A thunderball-based coastal defense and private investigation evidence collection system

Abstract

The application discloses a kind of based on thunderball's coast defense private prosecution evidence obtaining system, including radar, ball machine, edge machine, client platform and network switch, radar and ball machine coaxial installation form thunderball front end of thunderball linkage, one edge machine is docked multiple thunderball front ends, multiple edge machines are communicated connection with client platform by network switch;Edge machine control thunderball front end adopts master-slave search type target tracking strategy or multi-target tracking strategy to track target, input tracking image into improved YOLOv7 network structure, output target class and target position;Master-slave search type target tracking is guided ball machine tracking with radar as main tracking, after discovering target, auxiliary tracking is carried out using target image position driving PTZ video tracking algorithm, and multi-target tracking strategy is carried out target tracking with fastest high-speed target as first priority;The application has the advantages that: multiple targets can be event analyzed, so as to carry out private prosecution evidence obtaining.

Description

Technical Field

[0001] This invention relates to intelligent anti-smuggling evidence collection equipment, and more specifically to a coastal defense anti-smuggling evidence collection system based on lightning balls. Background Technology

[0002] The Pearl River has numerous estuaries and a complex waterway network. Small vessels, especially speedboats and covered boats, often modified for propulsion, are small targets and move quickly, posing a significant challenge to anti-smuggling and evidence collection efforts. There is a need to design a maritime anti-smuggling system that can effectively track and collect evidence of smuggling by analyzing images of high-speed small vessels, thereby enhancing the coast guard's ability to prevent and control smuggling and maintain shipping order.

[0003] Chinese Patent Publication No. CN217739485U discloses an integrated PTZ camera, shore-based radar, and AIS linkage device, belonging to the field of surveillance technology. It includes an AIS device for receiving ship information, a PTZ camera body for acquiring image information, communication equipment, and a shore-based radar for tracking ships. The communication equipment is communicatively connected to the AIS device, shore-based radar, and PTZ camera body. The PTZ camera body is equipped with a preset module, a coordinate extraction module, and a control module. The preset module is used to preset the initial position of the PTZ camera body and the alarm area range. The coordinate extraction module is used to extract the coordinates of ships scanned by the shore-based radar and send them to the control module. The control module is used to continuously record the ship coordinates and the angle of the PTZ camera body. This patent application enables the PTZ camera to automatically monitor ships scanned by the shore-based radar and can adjust the camera attitude in real time to continuously maintain the camera pointing at the ship target, achieving video linkage tracking of ships on the water. However, this patent application only tracks the target and cannot perform event analysis on the target. In areas with many ships, simply tracking the target cannot determine whether the target is engaged in smuggling activities, thus this patent application cannot be used for anti-smuggling evidence collection. Summary of the Invention

[0004] The technical problem to be solved by this invention is how to provide a coastal defense anti-smuggling evidence collection system that can perform event analysis on multiple targets in order to carry out anti-smuggling evidence collection.

[0005] This invention solves the above-mentioned technical problems through the following technical means: a maritime defense anti-smuggling evidence collection system based on a radar ball, including a radar, a PTZ camera, an edge device, a client platform, and a network switch. The radar and PTZ camera are coaxially mounted vertically to form a radar ball front-end with radar ball linkage. The radar front-end is located above the area to be tracked. One edge device connects to multiple radar ball front-ends, and the multiple edge devices communicate with the client platform through the network switch. The edge device controls the radar ball front-end to track targets using a master-slave search-type target tracking strategy or a multi-target tracking strategy. The tracked images are input into an improved YOLOv7 network structure, and the target category and target location are output. The master-slave search-type target tracking uses radar-guided PTZ tracking as the main tracking method. After the target is detected, a video tracking algorithm driven by the target image position is used as an auxiliary tracking method. The multi-target tracking strategy prioritizes the fastest high-speed target for target tracking.

