A camera automatic target tracking method and system
By analyzing lighting and shadow features and combining them with target recognition algorithms, the problem of inaccurate target recognition by traditional cameras in complex scenes has been solved, and efficient automatic target tracking has been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG TUSHENG ULTRA HD INNOVATION CENT CO LTD
- Filing Date
- 2026-05-14
- Publication Date
- 2026-07-14
AI Technical Summary
Traditional video image target recognition algorithms struggle to accurately identify targets in complex scenes, resulting in poor tracking performance, especially when the subject is moving, which can easily lead to misjudgments.
By analyzing the illumination and shadow features of the area captured by the camera, the probability of the target's presence in each frame of the image is constructed. Automatic tracking is then performed using a target recognition algorithm, and corrections are made using illumination concentration and target compensation coefficients to improve target recognition accuracy.
It improves the accuracy and efficiency of camera target tracking, enabling accurate identification and tracking of subjects in complex scenes and reducing misjudgments.
Smart Images
Figure CN122391280A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of video target tracking technology, specifically to a method and system for automatic target tracking with a camera. Background Technology
[0002] Virtual filmmaking, a revolutionary advancement in the film and television industry, is fundamentally based on the deep integration of real-time rendering engines such as Unreal Engine, high-precision camera tracking technology, and LED backdrops. This integration creates a visual environment that allows creators to observe the final or near-final visual effects on set in real time. This technological paradigm enables real-time visual feedback and interaction from dynamic pre-visualization to the shooting process, allowing creators to make artistic decisions regarding composition, lighting, and performance scheduling based directly on the real-time images of the subjects and the virtual scene. This significantly enhances creative freedom and production efficiency, and brings much of the work traditionally done in post-production special effects compositing to the live-action shooting stage.
[0003] However, during video recording, the target is not stationary. When the subject moves within the camera's field of view, video image recognition algorithms are needed to identify the subject and determine its direction of movement. This allows automated shooting equipment to be controlled to move, ensuring the subject remains within the camera's field of view. Common tracking methods include mechanical tracking, optical tracking, and image recognition tracking. Image recognition tracking generally does not require pre-planning of the shooting environment. It can obtain sufficient feature points through image recognition algorithms to identify the target. Therefore, image recognition tracking has a large shooting range and relatively high tracking accuracy, and is currently widely used in sports events, aerial landscape photography, and other fields.
[0004] Traditional video image target recognition algorithms typically identify the target based on the feature points of the target object to determine its position in the video image, thereby enabling localization, recognition, and tracking. However, the complexity of the shooting location and layout, as well as the uncertainty of the object's movement, make it difficult to accurately identify the target in complex scenes. This leads to inaccurate prediction of the target's actual spatial location, or misjudgment of the target due to similar backgrounds, ultimately affecting the tracking performance. Summary of the Invention
[0005] To address the aforementioned technical problems, the purpose of this application is to provide a method and system for automatic target tracking with a camera, the specific technical solution of which is as follows:
[0006] This application provides an automatic target tracking method for a camera, including the following steps:
[0007] The system acquires video image data of the target area using a camera, and then performs edge extraction and region segmentation.
[0008] Based on the edge distribution and grayscale distribution differences within each region of a single-frame video image, the illumination concentration of each region is obtained.
[0009] Suspected shadow regions are extracted based on the grayscale features of each region in a single frame video image. The target compensation coefficient for each region is obtained by considering the distance relationship between each region and the suspected shadow region, the degree of grayscale difference between each region and its neighboring regions, and the distribution characteristics of the inner edge width of the neighboring regions of each region. The target compensation coefficient for each region is then corrected by considering the degree of difference between the grayscale fluctuations of each region in adjacent frames of video images.
[0010] By using the illumination concentration of each region and the corrected target compensation coefficient, the target presence coefficient of each region in a single frame video image is obtained, so as to determine the region containing the target in the single frame image, and thus achieve target tracking.
[0011] Preferably, for the edges in each region of a single frame video image, the total number of pixels contained in each edge is counted as the edge length of each edge. The edge length of all edges in each region is thresholded and segmented, and the segmentation threshold is recorded as the first threshold. If the edge length is greater than or equal to the first threshold, the corresponding edge is regarded as a long edge; otherwise, the corresponding edge is regarded as a short edge.
