability of containing a target is greater than a second set threshold, it is determined that the infrared image region contains a candidate target, otherwise, the candidate target is not contained;

[0021] Accordingly, image segmentation is performed on the infrared image region containing the candidate target.

[0022] Optionally, after the multiple interest regions extracted from the infrared image to be detected are input into a trained support vector machine, the support vector machine outputs the position coordinate value of each candidate target in the infrared image to be detected;

[0023] The continuous multiple frames of infrared images are subjected to inter-frame track association, and based on the track association result, the authenticity of the target in the infrared image to be detected is confirmed.

[0024] Optionally, the continuous multiple frames of infrared images are subjected to inter-frame track association, and based on the track association result, the authenticity of the target in the infrared image to be detected is confirmed:

[0025] The infrared image to be detected is used as the current frame image, and for each of the multiple consecutive frames of images before and after the current frame image, the position coordinate value of each target in each frame image is output based on the trained support vector machine;

[0026] If the position coordinate values ​​of multiple targets in two adjacent frames of images with a preset ratio can be matched in the multiple frames of images, the continuous multiple frames of images are successfully associated, and there is a real target in the infrared image to be detected; otherwise, the continuous multiple frames of images are failed to be associated, and there is no target in the infrared image to be detected.

[0027] Optionally, the position coordinate values ​​of multiple targets in two adjacent frames of images are matched in the following manner:

[0028] For a pair of position coordinate values ​​of any target in two adjacent frames of images, if the position coordinate value in the current frame of image falls within a preset radius circular gate of the corresponding position coordinate value in the previous frame of image, the position coordinate value of any target is successfully matched;

[0029] If the position coordinate values ​​of all targets in two adjacent frames of images are matched successfully, the two adjacent frames of images are successfully associated.

[0030] Optionally, also include:

[0031] Generate the track of weak infrared targets based on the successfully associated continuous multi-frame images;

[0032] Based on the infrared image of the infrared dim small target generating the track, the initial training sample is updated in real time to obtain the real-time training sample of the support vector machine;

[0033] Accordingly, the support vector machine is trained based on the real-time training samples.

[0034] According to a second aspect of the present invention, there is provided a small target detection system based on infrared sequence images, comprising:

[0035] A background suppression module, used for performing background suppression on the infrared image to be detected based on the gray value of each pixel in the infrared image to be detected;

[0036] An extraction module is used to perform threshold segmentation on the infrared image to be detected after background suppression, and extract the region of interest therein, wherein the region of interest is a region that may contain a target;

[0037] The target detection module is used to input the region of interest into a trained support vector machine to detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples containing infrared dim small targets with generated tracks.

[0038] According to a third aspect of the present invention, there is provided an electronic device comprising a memory and a processor, wherein the processor is used to implement the steps of a small target detection method based on infrared sequence images when executing a computer management program stored in the memory.

[0039] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium on which a computer management program is stored, and when the computer management program is executed by a processor, the steps of a small target detection method based on infrared sequence images are implemented.

[0040] The invention provides a small target detection method and a small target detection system based on infrared sequence images. When training a support vector machine, infrared images of small targets with generated tracks are used to perform online training of the support vector machine, thereby improving the target recognition accuracy of the support vector machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a small target detection method based on infrared sequence images provided by the present invention;

[0042] Figure 2 It is a schematic diagram of the overall process of the small target detection method based on infrared sequence images;

[0043] Figure 3 A schematic diagram of the structure of a small target detection system based on infrared sequence images provided by the present invention;

[0044] Figure 4 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0045] Figure 5 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0046] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0047] Embodiment 1

[0048] A small target detection method based on sequence images, see Figure 1 , the small target detection method includes:

[0049] S1, performing background suppression on the infrared image to be detected based on the grayscale value of each pixel in the infrared image to be detected.

