Unmanned aerial vehicle abnormity intelligent inspection, identification and analysis system and method thereof

Abstract

The invention provides an unmanned aerial vehicle abnormity intelligent inspection, identification and analysis method and system, and the method comprises the steps: a discoverer judges whether a ship has an abnormal condition or not according to the search of the ship and an abnormal condition identification strategy, calibrates the ship as a target ship if the ship has the abnormal condition, and tracks the target ship; in the process that a discoverer tracks a target ship, the discoverer obtains the position of the target ship and the state of a surrounding predator, selects the corresponding predator, adjusts the distance between the discoverer and the target ship and between the discoverer and the predator through a tracking strategy, adjusts the flight path of the discoverer, and transmits the target characteristics of the target ship to the selected predator. The target features comprise ship positions, anchor points and abnormal conditions; the predator flies to the position where the target ship is located after obtaining the target features, and after the target ship enters the predator collection range, the target ship is locked through a feature recognition strategy according to the anchor points on the target ship; and the predator monitors the target ship and judges whether an abnormal condition exists or not.

Description

Technical Field

[0001] The present invention relates to the field of drones, and in particular to a drone abnormality intelligent inspection, identification and analysis system and method. Background Art

[0002] Drones can be used to obtain important ground information, such as images, including still pictures and videos, from which timely and accurate battlefield information and precise positioning information can be obtained to capture strategic strike targets and complete tasks such as strike effect evaluation.

[0003] During aerial ground observation, moving targets on the water surface, such as ships and vessels, are of great value and are a key reconnaissance focus. Drones can be used to detect unusual emissions and illegal stays, often requiring continuous monitoring during flight. However, since moving targets are in motion along with the drone, they can often be difficult to capture due to movement beyond the drone's field of view.

[0004] Searching, tracking, and locating moving targets is a hot topic in machine vision research and is widely used in practice. The key lies in ensuring rapid and accurate large-scale search and positioning. Existing technical solutions fall into two main categories: one utilizes monocular or binocular vision to identify, track, and locate targets. While this approach can accurately identify and locate moving targets, it cannot guarantee large-scale search and precise positioning of multiple targets simultaneously. The other approach involves installing a sufficient number of cameras within the search area to ensure that the entire area is fully covered. While this approach addresses the aforementioned issues to a certain extent, it increases costs and introduces difficulties in subsequent image processing. In summary, current engineering technologies struggle to simultaneously achieve large-scale search and precise positioning of multiple moving targets. For example, Application No. CN201510684345.8, "Multi-Moving Target Search and Positioning Device and Method Based on Bird Vision Characteristics," discloses a multi-moving target search and position device and method based on bird vision. From a biomimetic perspective, this device leverages the visual characteristics of animals in predator-prey relationships and incorporates the biconcave structure of birds for a biomimetic design. The device consists of a horizontally mounted circular base, four identical 2-DOF monocular mechanisms evenly distributed along the circle, a fixed camera mounted in the center of the base, and the corresponding computer and circuitry. This device combines the wide field of view of prey animals with their well-developed binocular vision, enabling rapid search and precise location of multiple moving targets across a panoramic area. It is suitable for application and widespread adoption in visual surveillance in engineering applications. However, due to the vast expanse of the water surface, it is not possible to track moving targets above the surface. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a drone abnormality intelligent inspection, identification and analysis system and method thereof to overcome the above-mentioned defects in the existing technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: Intelligent inspection, identification and analysis methods for drone anomalies, including In the target search step, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel according to the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked; Predator docking step: When the discoverer tracks the target vessel, it obtains the target vessel's position and the status of the surrounding predators, selects the corresponding predator, and adjusts the distance between the discoverer and the target vessel and predator through the tracking strategy, adjusts the discoverer's flight path, and transmits the target vessel's target characteristics to the selected predator. The target characteristics include the vessel's position, anchor point, and abnormal conditions. Target vessel identification step: After the predator obtains the target features, it flies to the location of the target vessel. When the target vessel enters the predator's acquisition range, it locks the target vessel through feature recognition strategy based on the anchor point on the target vessel; In the evidence collection step, the predator monitors the target vessel to determine whether there are any abnormalities. If so, it obtains evidence frames. If not, it obtains key frames with target features collected by the discoverer and obtains detailed images of the target vessel collected by the predator. These images are cross-compared to form a complete chain of evidence.