[0006] Beneficial effects: This invention provides a maritime anti-smuggling evidence collection system based on lightning ball. The master-slave search-style target tracking combines master tracking and auxiliary tracking to ensure that targets can be tracked and prevent missed captures or reports. The multi-target tracking strategy prioritizes the fastest high-speed targets, ensuring that high-speed targets that may be involved in smuggling are captured as soon as they appear. Furthermore, the tracked targets are input into an improved YOLOv7 network structure to determine the target category and location, effectively realizing multi-target event analysis and contributing to maritime anti-smuggling evidence collection.

[0007] Furthermore, the radar dome camera's front end is calibrated to unify the radar coordinate system with the PTZ camera coordinate system, achieving radar-dome camera linkage through the following formula.

[0008]

[0009]

[0010] Where x is the radial distance of the target detected by the radar, y is the tangential distance of the target detected by the radar, h is the height of the PTZ camera, p is the horizontal angle that the PTZ camera should rotate, and t is the complementary angle of the vertical angle that the PTZ camera should rotate.

[0011] Furthermore, the master-slave search-based target tracking strategy includes:

[0012] The system uses radar-guided tracking, also known as radar-ball linkage calculation, as the primary tracking point. Once the primary tracking detects a target, it switches to a video tracking algorithm driven by the target image position in the PTZ for auxiliary tracking. If the auxiliary tracking fails to detect a target, the PTZ camera zooms out to search for the target in the vicinity of the primary tracking point. Once a target is found, it switches to video tracking. If single-target tracking reaches a threshold, it switches to tracking other targets. If video tracking fails, it enters radar-guided tracking until the target disappears, at which point it exits tracking, forming a complete closed loop.

[0013] Furthermore, the target image position-driven PTZ video tracking algorithm includes:

[0014] Based on the distance of the target from the center of the image and the difference between the target size and the desired size, calculate the horizontal rotation angle, vertical rotation angle, and zoom value that the PTZ should adjust relative to the current position. The formula for adjusting the current position using PTZ is as follows:

[0015] d p =-(ab·Zoom)·d x

[0016] d t =-(cd·Zoom)·d y

[0017] d zoom =-e·d w

[0018] Where, d p d is the horizontal angle that the PTZ camera should adjust based on the current horizontal angle. t d is the vertical angle that the PTZ camera should adjust based on the current angle. zoom d represents the zoom level that the PTZ camera should adjust based on the current zoom level. x d represents the horizontal pixel value of the target center in the image, which is offset from the center of the image. y d represents the vertical pixel value of the target center in the image, which is offset from the center of the image. w The difference between the width of the target in the image and the expected width of the target category is denoted by a, b, c, d, and e, which are preset constants. Zoom is the current zoom value.

[0019] Furthermore, the multi-target tracking strategy includes:

[0020] The edge machine sets multi-target tracking priorities and tracks multiple targets in priority order. Low-priority targets can be interrupted by high-priority targets. When the tracking of a high-priority target reaches a threshold, it automatically switches to the next priority target. Among them, the fastest high-speed target is the first priority, non-high-speed targets that have not reached the set tracking threshold are the second priority, non-high-speed targets that have not been tracked are the third priority, and non-high-speed targets that have been tracked are sorted according to the last tracking time as the fourth priority.

[0021] Furthermore, the edge processor sets a threshold for the number of radar batches starting for a single target and a threshold for the offset distance from the starting point of the batch after the radar batch starts. Only targets exceeding these thresholds are considered valid targets. The edge processor also sets a defined value for the speed of high-speed targets and a threshold for the number of consecutive batches of high-speed targets to be locked. Only targets exceeding these thresholds are considered valid high-speed targets, in order to facilitate locking high-speed targets.

[0022] Furthermore, the improved YOLOv7 network structure includes:

[0023] The high downsampling rate detector heads responsible for large object detection in the original YOLOv7 network's three output scales (small, medium, and large) are removed. Simultaneously, an additional convolutional layer branch is added before the outputs of the small and medium object detector heads in the original YOLOv7 network. The output features of the small and medium object detector heads in the original YOLOv7 network are concatenated and fused with the features output by their corresponding convolutional layer branches before being output together. The small and medium object detector heads in the original YOLOv7 network are both part of the head region.