[0050] As an embodiment, the background suppression of the infrared image to be detected is performed based on the grayscale value of each pixel in the infrared image to be detected, including: selecting a filter window around any pixel in the infrared image to be detected; arranging the grayscale value of each pixel in the filter window in order according to size, and replacing the grayscale value of any pixel with the median of the grayscale value of the arrangement result; traversing each pixel in the infrared image to be detected, changing the grayscale value of each pixel, and completing the background suppression of the infrared image to be detected.

[0051] Specifically, an infrared image including the current frame corresponding to the target to be identified is obtained, and the infrared image of the current frame is used as the infrared image to be detected. For any pixel in the infrared image to be detected, a filter window is selected around it, and the grayscale value of each pixel in the filter window is arranged in order according to size, and the grayscale value of the pixel is replaced by the median of the grayscale value of the arrangement result, and then a traversal operation is performed on each pixel in the entire image to complete the background suppression of the infrared image to be detected.

[0052] S2, performing threshold segmentation on the infrared image to be detected after background suppression, and extracting the region of interest therein, wherein the region of interest is a region that may contain the target.

[0053] As an embodiment, the infrared image to be detected after background suppression is threshold segmented to extract the region of interest therein, including: segmenting the infrared image to be detected after background suppression based on a preset size window to obtain multiple infrared image regions, wherein each infrared image region contains multiple pixels; performing a traversal operation on each infrared image region, using h(t) to represent the number of pixels corresponding to each gray value t, and subtracting adjacent h(t)s from each other to obtain g(t)=h(t)-h(t-1); at this time, if g(t-2)+g(t-1)+g(t+1)+g(t+2)≥0, then g(t)=1; if g(t-2)+g(t-1)+g(t+1)+g(t+2)<0, then g(t)=-1 to obtain g(t), at this time, traverse g(t), add the number of consecutive positive 1s and negative 1s from negative to positive, and the turning point from negative to positive with the largest sum is the first set threshold T. The pixel values ​​inside each infrared image area are traversed and compared with the threshold T. If the pixel value is greater than T, the mark value of the area is 1; if the pixel value is less than T, the mark value is 0, wherein the mark value 1 indicates that the infrared image area may contain a target, and the mark value 0 indicates that the infrared image area does not contain a target; the infrared image area with a mark value of 1 is segmented, and the coordinate value of each candidate target area in the infrared image area is calculated; based on the coordinate value of each candidate target area, the interest area of ​​the infrared image area is segmented to extract the interest area, wherein each interest area contains a candidate target.

[0054] As an embodiment, the infrared image area with a mark value of 1 is subjected to image segmentation, and the coordinate value of each candidate target area in the infrared image area is calculated. Previously, it also includes: for the infrared image area with a mark value of 1, the probability of containing the target in each infrared image area is calculated; when the probability of containing the target is greater than a second set threshold, it is determined that the infrared image area contains the candidate target; otherwise, the candidate target is not contained; accordingly, the infrared image area containing the candidate target is subjected to image segmentation.

[0055] Specifically, a window of preset pixel size (preset size window) is used to divide the infrared image to be detected to obtain multiple infrared image regions. The entire image is traversed, and h(t) is used to represent the number of pixels corresponding to each gray value t. The adjacent h(t)s are subtracted from each other to obtain g(t)=h(t)-h(t-1); at this time, if g(t-2)+g(t-1)+g(t+1)+g(t+2)≥0, then g(t)=1; if g(t-2)+g(t-1)+g(t+1)+g(t+2)<0, then g(t)=-1, and g(t) is obtained. At this time, g(t) is traversed, and the number of consecutive positive 1s and negative 1s from negative to positive is added. The turning point from negative to positive with the largest sum is the first set threshold T. The pixel values ​​inside each region are compared with the threshold T. If the pixel value is greater than T, the region is marked as 1; if the pixel value is less than T, the region is marked as 0. The mark value of 1 indicates that the infrared image region may contain a target, and the mark value of 0 indicates that the infrared image region does not contain a target. Threshold segmentation can be used to extract feature points with large signal grayscale changes in the infrared image, and the threshold segmentation of the infrared image can be completed.