[0007] Preferably, the tracking strategy includes a method for obtaining the moving speed, moving trajectory and environmental visibility of the target vessel, and through the positional relationship between the target vessel, the predator and the discoverer, obtaining the corresponding docking weight and positioning weight according to the data index table, calculating the tracking value, and presetting a tracking threshold. When the tracking value is less than the tracking threshold, an adjustment strategy is generated to adjust the docking weight and positioning weight so that the tracking value is greater than the tracking threshold.

[0008] Preferably, the tracking strategy further comprises a docking step, wherein the docking step comprises The data acquisition sub-step is used to dynamically adjust the communication frequency and the transmission frequency according to the distance between the discoverer and the predator, the environmental conditions, and the flight speed of both, so as to obtain the real-time communication quality between the discoverer and the predator; The docking distance calculation sub-step is used to obtain the docking weight and real-time communication quality, and obtain the docking distance through the distance adjustment algorithm, and dynamically adjust the flight speed and flight path between the discoverer and the predator according to the docking distance.

[0009] Preferably, the tracking strategy also includes a positioning step, which is used to obtain a positioning weight, obtain a positioning distance based on the positioning weight, and dynamically adjust the flight speed and flight route of the discoverer based on the positioning distance.

[0010] Preferably, the predator docking strategy is provided with an anchor point acquisition step, which is used to obtain images collected by several frames of the finder as target images, and obtain image information of the target ship and other ships in the target image, compare the target ship with the other ships, obtain several feature images of the target ship as anchor points, and obtain anchor point weights according to the frequency of occurrence of the anchor points on the other ships.

[0011] Preferably, the feature recognition strategy includes obtaining anchor points and obtaining images collected by the predator as images to be analyzed, respectively calculating the similarity between the ship images in the images to be analyzed and the anchor point images, and calculating the similarity based on the anchor point weights, and performing priority sorting based on the similarity, and selecting the ship with the highest similarity as the target ship.

[0012] Preferably, the predator docking step also includes a relay matching sub-step. When the predator has not entered the predation range during the tracking process of the current discoverer and the target vessel is about to leave the monitoring range of the current discoverer, the current discoverer searches for the remaining discoverers and passes the obtained target features to the next discoverer, and the next discoverer docks with the next predator. The current discoverer continues to track the target vessel until the next discoverer or predator tracks the target vessel.

[0013] Preferably, the relay matching sub-step also includes a predator value assessment strategy, which obtains the distance and movement trajectory between the current predator and the target predator respectively, and the cross-zone parameters are preset in the current predator, calculates the predator tracking cost respectively, compares the tracking cost of the current predator and the next predator, and selects the predator with the lower tracking cost to track the target vessel.

[0014] UAV abnormal intelligent inspection, identification and analysis system, including In the target search module, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel according to the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked; The predator docking module, when the discoverer tracks the target vessel, obtains the target vessel's location and the status of surrounding predators, selects the corresponding predator, and adjusts the distance between the discoverer and the target vessel and predator through the tracking strategy, adjusts the discoverer's flight path, and transmits the target vessel's target characteristics to the selected predator. The target characteristics include the vessel's location, anchor point, and abnormal conditions. In the evidence collection step, the predator monitors the target vessel to determine whether there are any abnormalities. If so, it obtains evidence frames. If not, it obtains key frames with target features collected by the discoverer and obtains detailed images of the target vessel collected by the predator. These images are cross-compared to form a complete chain of evidence.