[0024] The feature dimension of the head region output is F = W*H / (S*S)*N*(K+4+1), where W and H are the width and height of the input image, S is the downsampling rate of the detection head, N is the number of anchor points, K is the total number of categories of the target to be detected, 4 represents the four-dimensional coordinates of the target in the plane, and 1 represents the confidence of the foreground object. Let the number of convolution kernels in the newly added convolutional layer branch be M. Then, the main feature of the original YOLOv7 network is convolved into an embedding vector of W*H / (S*S)*N*M dimensions after this convolution. This embedding vector is then concatenated and fused with the head region output feature F and output together.

[0025] Furthermore, the radar has a detection range of 1km, a detection accuracy of 1m, a maximum detection speed of 120km / h, and a coverage angle of 140 degrees. The PTZ camera has a range comparable to the radar, a magnification of 48 times, and is equipped with high-power laser illumination.

[0026] Furthermore, the client platform consists of a software system deployed on a server, which interfaces with the edge machine for real-time alarm information display, historical alarm event query, warning zone setting, and device management.

[0027] Furthermore, the system also includes a DTU device for sending real-time alarm information to a preset mobile phone in the form of SMS, MMS or voice broadcast.

[0028] The advantages of this invention are:

[0029] (1) This invention provides a maritime defense anti-smuggling evidence collection system based on lightning ball. The master-slave search target tracking combines master tracking and auxiliary tracking to ensure that the target can be tracked and prevent missed capture and reporting. The multi-target tracking strategy prioritizes the fastest high-speed target to ensure that high-speed targets that may be suspected of smuggling are captured as soon as they appear. The tracked target is input into the improved YOLOv7 network structure to determine the target category and target location, effectively realizing multi-target event analysis and helping to carry out maritime defense anti-smuggling evidence collection.

[0030] (2) The target image position driven PTZ video tracking algorithm of the present invention calculates the horizontal rotation angle, vertical rotation angle and magnification value that the PTZ should be adjusted relative to the current position based on the distance of the target in the image from the center of the screen and the difference between the target size and the expected size, so as to ensure that the captured target is always located in the center of the screen and effectively reduce the problem of missed capture and missed reporting.

[0031] (3) The present invention sets a multi-target tracking priority. Low-priority targets can be interrupted by high-priority targets. When the tracking of a high-priority target reaches the threshold, it automatically switches to the next priority target tracking. The fastest high-speed target is the first priority, ensuring that high-speed targets that may be suspected of smuggling are captured first, thus improving tracking efficiency and accuracy.

[0032] (4) This invention sets a threshold for the number of single-target radar batches and a threshold for the offset distance from the starting point of the batch after radar batching. Targets exceeding these thresholds are considered valid moving targets to overcome the false image caused by sea clutter and other noise. It also sets a defined speed value for high-speed targets and a threshold for the number of consecutive batches of high-speed targets to be locked. Targets meeting these thresholds are considered valid high-speed moving targets, facilitating locking on high-speed targets. This improves the accuracy of target tracking and avoids erroneous tracking.

[0033] (5) This invention improves the YOLOv7 structure, which not only simplifies the network structure and avoids the high complexity of computation caused by adding high-dimensional features, but also speeds up the network's operating efficiency. At the same time, the addition of convolutional layer branches achieves stronger representation capabilities without rebuilding the feature extraction network, reducing the demand for computing resources. Attached Figure Description

[0034] Figure 1 An architecture diagram of a maritime defense anti-smuggling evidence collection system based on lightning ball provided in an embodiment of the present invention;

[0035] Figure 2 A basic principle diagram of the lightning ball linkage of a coastal defense anti-smuggling evidence collection system based on lightning ball provided in an embodiment of the present invention;