[0056] The high-brightness area of ​​the infrared image, that is, the infrared image area with a mark value of 1, is obtained through threshold segmentation. The probability of containing the target in each infrared image area is calculated, and the probability of containing the target in the infrared image area is compared with the set threshold. If it is greater than the set threshold, it is judged that the infrared image area contains the candidate target. If it is less than the set threshold, it is judged that the infrared image area does not contain the target, and these infrared image areas that do not contain the target are eliminated.

[0057] The infrared image area containing the candidate target is projected and segmented, and the coordinate value of each candidate target area in the infrared image area is calculated, wherein each infrared image area may contain multiple targets, and the above is called a candidate target because it cannot be fully determined whether it is a real target through projection segmentation at this time. Based on the coordinate value of each candidate target area, the region of interest of the infrared image area is segmented to extract the region of interest, wherein each region of interest contains a candidate target.

[0058] S3, inputting the region of interest into a trained support vector machine to detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples containing infrared dim small targets with generated tracks.

[0059] It can be understood that the region of interest extracted in step S2 is input into the trained support vector machine, and the trained support vector machine outputs the position coordinate value of each candidate target in the infrared image to be detected, as well as the number of targets, to detect and classify the targets in the infrared image to be detected.

[0060] Among them, the target in the infrared image to be detected identified by the trained support vector machine cannot be completely determined, and needs to be determined through the matching association of multiple frames of continuous infrared images.

[0061] As an embodiment, inter-frame track association is performed on continuous multiple frames of infrared images, and based on the track association result, the authenticity of the existence of the target in the infrared image to be detected is confirmed, including: taking the infrared image to be detected as the current frame image, for each frame of the continuous multiple frames before and after the current frame image, the position coordinate value of each target in each frame image is output based on the trained support vector machine; if the position coordinate values ​​of multiple targets in two adjacent frames of a preset ratio in the multiple frames can be matched, the continuous multiple frames are successfully associated, and the target actually exists in the infrared image to be detected; otherwise, the continuous multiple frames fail to be associated, and the target does not exist in the infrared image to be detected.

[0062] Among them, the position coordinate values ​​of multiple targets in two adjacent frame images are matched in the following manner: for a pair of position coordinate values ​​of any target in two adjacent frame images, if the position coordinate value in the current frame image falls within a preset radius circular gate of the corresponding position coordinate value in the previous frame image, then the position coordinate value of any target is matched successfully; if the position coordinate values ​​of all targets in two adjacent frame images are matched successfully, then the two adjacent frame images are successfully associated.

[0063] Specifically, when matching and associating multiple frames of continuous infrared images, multiple frames of images before and after the infrared image to be detected are input into the trained support vector machine, and the position coordinate values ​​of multiple candidate targets in each frame of the image are input.

[0064] For the two adjacent frames of images, the position coordinate values ​​of the identified multiple candidate targets are matched accordingly. If they can all match, the two adjacent frames of images can be associated. For continuous multi-frame images, if a certain proportion of two frames of images can be successfully associated, it is considered that the continuous multi-frame images are successfully associated. Only then can the authenticity of the multiple candidate targets in the infrared image to be detected be confirmed, and the detection and classification of the targets in the infrared image to be detected can be realized.

[0065] It should be noted that for the training samples of the support vector machine, before the support vector machine is trained, the weak targets in the infrared image are judged and extracted by the threshold criterion method, that is, for the extracted region of interest, the grayscale value of the pixels in the region of interest is used to determine whether the target actually exists in the region of interest. In this way, the infrared images in the training samples are annotated. When the support vector machine is trained, the trained support vector machine is used to detect and classify the targets in the infrared image, and the real existence of the targets in the infrared image is determined by the track association method.

[0066] Based on the successfully associated continuous multi-frame images, the track of the weak infrared target is generated; based on the infrared image of the weak infrared target with the generated track, the initial training sample is updated in real time to obtain the real-time training sample of the support vector machine.