[0015] The present invention provides the following beneficial effects: During the target search step, the discoverer uses an anomaly identification strategy to make a preliminary assessment of the vessel and flag any target vessel that may be anomaly. Subsequently, during the evidence collection step, the predator monitors and assesses the target vessel again. This multi-stage anomaly assessment mechanism significantly reduces the likelihood of misjudgment and improves the accuracy of anomaly detection. For example, in complex marine environments, interference factors may cause initial assessments to deviate, but these interferences can be effectively eliminated through the predator's reconfirmation. During the target vessel identification step, the predator uses a feature recognition strategy to locate the target vessel based on anchor points on the target vessel. These anchor points are carefully selected and identified during the target search step, ensuring high uniqueness and recognition. This ensures that the predator accurately locates the target vessel, avoiding misjudgment in complex scenarios. During the predator docking step, the discoverer tracks the target vessel, obtains its location and the status of surrounding predators, and selects an appropriate predator for docking. Simultaneously, the discoverer adjusts its flight path and transmits the target vessel's target features to the selected predator. This collaborative model allows the discoverer and predator to leverage their respective strengths, achieving efficient task allocation and execution. For example, the discoverer has broader search capabilities, while the predator has more precise target locking and evidence collection capabilities. During the evidence collection step, the predator not only monitors the target vessel in real time to obtain possible anomaly evidence frames, but also obtains key frames containing target features captured by the discoverer and cross-compares them with its own detailed images of the target vessel. This multi-source evidence collection and cross-comparison creates a complete chain of evidence, providing strong support for subsequent anomaly analysis and resolution. For example, when addressing vessel violations, a complete chain of evidence can accurately prove the time, location, and specific circumstances of the violation. During the predator docking step, the discoverer adjusts its flight path based on the tracking strategy, optimizing the distance between itself and the target vessel and the predator. This intelligent path adjustment mechanism improves inspection efficiency and reduces unnecessary flight time and energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall flow chart of the present invention; Figure 2 It is a step diagram of the present invention; Figure 3 is a schematic diagram of the tracking strategy of the present invention; Figure 4 It is a module connection diagram of the present invention; Figure 5 It is the target vessel identification diagram of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a central component. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a central component. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: Intelligent inspection, identification and analysis methods for drone anomalies, including In the target search step, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel based on the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked. The discoverer searches in the water inspection area based on the sparrow algorithm and is equipped with a camera, infrared sensor and radar to perform a full-scale scan of the vessels in the inspection area. The camera is used to capture images at a certain frame rate, the infrared sensor detects the heat characteristics and distance emitted by the vessel, and the radar detects the position and operating status of the vessel in real time. The convolutional neural network model is pre-trained to identify the appearance characteristics of the vessel and whether the vessel has abnormal emissions, oil leaks, etc., and compares the normal The image features of the ship are detected. If the deviation value between the detected ship features and the normal features exceeds a preset threshold, it is judged as an abnormal situation. The temperature data of the ship and the surrounding waters detected by the infrared sensor are compared with the normal temperature data. If there is a significant abnormality between the temperature of the waters around the ship and the temperature of the rest of the area, the ship has abnormal emissions. When an abnormal situation is detected in a ship, the detector calibrates the target ship and marks the outline of the ship in the image as the target ship. A unique identification number is assigned to it. Based on the current position and movement direction of the target ship, the detector predicts the future position, adjusts its own flight path, and continuously maintains the detection of the target ship. During the tracking process, the target ship's position information is continuously updated, and related data such as abnormal situations are stored. Predator docking steps: During the process of tracking the target ship, the discoverer obtains the target ship's position and the status of the surrounding predators, and selects the corresponding predator. Through the tracking strategy, the distance between the discoverer and the target ship and the predator is adjusted, the flight path of the discoverer is adjusted, and the target characteristics of the target ship are transmitted to the selected predator. The target characteristics include the ship's position, anchor point, and abnormal conditions; the discoverer uses the positioning system and continuous tracking of the target ship to try to obtain the target ship's latitude and longitude coordinates, speed, and heading information. The discoverer establishes a communication link with the surrounding predators to obtain the predator's status. The predator's status includes location, remaining power, device working status (whether the camera is normal, data storage capacity, etc.) and whether it is busy. According to the obtained predator status, the appropriate predator is selected by optimizing the U-shaped Anze algorithm. The selection reference factors include the distance between the predator and the target ship, and the predator with a closer distance is given priority. Shorten the response time; select a predator with sufficient power, fault-free equipment, and in idle state; the discoverer dynamically adjusts the distance between the three based on its relative position to the target vessel, the relative position to the selected indirect target, and the current environmental factors through tracking strategies to ensure that the discoverer can continue to track the target vessel and successfully dock with the predator; the discoverer transmits the target characteristics of the target vessel to the selected predator, and the target characteristics include the position of the vessel, which is used to guide the predator to quickly approach the target; anchor points, such as unique frost or color structures on the vessel, serve as key references for the predator to subsequently lock on to the target, as well as detailed descriptions of abnormal situations, such as the type of abnormality and the location of occurrence, to help the predator understand the target situation in advance and conduct targeted monitoring.

