Target monitoring, detecting and tracking intelligent system and method

Abstract

The invention discloses a target monitoring, detecting and tracking intelligent system and method. The system comprises a real-time control module, a monitoring module, a detecting module, a multi-target detecting module, a tracking module and an exit module. According to the invention, by introducing an adaptive algorithm, the system can automatically adjust a threshold according to real-time data, in addition, the real-time control module can accurately control each sub-module to take action, and can clearly explain the practical significance of each decision, reduce the phenomena of false alarm and missing alarm, and more accurately detect and identify a target; the system has high intelligence, interpretability and stability, and has a good engineering application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of target detection and recognition, and particularly relates to an intelligent system and method for target monitoring, detection, and tracking. Background Art

[0002] In the field of military security, the importance of intelligent systems for surveillance, detection, and tracking is self-evident; these systems are widely used in multiple aspects such as border monitoring, critical infrastructure protection, and battlefield situation awareness; with the increasing requirements for the military security field and the increasingly complex battlefield environment, the performance requirements for such systems are also continuously improving, not only requiring higher accuracy, but also stronger adaptability and intelligence level, especially for the detection and recognition of small targets at long distances in complex battlefield environments.

[0003] Traditional algorithms rely on setting fixed thresholds to complete the surveillance, detection, and tracking of suspicious targets in a certain area. The following problems exist in this algorithm:

[0004] (1) Static threshold problem: Due to the lack of a dynamic adjustment mechanism, traditional fixed-threshold detection algorithms are unable to cope when facing complex and changeable environments; fixed thresholds are difficult to adapt to changing environmental conditions and task requirements, which may lead to excessive or insufficient targets being marked as candidates, thus affecting the quality and efficiency of decision-making.

[0005] (2) Dependence on manual intervention: The fixed-threshold method usually requires operators to be in a working state all the time, constantly intervening to adjust the threshold settings to ensure optimal performance; this method not only consumes a large amount of human resources, but also is prone to errors due to human factors, reducing the reliability and response speed of the system; especially in a military environment, long-term manual supervision not only increases the operation cost, but also limits the automation level of the system, which does not conform to the development trend of unmanned and intelligent.

[0006] (3) Insufficient adaptability: The modern battlefield environment changes rapidly, and factors such as weather, light, and background noise will affect the data quality of sensors; fixed thresholds cannot flexibly cope with these changes, limiting the application scope of the system; for example, under different meteorological conditions: sunny, rainy, foggy, etc., the same set of fixed thresholds may lead to completely different detection results, thus affecting the overall effectiveness of the system.

[0007] (4) Low intelligence level: For military applications, a high intelligence level means that it can be better integrated into the command and control system to achieve more efficient combat coordination; the existing fixed-threshold systems obviously cannot meet this trend, especially in the case of pursuing unmanned combat platforms. Summary of the Invention

[0008] The object of the present invention is to overcome the deficiencies of the prior art and provide an intelligent system and method for target monitoring, detection, and tracking, which can operate efficiently and reliably in a complex and changeable military environment, while ensuring that the operator has a clear understanding of the decision-making logic of the system, significantly improving the intelligence level of the monitoring, detection, and tracking systems in the field of military security.

[0009] The technical solution adopted by the present invention is as follows:

[0010] An intelligent system for target monitoring, detection, and tracking, comprising a real-time control module, a monitoring module, a detection module, a multi-target detection module, a tracking module, and an exit module;

[0011] The monitoring module performs real-time monitoring of suspicious areas, and when a suspicious target is found, it sends the suspicious target information to the real-time control module. The real-time control module judges the current state and selects a suitable module to execute the current task; if it is a false target, the real-time control module continues to select the monitoring module to perform the monitoring task; if the target is a real target, the real-time control module selects the detection module to perform the detection task;

[0012] The detection module performs the detection task on the suspicious target. If multiple targets are found, it manipulates the real-time control module to select the multi-target module to execute the task; if an important valuable target is found, it manipulates the real-time control module to select the tracking module to execute the task;

[0013] The multi-target detection module performs the task scenario of detecting multi-targets and swarm targets. By controlling the target growth button of the multi-target detection module, the number of detected targets increases exponentially;

[0014] The tracking module performs the real-time tracking task of the target;

[0015] The exit module performs the exit or restart task.