[0067] When the support vector machine is subsequently trained, the support vector machine is trained based on the updated real-time training samples. It should be noted that the embodiment of the present invention is an online training process, that is, the training process and the detection and recognition process are carried out simultaneously, and the training samples are updated in real time using the detection and recognition results, and the support vector machine is continued to be trained based on the updated training samples, and the classification hyperplane parameters of the support vector machine are adjusted in real time, and the support vector machine is optimized in real time to improve the detection and recognition rate of the support vector machine.

[0068] Embodiment 2

[0069] A small target detection method based on infrared sequence images includes: based on the gray value of each pixel in the infrared image to be detected, background suppression is performed on the infrared image to be detected; threshold segmentation is performed on the infrared image to be detected after background suppression, and a region of interest is extracted therein, wherein the region of interest is a region that may contain a target; the region of interest is input into a trained support vector machine, and the target in the infrared image to be detected is detected and classified, wherein the support vector machine is trained based on training samples of infrared small targets containing generated tracks.

[0070] See also Figure 2 The small target detection method provided by the embodiment of the present invention mainly includes two parts: online training and online operation. The online training part is to use the infrared image of the infrared small target including the generated track as the training sample to train the support vector machine. The online operation part is to first suppress the background of the input infrared image, take the threshold to segment the residual image after the background suppression, and obtain the local area of ​​interest. Before the support vector machine completes the training, the threshold judgment method is used to complete the target detection. After the support vector machine completes the training, the target is sent to the support vector machine for classification, and the local area identified as containing the infrared small target is the single-frame detection result; finally, the target detected in the single frame is track-associated to generate the target track, and the target track is used to update the training sample to continuously train and optimize the support vector machine.

[0071] Among them, because of the complex scenes and changeable target appearance of sequence images, it is difficult to establish a universal detection method suitable for various environments. Motion target detection and recognition based on statistical learning can fully and specifically learn the temporal and spatial characteristics of the target, with high recognition accuracy and low false alarm rate.

[0072] The idea of ​​most existing statistical learning algorithms is to use offline training images as prior knowledge to train classifiers. On the one hand, the offline training machine needs to prepare a large number of samples, and the samples need to be manually labeled. On the other hand, due to the limitations of sample space and quantity, when the scene and target characteristics change, the original classifier parameters will cause the extracted target recognition rate to decrease. The embodiment of the present invention uses track targets with high posterior probability as online training samples, without the need for manual labeling, and can update the classifier parameters in real time after the scene and target characteristics change, effectively improving the target detection recognition rate.

[0073] Embodiment 3

[0074] A small target detection system based on infrared sequence images, see Figure 3 The small target detection system includes a background suppression module 301, an extraction module 302 and a target detection module 303, wherein:

[0075] The background suppression module 301 is used to suppress the background of the infrared image to be detected based on the grayscale value of each pixel in the infrared image to be detected; the extraction module 302 is used to perform threshold segmentation on the infrared image to be detected after background suppression, and extract the region of interest therein, wherein the region of interest is a region that may contain a target; the target detection module 303 is used to input the region of interest into a trained support vector machine, and detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples of infrared weak and small targets containing generated tracks.

[0076] It can be understood that the small target detection system based on infrared sequence images provided by the present invention corresponds to the small target detection method based on infrared sequence images provided in the aforementioned embodiments. The relevant technical features of the small target detection system based on infrared sequence images can refer to the relevant technical features of the small target detection method based on infrared sequence images, which will not be repeated here.

[0077] Embodiment 4

[0078] See also Figure 4 , Figure 4 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 4As shown, an embodiment of the present invention provides an electronic device 400, including a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and executable on the processor 420. When the processor 420 executes the computer program 411, the following steps are implemented: background suppression is performed on the infrared image to be detected based on the grayscale value of each pixel in the infrared image to be detected; threshold segmentation is performed on the infrared image to be detected after background suppression, and a region of interest is extracted therein, wherein the region of interest is a region that may contain a target; the region of interest is input into a trained support vector machine, and the target in the infrared image to be detected is detected and classified, wherein the support vector machine is trained based on training samples of infrared weak targets containing generated tracks.