[0021] Target vessel identification steps: After acquiring the target features, the predator flies to the target vessel's location. When the target vessel enters the predator's acquisition range, it locks onto the target vessel using a feature recognition strategy based on the target vessel's anchor points. After receiving the target features, the predator uses a built-in navigation algorithm and a positioning pass to fly to the target vessel's location. During flight, it adjusts its flight path in real time based on the target vessel's location information updated by the discoverer to ensure accurate approach to the target. When the target vessel enters the predator's image acquisition range, the predator uses a high-definition camera to capture images of the target vessel. Using an image recognition algorithm, it identifies the anchor points provided by the discoverer. Using a feature point matching algorithm, it searches for areas in the collected images that coincide with the anchor points, thereby accurately locking onto the target vessel's position in the image. After identifying the anchor points, it further combines the target vessel's overall appearance features, navigation posture, and other information to compare and verify with the target features provided by the discoverer. If the matching degree of each feature reaches the set threshold, the vessel is confirmed to be the target vessel and locked onto the target.

[0022] During the evidence collection process, the predator monitors the target vessel to determine if any anomalies exist. If so, it captures evidence frames. If not, it captures key frames containing target features captured by the discoverer, along with detailed images of the target vessel captured by the predator. These images are then cross-referenced to form a complete chain of evidence. The predator uses its onboard specialized monitoring equipment, such as high-resolution cameras, multispectral sensors, and gas detectors, to comprehensively monitor the target vessel. Image analysis algorithms and sensor data are then used to determine if any anomalies exist on the target vessel. For example, multispectral sensors can detect the spectral signature of oil spills on the sea surface, while gas detectors can detect abnormal pollutant emissions. If an anomaly is detected on the target vessel, the predator immediately captures evidence frames. For image-based evidence, high-definition images are captured, clearly detailing the anomaly to ensure that the features are clearly discernible. Other types of evidence, such as gas detection data and pollutant sample data, are accurately recorded and stored. If no anomalies are detected, the predator captures key frames containing target features captured by the discoverer, along with detailed images of the target vessel captured by itself. The two are cross-compared to check for any anomalies missed due to differences in monitoring equipment accuracy or different monitoring angles. Comparative analysis of the vessel's appearance, structural integrity, and other features ensures that no anomalies have been overlooked. The evidence frames captured by the predator, the key frames of the discoverer, and related sensor data, monitoring time, and location information are integrated to form a complete chain of evidence. The evidence is numbered, timestamped, and annotated with relevant descriptions, creating an evidence database for subsequent query, analysis, and processing. This chain of evidence ensures the accuracy and traceability of monitoring anomalies on the target vessel, while also preventing the vessel from ceasing its abnormal behavior after the discoverer or predator approaches, making it impossible to promptly punish the vessel.

[0023] The tracking strategy involves obtaining the target vessel's speed, trajectory, and environmental visibility. Based on the positional relationships between the target vessel, predator, and discoverer, the system calculates the tracking value using a data index table to obtain the corresponding docking and positioning weights. A preset tracking threshold is set. If the tracking value falls below the threshold, an adjustment strategy is generated to adjust the docking and positioning weights to bring the tracking value above the threshold. The system also obtains the target vessel's speed and trajectory, as well as environmental visibility, which affects the accuracy of image analysis. Based on the acquired data, the system queries a data index table. This pre-established data index table contains the docking and positioning weights corresponding to different drone performance levels. Using these weights, the system calculates the tracking value. A preset tracking threshold is set as a metric to measure tracking effectiveness. If the calculated tracking value falls below the threshold, the current tracking strategy is ineffective and needs adjustment. A dynamic tracking strategy is used to ensure that the predator is always within the discoverer's information docking range while the discoverer is tracking the target vessel, avoiding disconnection between the discoverer and the predator due to the target vessel moving too fast. At the same time, the target vessel is accurately locked to avoid the discoverer being too close to the target vessel and being unable to effectively guide the predator due to limited field of view, and to avoid the target positioning accuracy being reduced due to the distance from the orchard, which may actually lead to the loss of the target vessel.