[0016] A method for target monitoring, detection, and tracking, the specific steps are as follows:

[0017] S1: Read the frame images to be processed in sequence according to the image sequence;

[0018] S2: The real-time control module defaults to selecting the monitoring module to execute the current task; at the same time, the real-time control module can select the monitoring module, detection module, multi-target detection module, tracking module, or exit module of the system at any time to execute the current task;

[0019] S3: If the monitoring module is selected to execute the current task in S2, set M to 0, where M represents the maximum detection quantity of the current frame image. Calculate the detection threshold T of the current frame image according to the M value, and enter step 4;

[0020] If the detection module is selected in S2 to execute the current task, the system generates an M value setting module. Through the M value setting module, an adjustable M value at any time, as well as two values M1 = 2 and M2 = 10, are input. The range of M is M1 < M < M2. The detection threshold T of the current frame image is calculated according to the M value, and then step 5 is entered;

[0021] If the multi-object detection module is selected in S2 to execute the current task, the system generates an M value setting module. Through the M value setting module, an adjustable M value at any time is input, and the target proliferation coefficient is set to 3. The detection threshold T of the current frame image is calculated according to the M value, and then step 6 is entered;

[0022] If the tracking module is selected in S2 to execute the current task, then step 7 is entered;

[0023] If the exit module is selected in S2, step 1 is entered;

[0024] S4: After entering the monitoring module to execute the task, according to the detection threshold T value and the input frame image, the detection result and the actual number of detected targets NUM are obtained, and the T value is updated every 5 frames; if NUM is greater than 0, a task switching prompt is generated and S2 is entered. If there is no operation within 3 seconds, S4 continues to be executed;

[0025] S5: After entering the detection module to execute the task, according to the detection threshold T value and the input frame image, the detection result and the actual number of detected targets NUM are obtained; if M1 < NUM < M2, the next frame of input image continues to be detected with the detection threshold T value; otherwise, the detection threshold T of the next frame image is recalculated, and according to the updated detection threshold T value and the current frame input frame image, a new NUM value is obtained;

[0026] S6: After entering the multi-object detection module to execute the task, according to the detection threshold T value and the input frame image, the detection result and the actual number of detected targets NUM are obtained; a target addition button is generated. Each time it is clicked, M is multiplied by the target proliferation coefficient 3, that is, the M value is updated to M = M * 3. The updated M value is input, and the detection threshold T of the next frame image is calculated, and the detection result and the number of detected targets NUM are obtained;

[0027] S7: After entering the tracking module to execute the task, the M value is set to 1. According to the M value and the current input frame image, the detection threshold T is calculated, and according to the detection threshold T value, the detection result and the number of detected targets NUM are obtained.

[0028] Specifically, when the monitoring module, the detection module, and the multi-object detection module in S3 and the tracking module in S7 execute tasks, the calculation method of the detection threshold T is specifically as follows:

[0029] a: A feature extraction method based on moving targets, which constructs a saliency map on the original image to highlight potential target areas;

[0030] b: Set a reasonable threshold to binarize the image and distinguish targets from the background

[0031] c: Perform morphological operations to determine the target areas and the number of targets;

[0032] d: Take the maximum D of the saliency map of the target area and set the initial value of the threshold Th to D;

[0033] e: Gradually decrease the value of the threshold Th as the detection threshold T until the number of detected targets is greater than the maximum detection number M of the current frame image. At this time, the threshold Th is denoted as D min ; Let the detection threshold T be equal to D min + 1. At this time, the actual number of targets NUM that can be obtained using this detection threshold T is less than or equal to M.

[0034] Due to the above - mentioned technical solution, the present invention has the following advantages:

[0035] (1) High flexibility: By introducing an advanced adaptive algorithm, the system can automatically adjust the threshold according to real - time data, thereby more accurately identifying targets and reducing false alarms and missed detections; this ability is particularly suitable for complex environments in military scenarios, such as continuous monitoring during day - night alternation and different weather conditions.