[0036] Figure 3 A schematic diagram of a master-slave search-type target tracking algorithm used in a coastal defense anti-smuggling evidence collection system based on a lightning ball, provided in an embodiment of the present invention;

[0037] Figure 4 A schematic diagram of a multi-target tracking strategy for a coastal defense anti-smuggling evidence collection system based on a lightning ball, provided in an embodiment of the present invention;

[0038] Figure 5 A schematic diagram of the improved YOLOv7 algorithm network structure used in a coastal defense anti-smuggling evidence collection system based on a lightning ball, provided in an embodiment of the present invention;

[0039] Figure 6 The image shown is a suspected smuggling vessel captured by a marine anti-smuggling evidence collection system based on a lightning ball, as provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Figure 1 This invention provides a system architecture diagram for a maritime defense anti-smuggling evidence collection system based on a radar dome camera. The radar dome camera linkage system includes: a high-precision radar 1, a high-definition dome camera 2, an edge device 3, a network switch 4, and a client platform 5. Figure 1 The intelligent analysis controller is the edge device 3. The high-precision radar 1 and the high-definition PTZ camera 2 are mounted coaxially. One high-definition PTZ camera 2 and one high-precision radar 1 constitute the front end of radar 1.

[0042] The high-precision radar 1 has high positioning accuracy and measurement range. For example, it has a detection range of 1km, a detection accuracy of 1m, a maximum detection speed of 120km / h, and a coverage angle of 140 degrees, to achieve the capability of covering river estuaries and speed coverage.

[0043] The high-definition PTZ camera 2 has a range comparable to that of the high-precision radar 1, for example, 48x zoom, and comes with high-power laser illumination to ensure nighttime capture performance and guarantee reliable 24-hour operation.

[0044] The edge device 3 is deployed at the front end and independently realizes functions such as lightning ball linkage, multi-target tracking, image capture and recognition, and alarm information encapsulation and uploading, ensuring that the front end can still work normally even when the back end is offline.

[0045] The client platform 5 mainly consists of a software system deployed on the server, which interfaces with the edge machine 3 and is mainly used for functions such as real-time alarm information display, historical alarm event query, alarm zone setting, and device management.

[0046] As a further improvement of the present invention, the system architecture satisfies that one edge machine 3 can support multiple lightning ball front-ends, such as 5, and one client platform 5 can support multiple edge machines 3, such as 20. When the read and write capabilities of the client platform 5 are limited, an application server can be added, and load balancing can be achieved through distributed deployment of the client platform 5 to improve the overall system performance.

[0047] As a further improvement of the present invention, the system can also be equipped with a DTU device to enable real-time alarm information to be sent to a preset mobile phone in the form of SMS, MMS or voice broadcast, so as to realize the function of immediate handling.

[0048] Figure 2 This is a basic schematic diagram of the linkage principle of a marine anti-smuggling evidence collection system based on a radar dome camera, provided in an embodiment of the present invention. The high-precision radar 1 and the high-definition dome camera 2 are coaxially mounted vertically. Through equipment calibration, the coordinate system of radar 1 and the coordinate system of dome camera 2 are unified, realizing the function of radar 1 guiding dome camera 2 to rotate, track, and capture images.

[0049] Figure 2 In the diagram, PQRS represents a flat water surface, O is the equipment installation point, C is the position of PTZ camera 2, h is the mounting height of PTZ camera 2, A is a target within the surveillance area, the line containing OB is the X-axis of the horizontal coordinate system of radar 1, the direction of OB is the horizontal zero azimuth direction of PTZ camera 2, B is the projection of the target onto the X-axis of the horizontal coordinate system of radar 1, x is the radial distance of the target detected by radar 1, y is the tangential distance of the target detected by radar 1, and p is the horizontal angle that PTZ camera 2 should rotate. The line passing through point C and parallel to the direction of OB is the vertical zero azimuth direction of PTZ camera 2; therefore, t is the complementary angle of the vertical rotation angle that PTZ camera 2 should rotate.