[0024] Formula 1 ; in, is the space-time-visibility coupling value, For time Normalized ambient visibility at the moment, For the maximum visibility of the discoverer's theory, is the instantaneous speed of the target ship, is the reference speed, is the signal attenuation rate, is the time oscillation period; Specific calculation formula =0.9, =20 sections, =0.12 , =8s; , =18 sections; ; ; ; Formula 2 ; is the dual distance cooperative entropy, is the complementary error function, , is the real-time distance between the discoverer and the target vessel, is the standard deviation of the observation error, is the hyperbolic secant function, is the real-time distance between the discoverer and the predator, is the distance scale factor, is the smoothing parameter, is the docking weight, Specific calculation formula =1.5km, =0.6, =1.2km; =4km, =3km, =0.8; ; ; ; ; right After correction, we get 0.68; Formula 3 ; in, is the trajectory locking factor, is the positioning weight, is the prediction error vector of the k-th trajectory point, is the error tolerance radius, is the azimuth The probability density function of is just a function, At the effective observation angle If yes, it is 1, if no, it is 0. The maximum observation angle is 30°. If the angle exceeds this, the data will be invalid. Specific calculation formula =0.9, ; =0.4km, ; ; ; Corrected to 0.0966; , revised to 1.18; Formula 4 ; For comprehensive tracking cost, is the theoretical maximum value, 、 is the sensitivity index, To prevent zero factors, is the emergency correction factor, Specific calculation formula =1.2, =1.2, =0.8, =0.01, =0.05; ; ; ; ; ; like It needs to be corrected.

[0025] The tracking strategy also includes a docking step, which includes a data acquisition substep. This step dynamically adjusts the communication and transmission frequencies based on the distance between the finder and predator, environmental conditions, and the flight speeds of both, to monitor the real-time communication quality between the two. The distance between the finder and predator is a significant factor affecting communication and signal transmission. Generally speaking, the greater the distance, the greater the signal attenuation, and the lower the communication quality. Environmental conditions, such as weather (rain, fog, strong winds, etc.) and electromagnetic interference, can also significantly impact communication and signal transmission. The flight speed of both parties also affects communication. High speeds can cause problems such as Doppler shift, further impacting communication quality. The system dynamically adjusts the communication and transmission frequencies. For example, when the distance is long or the environmental interference is high, the transmission frequency is appropriately increased to enhance signal strength, while the communication frequency is adjusted to avoid interference bands, thereby improving communication stability and reliability. By continuously adjusting the communication and transmission frequencies, the system can monitor the communication quality between the finder and predator in real time.

[0026] The docking distance calculation substep is used to obtain the docking weight and real-time communication quality, and then use the distance adjustment algorithm to obtain the docking distance. Based on the docking distance, the system uses the distance adjustment algorithm to calculate the docking distance. Based on the calculated docking distance, the system dynamically adjusts the flight speed and flight path between the discoverer and predator. If the docking distance needs to be shortened, the discoverer and predator may reduce their flight speed and adjust their flight path to gradually approach each other. If the docking distance can be increased, they may appropriately increase their flight speed and optimize their flight path to improve docking efficiency.

[0027] The distance adjustment algorithm is: Formula 1 ; is the communication quality attenuation factor, Q is the current communication quality, The communication quality threshold, below which the distance adjustment is initiated, is the docking weight, is the weight saturation parameter, erfc is the complementary error function, is the hyperbolic secant function; Specific calculation formula Q= -85dB, =-80dB; ; ; ; ; 1-0.486=0.514; ; Formula 2 ; K(D) is the dynamic distance adjustment parameter, is the theoretical optimal distance, is the distance sensitivity index, is the current actual distance, is the distance parameter coefficient, is the reference distance, is the Hessian matrix of communication quality, Frobenius norm, is the mutation suppression coefficient; Specific calculation formula =600m, =3, =1500m, =0.8, =1000m, =0.5; ; ; ; ; ; ; Formula 3 ; is the docking distance, is the minimum allowed distance, is the maximum allowed distance, is the threshold response index, To adjust the smoothing parameter, ; Specific calculation formula =400m, =800m, =0.2, =1.2; ; ; ; .

[0028] The tracking strategy also includes a positioning step, which is used to obtain a positioning weight, determine a positioning distance based on the positioning weight, and dynamically adjust the finder's flight speed and flight path based on the positioning distance. The positioning distance is calculated based on the obtained positioning weight and the characteristics of different positioning technologies. GPS positioning offers ideal accuracy of meters, but in practice, this is subject to interference from various factors. When obtaining the positioning distance, a relatively accurate positioning distance range is calculated based on the current GPS signal strength, the number of satellites, and a positioning error model. Based on the obtained positioning distance, the finder dynamically adjusts its flight speed and flight path in real time. If the positioning distance indicates that the finder is too close to the target vessel, approaching a range that could result in a restricted field of view or other risks (e.g., less than 100 meters), the finder will appropriately increase its flight speed and distance from the target vessel to expand its monitoring field of view. It will also adjust its flight path to maintain an effective monitoring angle for the target vessel.