[0012] Preferably, the method for obtaining the illumination concentration of each region in a single frame video image is as follows: calculate the difference between the mean gray value of each region and the minimum mean gray value of all regions, and calculate the proportion of the number of short edges in each region to the total number of edges. The product of the mean edge length of all long edges in each region and the difference result, divided by the proportion, is used as the illumination concentration of each region.
[0013] Preferably, the extraction process of the suspected shadow region is as follows: threshold segmentation is performed on the gray-scale mean of each region in a single frame video image, and the segmentation threshold is recorded as the second threshold. Regions with a gray-scale mean less than or equal to the second threshold are regarded as suspected shadow regions.
[0014] Preferably, the method for obtaining the target compensation coefficient for each region is as follows:
[0015] The difference in the mean grayscale values of the left and right neighboring regions of each region is statistically analyzed, as well as the sum of the mean edge widths of the neighboring regions of each region. The minimum distance between the center point of each region and the center points of all suspected shadow regions is also calculated. The target compensation coefficient of each region is positively correlated with the difference in the mean grayscale values and the sum, and negatively correlated with the minimum distance.
[0016] Preferably, the eight neighboring regions of each region are taken as the neighboring regions of each region. For the neighboring regions of the i-th region, the neighboring regions whose center point x-coordinate is less than the center point x-coordinate of the i-th region are taken as the left neighboring regions of the i-th region, and the neighboring regions whose center point x-coordinate is greater than the center point x-coordinate of the i-th region are taken as the right neighboring regions of the i-th region. If the center point x-coordinate of the region is equal to the center point x-coordinate of the i-th region, the corresponding neighboring region is not included in the analysis of the left and right neighboring regions of the i-th region.
[0017] Preferably, the step of correcting the target compensation coefficient for each region further includes:
[0018] In the formula, Let be the correction target compensation coefficient for the i-th region. Let be the target compensation coefficient for the i-th region. Let be the variances of the grayscale values of the i-th region in the s-th and s-1-th frames, respectively, and S be the total number of frames in the video data.
[0019] Preferably, the target presence coefficient of each region in the single-frame video image is the normalized result of the product of the illumination concentration degree of each region and the corrected target compensation coefficient.
[0020] Preferably, the step of determining the region containing the target in a single frame image further includes: performing threshold segmentation on the target presence coefficient of all regions, recording the segmentation threshold as the third threshold, and if the target presence coefficient is greater than or equal to the third threshold, then the corresponding region is the region where the target is located.
[0021] This application also provides an automatic target tracking system for a camera, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described automatic target tracking methods for a camera.
[0022] As can be seen from the above, the automatic target tracking method and system for cameras provided in this application have at least the following beneficial effects:
[0023] This application analyzes the light concentration characteristics of the area where the camera is shooting the object and the shadow characteristics of the area where the object is shooting when the shooting situation changes, thereby constructing the probability that there is a target in each area of each frame image, thus selecting the area in a single frame image where there may be a shooting target, and then combining the target recognition algorithm to identify the shooting object, and then controlling the camera to perform fully automatic tracking and shooting according to the movement of the shooting object.
[0024] This application addresses the problem that traditional target recognition algorithms are easily affected by the lighting conditions of the shooting environment in complex scenes, resulting in low accuracy and efficiency in target recognition. This application improves the accuracy of camera target tracking. Attached Figure Description
[0025] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 A flowchart illustrating the steps of an automatic target tracking method for a camera provided in this application. Detailed Implementation
[0027] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a camera automatic target tracking method and system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0028] Unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0029] The following description, in conjunction with the accompanying drawings, details the specific scheme of the automatic target tracking method and system for cameras provided in this application.
[0030] Please see Figure 1 The diagram illustrates a flowchart of a camera automatic target tracking method according to an embodiment of this application, including the following steps:
[0031] Step 1: Acquire video image data of the target area using a camera, and perform edge extraction and region segmentation.