[0079] Embodiment 5

[0080] See also Figure 5 , Figure 5 Schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Figure 5 As shown, this embodiment provides a computer-readable storage medium 500, on which a computer program 511 is stored. When the computer program 511 is executed by a processor, the following steps are implemented: based on the grayscale value of each pixel in the infrared image to be detected, background suppression is performed on the infrared image to be detected; threshold segmentation is performed on the infrared image to be detected after background suppression, and a region of interest is extracted therein, wherein the region of interest is a region that may contain a target; the region of interest is input into a trained support vector machine, and the target in the infrared image to be detected is detected and classified, wherein the support vector machine is trained based on training samples of infrared weak and small targets containing generated tracks.

[0081] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0082] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0087] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A small target detection method based on infrared sequence images, characterized in that: include: Based on the gray value of each pixel in the infrared image to be detected, background suppression is performed on the infrared image to be detected; Performing threshold segmentation on the infrared image to be detected after background suppression, and extracting the region of interest therein, wherein the region of interest is a region that may contain the target; Inputting the region of interest into a trained support vector machine to detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples containing infrared dim small targets with generated tracks; The step of performing threshold segmentation on the infrared image to be detected after background suppression to extract the region of interest therein comprises: Segmenting the infrared image to be detected after background suppression based on a preset size window to obtain a plurality of infrared image regions, wherein each infrared image region contains a plurality of pixels; Perform a traversal operation on each infrared image area, use h(t) to represent the number of pixels corresponding to each gray value t, subtract adjacent h(t)s from each other, and obtain g(t)=h(t)-h(t-1); at this time, if g(t-2)+g(t-1)+g(t+1)+g(t+2)≥0, then g(t)=1; if g(t-2)+ g(t-1)+ g(t+1)+g(t+2)<0, then g(t)=-1, and g(t) is obtained. At this time, traverse g(t), add the number of consecutive positive 1s and negative 1s from negative to positive, and the turning point from negative to positive with the largest sum is the first set threshold T; traverse the pixel values ​​inside each infrared image area and compare them with the threshold T. If it is greater than T, the mark value of the area is 1; if it is less than T, the mark value is 0, wherein the mark value 1 indicates that the infrared image area may contain a target, and the mark value 0 indicates that the infrared image area does not contain a target; Performing image segmentation on the infrared image region with a mark value of 1, and calculating the coordinate value of each candidate target region in the infrared image region; Based on the coordinate value of each candidate target region, the region of interest of the infrared image region is segmented to extract the regions of interest, wherein each region of interest contains a candidate target.

2. The small target detection method according to claim 1, characterized in that: The method of performing background suppression on the infrared image to be detected based on the grayscale value of each pixel in the infrared image to be detected includes: For any pixel in the infrared image to be detected, a filtering window is selected around it; The grayscale value of each pixel in the filter window is arranged in order according to size, and the grayscale value median of the arrangement result is used to replace the grayscale value of any pixel; Each pixel in the infrared image to be detected is traversed, and the gray value of each pixel is changed to complete the background suppression of the infrared image to be detected.

3. The small target detection method according to claim 1, characterized in that: The image segmentation is performed on the infrared image region with a mark value of 1, and the coordinate value of each candidate target region in the infrared image region is calculated, and the above also includes: For the infrared image regions with a mark value of 1, the probability of containing a target in each infrared image region is calculated, and when the probability of containing a target is greater than a second set threshold, it is determined that the infrared image region contains a candidate target, otherwise, the candidate target is not contained; Accordingly, image segmentation is performed on the infrared image region containing the candidate target.