[0036] (2) Improved reliability: By precisely controlling and clearly explaining each decision, the possibility of human error is reduced, enhancing the reliability and stability of the system.

[0037] (3) Optimized resource allocation: Flexibly adjust the detection intensity according to different task requirements, avoid unnecessary resource waste, and achieve the best allocation of resources.

[0038] (4) Enhanced tactical coordination: A higher degree of automation and a better human - machine interaction experience promote close cooperation between different combat units, improving the overall combat effectiveness. Brief Description of the Drawings

[0039] Figure 1 is a flowchart of the present invention.

[0040] Figure 2 is an infrared image small - aircraft target detection and tracking dataset under ground / air background adopted in the embodiment of the present invention.

[0041] Figure 3 is a monitoring result map of the first - frame image by the monitoring module of the present invention.

[0042] Figure 4It is the monitoring result diagram of the 13th frame image by the monitoring module of the present invention.

[0043] Figure 5 It is the detection result diagram of the 13th frame image by the detection module of the present invention.

[0044] Figure 6 It is the detection result diagram of the 15th frame image by the detection module of the present invention.

[0045] Figure 7 It is the monitoring result diagram of the 16th frame image by the detection module of the present invention.

[0046] Figure 8 It is the monitoring result diagram of the 45th frame image by the monitoring module of the present invention.

[0047] Figure 9 It is the tracking result diagram of the 100th frame image by the tracking module of the present invention. Detailed implementation manners

[0048] The present invention will be further explained and illustrated below in conjunction with the accompanying drawings and embodiments. The protection scope of the present invention cannot be limited thereby. The purpose of disclosing the present invention is to protect all technical improvements within the scope of the present invention.

[0049] This embodiment is executed relying on a computer equipped with an AMD Ryzen 5 5600K CPU, 16GB RAM, RTX 3060 GPU hardware functions, and MATLAB 2024a software; the dataset is selected as the "Weak Small Aircraft Target Detection and Tracking Dataset of Infrared Images under Ground / Air Background", which is provided by the ATR Key Laboratory of the School of Electronic Science, National University of Defense Technology, providing 21 types of frame sequence infrared image data and having relatively complete information markings; this implementation case selects the data3 dataset, as Figure 2 shown. The scene of this data is the background of the ground-air intersection. There is no target at the beginning, then a single target flies into the field of view, then flies out of the field of view, and finally enters the field of view again; as shown in the corresponding image sequence display, there are a total of 100 frame images, among which there is no target from the 1st frame to the 12th frame, there is a target from the 13th frame to the 29th frame, there is no target from the 30th frame to the 44th frame, and there is a target from the 45th frame to the 100th frame.

[0050] An intelligent system for target monitoring, detection, and tracking includes a real-time control module, a monitoring module, a detection module, a multi-target detection module, a tracking module, and an exit module.

[0051] The monitoring module performs real-time monitoring of suspicious areas. When a suspicious target is detected, it sends the information of the suspicious target to the real-time control module. The real-time control module judges the current state and selects an appropriate module to execute the current task. If it is a false target, the real-time control module continues to select the monitoring module to perform the monitoring task. If the target is a real target, the real-time control module selects the detection module to perform the detection task.

[0052] The detection module performs the detection task on the suspicious target. If multiple targets are detected, it manipulates the real-time control module to select the multi-target module to execute the task. If an important valuable target is detected, it manipulates the real-time control module to select the tracking module to execute the task.

[0053] The multi-target detection module executes the task scenario for detecting multi-targets and swarm targets. By controlling the target growth button of the multi-target detection module, the number of detected targets increases exponentially.

[0054] The tracking module performs the real-time tracking task on the target.

[0055] The exit module executes the exit or restart task.

[0056] A method for target monitoring, detection, and tracking, the specific steps are as follows:

[0057] Execute S1, and sequentially read 100 frames of images to be processed according to the image sequence.

[0058] Execute S2: The real-time control module default selects the monitoring module to execute the current task. At the same time, the implementation control module can select the monitoring module, detection module, multi-target detection module, tracking module, or exit module of the system to execute the current task at any time.