[0032] In this embodiment, a high-resolution sensor mounted on a gimbal camera is first used to achieve 4K or 8K super-resolution output. A data television stabilizer and a DSP+FPGA architecture image processing module are integrated. This camera captures video image data of the target, obtaining multiple frames of video image data. Each frame of video image data is then used as input, converted to grayscale image data using grayscale averaging, and an edge detection algorithm is employed. Preferably, the Sobel operator is used in this embodiment to extract all edges in each frame. Simultaneously, a single frame is divided into z equally sized regions. In this embodiment, z=64, but the implementer can adjust this value according to actual conditions.
[0033] Thus, the video image data of the target can be obtained according to the above process in this embodiment.
[0034] Step 2: Based on the edge distribution and grayscale distribution differences within each region of a single-frame video image, obtain the illumination concentration of each region.
[0035] During the filming process, in order to ensure clear footage and sufficient lighting, the filming crew usually sets up spotlights and fill lights at various angles. This makes the outline of the subject clear and the details rich, avoiding problems such as loss of detail, color deviation, or decreased contrast due to insufficient lighting. Therefore, the location of the subject is usually well-lit throughout the filming process. In contrast, most areas other than the target area are the background, which contains less light. Compared to the target area, it has less detail and lower overall brightness. Moreover, due to the supplemental lighting of the subject in the filming scene, the shadows cast by the subject appear in the background area, making some background areas even darker.
[0036] Specifically, for the area where the subject is located, the overall brightness of the area is higher than that of other areas due to the illumination of spotlights and fill lights. Furthermore, due to the sufficient lighting, the details of the subject are clearer. Also, due to the sufficient lighting, there may be some highlight areas or light spots in the subject. The above characteristics are specifically manifested in that the overall gray value of a single area is more different from that of other areas in a single frame of video image, and there are more complete and longer edges, fewer short and scattered edges, smaller edge stripe width, and more points with extremely high gray values in the area.
[0037] To characterize the aforementioned features, for a single region in a single frame of video image, the total number of pixels contained within each edge is counted, and this total number is used as the edge length. Further, using the edge length data of all edges within each region as input, the Otsu thresholding method is used to output a segmentation threshold for the edge length, denoted as the first threshold. When the edge length is greater than or equal to this threshold, the edge is considered a long edge; otherwise, it is considered a short edge. Additionally, using the grayscale values of all pixels in the image as input, cross-validation is used to output a segmentation threshold for the pixel grayscale values. When the grayscale value of a pixel is greater than or equal to the segmentation threshold, the pixel is considered more likely to be a highlight or spot caused by illumination and is classified as a spot; otherwise, it is classified as a normal pixel.
[0038] Based on the above analysis, in this embodiment, the illumination concentration degree of each region will be constructed to characterize the illumination characteristics of a single region as the target area being illuminated by a spotlight or supplementary light. Preferably, in this embodiment, the difference between the mean gray value of each region and the minimum mean gray value of all regions will be calculated, and the proportion of the number of short edges in each region to the total number of edges will be calculated. The product of the mean edge length of all long edges in each region and the difference result, divided by the proportion, will be used as the illumination concentration degree of each region.
[0039] Among them, when the overall illuminance of the i-th region is higher than that of the other regions, and the proportion of short edges in the region is smaller and the average length of the long edges is longer, it indicates that the corresponding position in the region receives higher illuminance and the edge stripes are clearer. It is more likely that the area where the subject is located is illuminated by spotlights and fill lights, and the subject is more likely to be present in the region.
[0040] Step 3: Extract suspected shadow regions based on the grayscale features of each region in a single frame video image. Calculate the target compensation coefficient for each region by considering the distance relationship between each region and the suspected shadow region, the degree of grayscale difference between each region and its neighboring regions, and the distribution characteristics of the inner edge width of the neighboring regions. Then, adjust the target compensation coefficient for each region by considering the degree of difference in grayscale fluctuations between adjacent frames of video image.
[0041] Because the shooting requirements vary under different shooting conditions, such as the need to adjust focus, shooting angle, or change scenes and move the subject at certain times, the lighting of the spotlight or fill light will also change. This will lead to a certain deviation in the lighting effect when shooting the subject, and the lighting conditions of the subject area and the background area may be similar. Therefore, if the calculation is performed only in step S2, some areas where the subject is located may be misjudged as having no subject, resulting in a deviation in the overall target recognition and affecting the camera's tracking response time. Therefore, further calculation is required.