[0050] but:

[0051] It should be noted that the rise and fall of the tide will affect the height h of the PTZ camera 2, that is, the height of the PTZ camera 2 from the horizontal plane where the target is located.

[0052] Figure 3 This invention provides a diagram of a master-slave search-type target tracking algorithm used in a radar-based anti-smuggling evidence collection system for coastal defense. The master-slave search-type target tracking method uses radar 1 to guide PTZ camera 2 for primary tracking, i.e., using a reference point calculated based on the basic principle of radar-ball linkage as the primary tracking point; this can also be called coarse tracking. If the coarse tracking detects a target, the target image position drives a PTZ video tracking algorithm as auxiliary tracking; this can also be called fine tracking. If the coarse tracking fails to detect a target, PTZ camera 2 zooms out to search for the target in the vicinity of the reference point. Figure 3 Using point A as the reference point, after the target is detected, video tracking is switched to ensure that the target remains centered in the frame, overcoming the inaccurate reference point caused by tidal fluctuations and insufficient tangential accuracy of radar 1. If single-target tracking reaches a threshold, tracking is switched to other targets. If video tracking fails, radar 1-guided tracking is initiated until the target disappears, at which point tracking exits, forming a complete closed loop.

[0053] The target image position-driven PTZ video tracking algorithm calculates the horizontal rotation angle, vertical rotation angle, and zoom value that the PTZ should adjust relative to the current position based on the distance of the target from the center of the image and the difference between the target size and the desired size. The formula for adjusting the current position using PTZ is as follows:

[0054] d p =-(ab·Zoom)·d x

[0055] d t =-(cd·Zoom)·d y

[0056] d zoom =-e·d w

[0057] Where, d p d is the horizontal angle that PTZ camera 2 should adjust based on the current horizontal angle. t d is the vertical angle that the PTZ camera 2 should adjust based on the current angle. zoom d is the zoom value that PTZ camera 2 should adjust based on the current zoom level. x d represents the horizontal pixel value of the target center in the image, which is offset from the center of the image. y d represents the vertical pixel value of the target center in the image, which is offset from the center of the image. w The difference between the width of the target in the image and the expected width of the target category is denoted by a, b, c, d, and e, which are preset constants. Zoom is the current zoom value.

[0058] Figure 4 This invention provides a multi-target tracking strategy diagram for a coastal defense anti-smuggling evidence collection system based on a lightning ball. The multi-target tracking strategy is implemented using a set multi-target tracking priority. Low-priority targets can be interrupted by high-priority targets, and when a high-priority target reaches its tracking threshold, the system automatically switches to tracking the next lower priority target. To meet the needs of anti-smuggling operations, this invention prioritizes the fastest high-speed targets as the first priority, non-high-speed targets that have not reached the set tracking threshold as the second priority, non-high-speed targets that have not been tracked as the third priority, and non-high-speed targets that have not been tracked before as the fourth priority, sorted by their last tracking time. This ensures that potentially smuggled high-speed targets are captured as soon as they appear.

[0059] The aforementioned maritime anti-smuggling evidence collection system improves target tracking smoothness for moving targets within the warning area detected by radar 1 by setting the following core parameters: a threshold for the number of batches of a single target detected by radar 1, and a threshold for the offset distance from the starting point of the batch after radar 1 detects the target. Targets exceeding these thresholds are considered valid moving targets, thus overcoming the illusion caused by sea clutter and other artifacts. The invention also sets a defined speed value for high-speed targets and a threshold for the number of consecutive batches of high-speed targets to be locked. Targets exceeding these thresholds are considered valid high-speed moving targets, facilitating the locking of high-speed targets.