4. The small target detection method according to claim 1, characterized in that: After inputting a plurality of interest regions extracted from the infrared image to be detected into a trained support vector machine, the support vector machine outputs a position coordinate value of each candidate target in the infrared image to be detected; The continuous multiple frames of infrared images are subjected to inter-frame track association, and based on the track association result, the authenticity of the target in the infrared image to be detected is confirmed.

5. The small target detection method according to claim 4, characterized in that: The continuous multiple frames of infrared images are subjected to inter-frame track association, and based on the track association result, the authenticity of the target in the infrared image to be detected is confirmed: The infrared image to be detected is used as the current frame image, and for each of the multiple consecutive frames of images before and after the current frame image, the position coordinate value of each target in each frame image is output based on the trained support vector machine; If the position coordinate values ​​of multiple targets in two adjacent frames of images with a preset ratio in the multiple frames of images can match, the continuous multiple frames of images are successfully associated, and the target actually exists in the infrared image to be detected; Otherwise, the association of the continuous multiple frames of images fails, and there is no target in the infrared image to be detected.

6. The small target detection method according to claim 5, characterized in that: The position coordinate values ​​of multiple targets in two adjacent frames of images are matched in the following way: For a pair of position coordinate values ​​of any target in two adjacent frames of images, if the position coordinate value in the current frame of image falls within a preset radius circular gate of the corresponding position coordinate value in the previous frame of image, the position coordinate value of any target is successfully matched; If the position coordinate values ​​of all targets in two adjacent frames of images are matched successfully, the two adjacent frames of images are successfully associated.

7. The small target detection method according to claim 5 or 6, characterized in that: Also includes: Generate the track of weak infrared targets based on the successfully associated continuous multi-frame images; Based on the infrared image of the infrared dim small target generating the track, the initial training sample is updated in real time to obtain the real-time training sample of the support vector machine; Accordingly, the support vector machine is trained based on the real-time training samples.

8. A small target detection system based on infrared sequence images, characterized in that: include: A background suppression module, used for performing background suppression on the infrared image to be detected based on the gray value of each pixel in the infrared image to be detected; An extraction module is used to perform threshold segmentation on the infrared image to be detected after background suppression, and extract the region of interest therein, wherein the region of interest is a region that may contain a target; A target detection module, used for inputting the region of interest into a trained support vector machine to detect and classify the target in the infrared image to be detected, wherein the support vector machine is trained based on training samples containing infrared dim small targets with generated tracks; The step of performing threshold segmentation on the infrared image to be detected after background suppression to extract the region of interest therein comprises: Segmenting the infrared image to be detected after background suppression based on a preset size window to obtain a plurality of infrared image regions, wherein each infrared image region contains a plurality of pixels; Perform a traversal operation on each infrared image area, use h(t) to represent the number of pixels corresponding to each gray value t, subtract adjacent h(t)s from each other, and obtain g(t)=h(t)-h(t-1); at this time, if g(t-2)+g(t-1)+g(t+1)+g(t+2)≥0, then g(t)=1; if g(t-2)+ g(t-1)+ g(t+1)+g(t+2)<0, then g(t)=-1, and g(t) is obtained. At this time, traverse g(t), add the number of consecutive positive 1s and negative 1s from negative to positive, and the turning point from negative to positive with the largest sum is the first set threshold T; traverse the pixel values ​​inside each infrared image area and compare them with the threshold T. If it is greater than T, the mark value of the area is 1; if it is less than T, the mark value is 0, wherein the mark value 1 indicates that the infrared image area may contain a target, and the mark value 0 indicates that the infrared image area does not contain a target; Performing image segmentation on the infrared image region with a mark value of 1, and calculating the coordinate value of each candidate target region in the infrared image region; Based on the coordinate value of each candidate target region, the region of interest of the infrared image region is segmented to extract the regions of interest, wherein each region of interest contains a candidate target.

9. A computer-readable storage medium, characterized in that: A computer management program is stored thereon, and when the computer management program is executed by the processor, the steps of the method for detecting small targets based on infrared sequence images as described in any one of claims 1 to 7 are implemented.

Images