[0042] Specifically, during the shooting process, the overall light source is usually focused on the location of the subject. Therefore, shadow areas will exist around the subject area due to the lighting, and these shadow areas will be close to the subject area. Furthermore, for the subject area, due to good focusing and overall illumination, the edge stripes in the subject area are highly detailed, and the overall edge width is small. In addition, for the edge area of the subject's location, since one side is the subject area and the other side is the background area, the difference between the two sides of this area is quite obvious. Specifically, the closer the overall distance between the subject area and the shadow area, the smaller the edge width of the surrounding areas, and there is a certain difference between the left and right sides.
[0043] To characterize the above features, for regions in a single frame video image, the gray-scale mean of each region is first calculated. The gray-scale mean of all regions is then used as input. Using cross-validation, the output is a segmentation threshold of the region's gray-scale mean, denoted as the second threshold. When the gray-scale mean of a single region is less than or equal to the second threshold, the corresponding region contains a shadow region and is considered a suspected shadow region. The remaining regions are considered ordinary regions.
[0044] Based on the above analysis, and considering the grayscale differences between neighboring areas and the average distribution of edge widths within those areas, combined with the distribution characteristics of each area relative to suspected shadow areas, a target compensation coefficient is constructed for each area. This coefficient is used to compensate for the possibility of a subject being captured within each area. The target compensation coefficient is constructed by: calculating the mean grayscale differences between the left and right neighboring areas of each area, summing the mean edge widths of the neighboring areas of each area, and calculating the minimum distance between the center point of each area and the center points of all suspected shadow areas. The target compensation coefficient for each area is positively correlated with the mean grayscale differences and the summation, and negatively correlated with the minimum distance. It should be noted that the positive correlation means the dependent variable increases as the independent variable increases and decreases as the independent variable decreases, while the negative correlation means the dependent variable decreases as the independent variable increases and increases as the independent variable decreases. The specific calculation relationship is not specifically limited in this embodiment.
[0045] Preferably, in this embodiment, the specific formula for calculating the target compensation coefficient is as follows: In the formula, Let be the target compensation coefficient for the i-th region. Let be the minimum Euclidean distance between the center point of the i-th region and the center points of all suspected shaded regions. Let J be the average edge width of the j-th neighboring regions of the i-th region, and J be the total number of neighboring regions of the i-th region. Let be the average gray values of the left and right neighboring regions of the i-th region, respectively. To avoid constants with a denominator of zero, the value range is (0,0,1), and in this embodiment it is set to 0.001.
[0046] It should be noted that the eight neighboring regions of each region are the adjacent regions of each region. Also, when the i-th region is a suspected shaded region, Take the maximum value of the Euclidean distance between the center points of all regions in a single frame image.
[0047] Preferably, in this embodiment, the left and right sides of a single region mainly mean: taking the i-th region as an example, for the neighboring regions of the i-th region, the neighboring regions whose center coordinates are less than the center coordinates of the i-th region are designated as the left neighboring regions of the i-th region, and the neighboring regions whose center coordinates are greater than the center coordinates of the i-th region are designated as the right neighboring regions of the i-th region. Specifically, if the center coordinates of a region are equal to the center coordinates of the i-th region, then the corresponding neighboring region does not participate in the analysis of the left and right neighboring regions of the i-th region; that is, this neighboring region does not participate in the calculation and analysis. In practical application scenarios, the implementer can decide on the setting of the left and right neighboring regions for each region; this embodiment does not impose any special restrictions on this.
[0048] It is understandable that the closer the i-th region is to the suspected shadow region, the smaller the edge width in the neighboring regions of the region, and the greater the difference between the two sides of the region's neighborhood, the more likely there is a subject in the location of the region, and the more the region should be corrected to the target region.