[0059] Execute S3. The real-time control module first selects to execute the monitoring task, sets M to 0, and calculates the detection threshold T of the first frame image to be 39.44.

[0060] Execute S4, enter the monitoring module, and obtain the detection result of the first frame according to the detection threshold T. As Figure 3 shown, NUM is 0, and the T value is updated every 5 frames.

[0061] No target is detected in the first frame to the fourth frame, and NUM is 0. The threshold is updated in the fifth frame, and T is obtained as 45.47. Detection is performed using the updated threshold. No target is detected in the fifth frame to the ninth frame, and NUM is 0. The threshold is updated in the tenth frame, and T is obtained as 42.56. Detection is performed using the updated threshold until the 13th frame, when a target is detected and NUM is 1. As Figure 4As shown; since the NUM value is greater than 0, a task switching prompt is generated, enter S2, select to enter the detection module, set M1 to 2, M2 to 10, set M to 5, update the detection threshold T to 22.05, and continue to execute S5.

[0062] Execute S5. After entering the detection module to execute the task, according to the detection threshold T value and the current 13th frame image, the detection result and the detected target number NUM are obtained as 4, as Figure 5 shown; continue to execute the operation until the 15th frame, and the new NUM is 22. The detection result is as Figure 6 shown. Since there is a lot of interference in the ground background, there are many false targets.

[0063] Since the NUM value is greater than the M2 value, recalculate the detection threshold T of the next frame image as 36.46. According to the updated T value and the current frame input frame image, the new NUM value is 3. The detection result is as Figure 7 shown.

[0064] Continue to execute the operation until the 30th frame. When the target flies out of the field of view, switch to the monitoring task and set M to 0.

[0065] Execute S4. Enter the monitoring module to execute the monitoring task, and perform the frame detection tasks in the same steps as above until the 45th frame. The new NUM value is 1. The detection result is as Figure 8 shown; since the NUM value is greater than 0, a task switching prompt is generated, enter S2, switch to the tracking task, set M to 1, and calculate and execute S7.

[0066] Execute S7. After entering the tracking module, according to the M value and the input frame image, the detection result is obtained until the end of the sequence image. The NUM value obtained is always 1. The 100th frame image is as Figure 9 shown; from the 45th frame image to the end of the 100th frame image, the tracking module realizes the stable tracking of the target.

[0067] Since the multi-target detection is not included in this dataset, in this technical solution, if the multi-target detection module is selected to execute the current task in S2, the system generates an M value setting module, sets M = 5, and sets the target proliferation coefficient to 3. Calculate the detection threshold T of the current frame image according to the M value, execute S6, and according to the detection threshold T value and the input frame image, obtain the detection result and the actual detected target number NUM; generate a target increase button. Each time it is clicked, update the M value to M = M * 3, input the updated M value, calculate the detection threshold T of the next frame image, and obtain the detection result and the detected target number NUM.

[0068] In the above steps, the calculation method of the detection threshold T is specifically as follows:

[0069] a: A feature extraction method based on moving objects, which constructs a saliency map on the original image to highlight potential target regions.

[0070] b: Set a reasonable threshold to binarize the image and distinguish the target from the background.

[0071] c: Perform morphological operations to determine the target regions and the number of targets.

[0072] d: Take the maximum D of the saliency map of the target region and set the initial value of the threshold Th to D.

[0073] e: Gradually decrease the value of the threshold Th as the detection threshold T until the number of detected targets is greater than the maximum detection number M of the current frame image. At this time, the threshold Th is denoted as D min ; Let the detection threshold T be equal to D min +1. At this time, the actual number of targets NUM obtained using this detection threshold T is less than or equal to M.

[0074] The parts not detailed in the present invention are prior art.

[0075] The embodiments selected herein for disclosing the object of the present invention are considered to be suitable at present. However, it should be understood that the present invention is intended to cover all variations and improvements of all embodiments falling within the scope of this concept and invention.