[0060] Figure 5 This invention provides an improved YOLOv7 algorithm network structure diagram for a radar-based coastal defense anti-smuggling evidence collection system. The original YOLOv7 network structure is a conventional technology. This invention is based on the original structure, referencing the YOLOv7 network structure described in the article "Target Detection Algorithm - YOLOv7 - Detailed Explanation" published at https: / / blog.csdn.net / u012863603 / article / details / 126118799. To address the scaling characteristics of video footage in coastal defense applications and radar-based linkage, targets with excessively high proportions are avoided in captured images. The high downsampling rate detection heads responsible for large target detection in the original small, medium, and large output scale network structure are removed. Simultaneously, an additional convolutional layer branch is added before the original feature output to extract high-dimensional features of potential targets. IDs are assigned to different targets in the same frame based on the feature values, facilitating association with radar batch numbers and other information, making the video independent tracking process more robust in handling anomalies such as target occlusion. It is worth emphasizing that the newly added high-dimensional features will be concatenated and fused with the original head output features before being output together. This improvement not only simplifies the network structure and avoids the high complexity of computation caused by adding high-dimensional features, but also speeds up the network's operating efficiency.

[0061] For example, suppose the feature dimension of the original head region output is...

[0062] F = W * H / (S * S) * N * (K + 4 + 1)

[0063] Where W and H are the width and height of the input image, respectively, S is the downsampling rate of the detection head, N is the number of anchors, K is the total number of categories of the target to be detected, 4 represents the four-dimensional coordinates of the target in the plane, and 1 represents the confidence level of the foreground object.

[0064] Let M be the number of convolutional kernels in the newly added convolutional layer branches. After this convolution, the original main features will obtain an embedding vector V of dimensionality W*H / (S*S)*N*M, which is then concatenated and fused with the original head region output features F before being output together. This achieves stronger representational capabilities and reduces the computational resource requirements without rebuilding the feature extraction network.

[0065] Figure 6 This invention provides an embodiment of a maritime anti-smuggling evidence collection system based on lightning balls, capturing images of suspected smuggling vessels. It can be observed that the captured target images are all centered in the frame and include speed and location information, facilitating analysis by the coast guard.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A coastal defense anti-smuggling evidence collection system based on lightning balls, characterized in that, The system includes a radar, a PTZ camera, an edge device, a client platform, and a network switch. The radar and PTZ camera are coaxially mounted vertically to form a radar-PTZ front-end. The radar front-end is located above the area to be tracked. One edge device connects to multiple radar front-ends, and the multiple edge devices communicate with the client platform through the network switch. The edge device controls the radar front-end to track targets using a master-slave search-based target tracking strategy or a multi-target tracking strategy. The tracked images are input into an improved YOLOv7 network structure, and the target category and target location are output. The master-slave search target tracking uses radar-guided PTZ tracking as the main tracking method. After the target is detected, a video tracking algorithm driven by the target image position is used as an auxiliary tracking method. The multi-target tracking strategy prioritizes the fastest high-speed target for target tracking. The improved YOLOv7 network structure includes: The high downsampling rate detector heads responsible for large object detection in the original YOLOv7 network's three output scales (small, medium, and large) are removed. Simultaneously, an additional convolutional layer branch is added before the outputs of the small and medium object detector heads in the original YOLOv7 network. The output features of the small and medium object detector heads in the original YOLOv7 network are concatenated and fused with the features output by their corresponding convolutional layer branches before being output together. The small and medium object detector heads in the original YOLOv7 network are both part of the head region. The feature dimension of the head region output is F=W*H / (S*S)*N*(K+4+1), where W and H are the width and height of the input image, S is the downsampling rate of the detection head, N is the number of anchor points, K is the total number of categories of the target to be detected, 4 represents the four-dimensional coordinates of the target in the plane, and 1 represents the confidence of the foreground object. Let the number of convolution kernels in the newly added convolutional layer branch be M. Then, the main feature of the original YOLOv7 network is convolved into an embedding vector of W*H / (S*S)*N*M dimensions after this convolution. This embedding vector is then concatenated and fused with the head region output feature F and output together.