[0049] Furthermore, during the shooting process, the area where the subject is located is generally well-lit. However, as the subject moves, the positions of the spotlights and fill lights change, resulting in fluctuations in brightness in certain areas. These fluctuations manifest as bright areas (under illumination), dim areas (as the subject moves away from the light source), and bright areas (as the subject returns to the light source). Therefore, areas with significant differences in overall brightness within a short period are more likely to contain the subject. Based on the above analysis, the target compensation coefficient is further modified. In this embodiment, the calculation formula for the modified target compensation coefficient for each area is as follows:
[0050] In the formula, Let be the correction target compensation coefficient for the i-th region. Let be the target compensation coefficient for the i-th region. Let be the variances of the grayscale values of the i-th region in frames s and s-1, respectively, and S be the total number of frames in the video data. The larger the overall brightness variation difference in the i-th region, the larger the target compensation coefficient, indicating that the brightness variation in that region is more likely caused by the movement of the subject, and the more likely a subject is present in that region.
[0051] Step 4: By using the illumination concentration of each region and the corrected target compensation coefficient, the target presence coefficient of each region in a single frame video image is obtained, so as to determine the region containing the target in the single frame image, and thus realize target tracking.
[0052] According to the above process in this embodiment, the correction target compensation coefficient of each region in a single frame video image can be obtained. Furthermore, the presence or absence of a target in each region can be determined and identified by using the correction target compensation coefficient.
[0053] Specifically, the target presence coefficient of each region in a single frame image is obtained through the following relationship, which is used to characterize the probability that a single region is the location of a target. Preferably, in this embodiment, the normalized result of the product of the illumination concentration degree of each region and the corrected target compensation coefficient is used as the target presence coefficient of each region in a single frame video image. Among them, the higher the illumination concentration degree of the i-th region in a single frame image and the greater the probability of the target presence, the more likely that the region is the location of the shooting object.
[0054] Using the above method, the target presence coefficient of each region in each frame of the image can be calculated. Taking the target presence coefficient of all regions as input, cross-validation is used to output the segmentation threshold of the target presence coefficient, which is denoted as the third segmentation threshold. For each region, when the target presence coefficient is greater than or equal to the third segmentation threshold, it is characterized as the region where the target is located. During the target tracking process, attention should be paid to the changes in this region so that the camera can track and capture the target.
[0055] The specific process of camera target tracking is existing technology, and implementers can choose according to the actual situation. This embodiment does not impose any special restrictions on it. In this embodiment, specifically: through the above process, it can be determined that there is a shooting object in each region of each frame of video image. Further, the camera control system combines the target recognition algorithm to perform target recognition on the region where the shooting object exists in each frame of image, and obtain the location of the shooting target. Combined with the inter-frame prediction algorithm, this embodiment uses the Kalman filter prediction algorithm to predict the displacement and dynamic feature changes of the shooting object. The specific process of using the Kalman filter to predict the centroid in this embodiment is as follows:
[0056] 1) Initialize the state vector, covariance matrix, motion model matrix, and noise matrix;
[0057] 2) Prediction stage: The state vector is updated by the state transition matrix to achieve state prediction; the covariance matrix is updated to achieve covariance prediction.
[0058] 3) Update phase: Calculate the Kalman gain based on the predicted covariance and observation noise, and correct the predicted state using the observed values to achieve state update; update the covariance matrix, taking into account the influence of the Kalman gain, to achieve covariance update;
[0059] 4) Extract prediction results: Obtain the predicted centroid position;
[0060] 5) Repeat process 2) and 3), repeating the prediction and update steps in each video image frame to continuously track the centroid of the target.
[0061] It should be noted that the specific process of predicting the displacement and dynamic feature changes of the subject using a Kalman filter in this embodiment is a well-known technique to those skilled in the art, and will not be elaborated upon here. Furthermore, the coordinate deviation between the predicted target position coordinates and the current camera shooting center position is calculated, thereby outputting the shooting adjustment angle, displacement direction, and displacement distance required by the current camera.
[0062] Furthermore, after obtaining the camera's adjustment parameters, the camera control system sends corresponding control information to the automated equipment carrying the camera. The actuators adjust according to the control information, such as controlling the movement of the track and the angle of the robotic arm, and finally adjust the camera's shooting position to track the target.
[0063] Based on the same inventive concept as the above method, this application also provides an automatic target tracking system for a camera, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described automatic target tracking methods for a camera.