Claims

1. An intelligent target monitoring, detection and tracking system, characterized in that: It includes a real-time control module, a monitoring module, a detection module, a multi-target detection module, a tracking module, and an exit module; The monitoring module performs real-time monitoring of suspicious areas. When a suspicious target is found, it sends the suspicious target information to the real-time control module. The real-time control module judges the current state and selects an appropriate module to execute the current task. If it is a false target, the real-time control module continues to select the monitoring module to perform the monitoring task. If the target is a real target, the real-time control module selects the detection module to perform the detection task; The detection module performs the detection task on the suspicious target. If multiple targets are found, it manipulates the real-time control module to select the multi-target module to execute the task; If an important valuable target is found, it manipulates the real-time control module to select the tracking module to execute the task; The multi-target detection module executes the task scenario for detecting multi-targets and swarm targets. By controlling the target growth button of the multi-target detection module, the number of detected targets increases exponentially; The tracking module performs the real-time tracking task on the target; The exit module executes the exit or restart task.

2. A target monitoring, detection and tracking method, characterized in that: The specific steps are as follows: S1: Read the frame images to be processed in sequence according to the image sequence; S2: The real-time control module defaults to selecting the monitoring module to execute the current task. At the same time, the real-time control module can随时 select the monitoring module, detection module, multi-target detection module, tracking module, or exit module of the system to execute the current task; S3: If the monitoring module is selected to execute the current task in S2, set M to 0. M represents the maximum detection number of the current frame image. Calculate the detection threshold T of the current frame image according to the M value, and enter step 4; If the detection module is selected to execute the current task in S2, the system generates an M value setting module. Input an adjustable M value at any time through the M value setting module, as well as two values M1 = 2 and M2 = 10. The value range of M is M1 < M < M2. Calculate the detection threshold T of the current frame image according to the M value, and enter step 5; If the multi-target detection module is selected to execute the current task in S2, the system generates an M value setting module. Input an adjustable M value at any time through the M value setting module, and set the target proliferation coefficient to 3. Calculate the detection threshold T of the current frame image according to the M value, and enter step 6; If the tracking module is selected to execute the current task in S2, then enter step 7; If the exit module is selected in S2, enter step 1; S4: After entering the monitoring module to execute the task, according to the detection threshold T value and the input frame image, obtain the detection result and the actual number of detected targets NUM. Update the T value every 5 frames. If NUM is greater than 0, generate a task switch prompt and enter S2. If there is no operation within 3 seconds, continue to execute S4; S5: After entering the detection module to execute the task, according to the detection threshold T value and the input frame image, obtain the detection result and the actual number of detected targets NUM; If M1 < NUM < M2, continue to detect the next frame of input image with the detection threshold T value; Otherwise, recalculate the detection threshold T of the next frame image. According to the updated detection threshold T value and the current frame input frame image, obtain a new NUM value; S6: After entering the multi-target detection module to perform the task, the detection result and the number of actually detected targets NUM are obtained according to the detection threshold T value and the input frame image; Generate a target increase button. Each time you click it, M is multiplied by the target multiplication coefficient 3, that is, the M value is updated to M=M*3. Enter the updated M value, calculate the detection threshold T for the next frame image, and obtain the detection result and the number of detected targets NUM. S7: After entering the tracking module to perform the task, the M value is set to 1. According to the M value and the current input frame image, the detection threshold T is calculated. According to the detection threshold T value, the detection result and the number of detected targets NUM are obtained.

3. The target monitoring, detection and tracking method according to claim 2, characterized in that: When the monitoring module, detection module and multi-target detection module in S3 and the tracking module in S7 perform tasks, the calculation method of the detection threshold T is specifically as follows: a: Based on the feature extraction method of moving targets, a saliency map that can highlight the potential target area is constructed on the original image; b: Set a reasonable threshold to binarize the image and distinguish the target from the background; c: Implement morphological operations to determine the target area and target quantity; d: Take the maximum D of the saliency map of the target area and set the initial value of the threshold Th to D; e: Gradually reduce the value of the threshold Th as the detection threshold T until the number of detected targets is greater than the maximum detection number M of the current frame image. The threshold Th at this time is recorded as D min ; Let the detection threshold T equal to D min +1, at this time, the actual target NUM that can be obtained using the detection threshold T is less than or equal to M.

Images