2. The coastal defense anti-smuggling evidence collection system based on lightning ball as described in claim 1, characterized in that, The radar dome camera's front end is calibrated to unify the radar coordinate system with the PTZ camera coordinate system, and radar-dot camera linkage is achieved through the following formula. in, The radial distance of the target detected by the radar. The tangential range of the target detected by radar. Setting the height for the PTZ camera, The horizontal angle at which the PTZ camera should rotate. This is the complementary angle of the vertical rotation angle of the PTZ camera.

3. The coastal defense anti-smuggling evidence collection system based on lightning ball as described in claim 1, characterized in that, The master-slave search-based target tracking strategy includes: The system uses radar-guided tracking, also known as radar-ball linkage calculation, as the primary tracking point. Once the primary tracking detects a target, it switches to a video tracking algorithm driven by the target image position in the PTZ for auxiliary tracking. If the auxiliary tracking fails to detect a target, the PTZ camera zooms out to search for the target in the vicinity of the primary tracking point. Once a target is found, it switches to video tracking. If single-target tracking reaches a threshold, it switches to tracking other targets. If video tracking fails, it enters radar-guided tracking until the target disappears, at which point it exits tracking, forming a complete closed loop.

4. A coastal defense anti-smuggling evidence collection system based on lightning balls according to claim 3, characterized in that, The target image position-driven PTZ video tracking algorithm includes: Based on the distance of the target from the center of the image and the difference between the target size and the desired size, calculate the horizontal rotation angle, vertical rotation angle, and zoom value that the PTZ should adjust relative to the current position. The formula for adjusting the current position using PTZ is as follows: in, The horizontal angle that the PTZ camera should adjust based on the current horizontal angle. This is the vertical angle that the PTZ camera should adjust based on the current angle. This is the zoom value that the PTZ camera should adjust based on the current zoom level. This represents the horizontal pixel value of the target center in the image, which is offset from the center of the image. This represents the vertical pixel value of the target center in the image, offset from the center of the image. This represents the difference between the width of the target in the image and the expected width for that target category. , , , , All are preset constants. This is the current multiplication value.

5. A coastal defense anti-smuggling evidence collection system based on lightning balls according to claim 1, characterized in that, The multi-target tracking strategy includes: The edge machine sets multi-target tracking priorities and tracks multiple targets in priority order. Low-priority targets can be interrupted by high-priority targets. When the tracking of a high-priority target reaches a threshold, it automatically switches to the next priority target. Among them, the fastest high-speed target is the first priority, non-high-speed targets that have not reached the set tracking threshold are the second priority, non-high-speed targets that have not been tracked are the third priority, and non-high-speed targets that have been tracked are sorted according to the last tracking time as the fourth priority.

6. A coastal defense anti-smuggling evidence collection system based on lightning balls according to claim 5, characterized in that, The edge processor sets a threshold for the number of radar batches starting for a single target and a threshold for the offset distance from the starting point of the batch after the radar batch starts. Only targets exceeding these thresholds are considered valid targets. The edge processor also sets a defined value for the speed of high-speed targets and a threshold for the number of consecutive batches of high-speed targets to be locked. Only targets exceeding these thresholds are considered valid high-speed targets, in order to facilitate locking high-speed targets.

7. A coastal defense anti-smuggling evidence collection system based on lightning balls according to claim 1, characterized in that, The radar has a detection range of 1km, a detection accuracy of 1m, a maximum detection speed of 120km / h, and a coverage angle of 140 degrees. The PTZ camera has a range comparable to the radar, a magnification of 48 times, and comes with high-power laser illumination.

8. A coastal defense anti-smuggling evidence collection system based on lightning balls according to claim 1, characterized in that, The client platform consists of a software system deployed on a server, which interfaces with the edge device for real-time alarm information display, historical alarm event query, alert zone setting, and device management.

9. A coastal defense anti-smuggling evidence collection system based on lightning ball as described in claim 1, characterized in that, The system also includes a DTU device, which is used to send real-time alarm information to a preset mobile phone in the form of SMS, MMS or voice broadcast.

Images