[0064] It is understood that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0065] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0066] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the protection scope of this application.
Claims
1. A method for automatic target tracking with a camera, characterized in that, Includes the following steps: The system acquires video image data of the target area using a camera, and then performs edge extraction and region segmentation. Based on the edge distribution and grayscale distribution differences within each region of a single-frame video image, the illumination concentration of each region is obtained. Suspected shadow regions are extracted based on the grayscale features of each region in a single frame video image. The target compensation coefficient for each region is obtained by considering the distance relationship between each region and the suspected shadow region, the degree of grayscale difference between each region and its neighboring regions, and the distribution characteristics of the inner edge width of the neighboring regions of each region. The target compensation coefficient for each region is then corrected by considering the degree of difference between the grayscale fluctuations of each region in adjacent frames of video images. By using the illumination concentration of each region and the corrected target compensation coefficient, the target presence coefficient of each region in a single frame video image is obtained, so as to determine the region containing the target in the single frame image, and thus achieve target tracking.
2. The automatic target tracking method for a camera as described in claim 1, characterized in that, For each region of a single-frame video image, the total number of pixels contained in each edge is counted and used as the edge length of each edge. The edge length of all edges in each region is thresholded and the threshold is recorded as the first threshold. If the edge length is greater than or equal to the first threshold, the corresponding edge is regarded as a long edge; otherwise, the corresponding edge is regarded as a short edge.
3. The automatic target tracking method for a camera as described in claim 2, characterized in that, The method for obtaining the illumination concentration of each region in a single frame video image is as follows: calculate the difference between the mean gray value of each region and the minimum mean gray value of all regions, and calculate the proportion of the number of short edges in each region to the total number of edges. Then, multiply the mean edge length of all long edges in each region by the difference and divide the product by the proportion to obtain the illumination concentration of each region.
4. The automatic target tracking method for a camera as described in claim 1, characterized in that, The extraction process of the suspected shadow region is as follows: the gray-scale mean of each region in a single frame video image is thresholded, and the threshold is recorded as the second threshold. Regions with a gray-scale mean less than or equal to the second threshold are identified as suspected shadow regions.
5. The automatic target tracking method for a camera as described in claim 1, characterized in that, The method for obtaining the target compensation coefficient of each region is as follows: calculate the difference in grayscale mean between the left and right adjacent regions of each region, and the sum of the average edge widths of the adjacent regions of each region, and calculate the minimum distance between the center point of each region and the center points of all suspected shadow regions. The target compensation coefficient of each region is positively correlated with the difference in grayscale mean and the sum, and negatively correlated with the minimum distance.
6. The automatic target tracking method for a camera as described in claim 5, characterized in that, The eight neighboring regions of each region are taken as the neighboring regions of each region. For the neighboring regions of the i-th region, the neighboring regions whose center coordinates are less than the center coordinates of the i-th region are taken as the left neighboring regions of the i-th region, and the neighboring regions whose center coordinates are greater than the center coordinates of the i-th region are taken as the right neighboring regions of the i-th region. If the center coordinates of the region are equal to the center coordinates of the i-th region, the corresponding neighboring regions are not included in the analysis of the left and right neighboring regions of the i-th region.
7. The automatic target tracking method for a camera as described in claim 1, characterized in that, The correction of the target compensation coefficient for each region further includes: In the formula, Let be the correction target compensation coefficient for the i-th region. Let be the target compensation coefficient for the i-th region. Let be the variances of the grayscale values of the i-th region in the s-th and s-1-th frames, respectively, and S be the total number of frames in the video data.
8. The automatic target tracking method for a camera as described in claim 1, characterized in that, The target presence coefficient in each region of the single-frame video image is the normalized result of the product of the illumination concentration degree of each region and the corrected target compensation coefficient.
9. The automatic target tracking method for a camera as described in claim 1, characterized in that, The step of determining the region containing the target in a single frame image further includes: performing threshold segmentation on the target presence coefficient of all regions, recording the segmentation threshold as the third threshold, and if the target presence coefficient is greater than or equal to the third threshold, then the corresponding region is the region where the target is located.
10. A camera automatic target tracking system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the automatic target tracking method for a camera as described in any one of claims 1-9.