[0029] The predator docking strategy is provided with an anchor point acquisition step, which is used to obtain several frames of images collected by the discoverer as target images, and obtain the target ship and other ship image information in the target image, compare the target ship with the other ships, obtain several feature images in the target ship as anchor points, and obtain the anchor point weight according to the frequency of occurrence of the anchor points on the other ships. The discoverer continuously collects images and selects several frames of images as target images. The target image contains information about several ships in the current scene. Several frames of images are used to capture ship features from multiple angles and different time points to avoid inaccurate feature extraction caused by ship movement or the limitations of single-shot images. The target image is processed and the target detection algorithm is used to quickly identify the position of the target ship in the image and obtain image information of the target ship and other ships; the target ship is compared with the other ships to find a unique image in the target ship as an anchor point. The features can be the outline, color distribution, and texture of the ship, such as the target. If a ship has a chimney with a special shape, while other ships do not, then the chimney feature can be used as an anchor point; use the feature extraction algorithm to extract the feature points in the image, and screen the unique features of the target ship as anchor points based on the comparison results; count the frequency of the anchor points on other ships; if an anchor point appears very frequently on other ships, it means that the uniqueness of this anchor point is low and the distinction in subsequent feature recognition is not great, so it is given a lower weight; conversely, if an anchor point hardly appears on other ships, it means that it has strong uniqueness and can well distinguish the target ship from other ships, so it is given a higher weight.

[0030] The feature recognition strategy includes obtaining anchor points and images captured by the predator as images to be analyzed, calculating the similarity between the images of ships in the images to be analyzed and the images of the anchor points, calculating the similarity based on the weights of the anchor points, and prioritizing the similarity to select the ship with the highest similarity as the target ship; obtaining the anchor points and their weights obtained in the anchor point acquisition step; the predator device captures images as images to be analyzed, which contain information about the ships in the current scene. The predator device may be at different positions and angles, and the images captured may differ from the target images captured by the discoverer, so feature recognition is required on these images to determine the target ship; calculating the similarity between the images of ships in the images to be analyzed and the images of the anchor points, which may be calculated using an image matching algorithm; and calculating the similarity score between each anchor point and the image of the ship in the image to be analyzed. These similarity scores are then weighted and summed based on the weights of the anchor points to obtain a comprehensive similarity score between each vessel and the target vessel. For example, if there are three anchor points, A, B, and C, with weights of 0.2, 0.3, and 0.5, respectively, and the similarity scores between a vessel in the image to be analyzed and these three anchor points are 0.8, 0.6, and 0.7, respectively, then the comprehensive similarity score between this vessel and the target vessel is 0.2 × 0.8 + 0.3 × 0.6 + 0.5 × 0.7 = 0.69. The vessels in the image to be analyzed are prioritized based on their comprehensive similarity scores; the higher the score, the more likely it is to be the target vessel. Finally, the vessel with the highest similarity is selected as the target vessel.

[0031] The predator docking step also includes a relay matching sub-step. When the predator has not entered the predation range during the tracking process of the current discoverer and the target vessel is about to leave the monitoring range of the current discoverer, the current discoverer searches for the remaining discoverers and passes the obtained target features to the next discoverer. The next discoverer docks with the next predator, and the current discoverer continues to track the target vessel until the next discoverer or predator tracks the target vessel. During the process of the current discoverer tracking the target vessel, the position status of the predator and the target vessel is continuously monitored. When the predator has not entered the range where it can prey on the target vessel and the target vessel is about to leave the monitoring range of the current discoverer, the relay matching mechanism is triggered. This situation may be caused by the high speed of the target vessel and the limited monitoring range of the current discoverer. The current discoverer will start searching for other available discoverers; the search and docking of the next discoverer will be realized through a pre-established communication network or location information system, and the current discoverer can quickly obtain the location and status information of other discoverers; the current discoverer will pass the characteristic information of the target ship obtained in the anchor point acquisition step, such as the anchor point, anchor point weight, etc., to the selected next discoverer to help the next discoverer quickly and accurately locate the target ship; after receiving the target characteristic information, the next discoverer will dock with the next predator and guide it to the location of the target ship; after completing the information transmission and docking, the current discoverer will not stop tracking the target ship immediately, it will continue tracking until the next discoverer or predator successfully takes over the tracking of the target ship, ensuring that the target ship will not be lost during the relay process.

[0032] The relay matching sub-step also includes a predator value assessment strategy, which obtains the distance and movement trajectory between the current predator and the target predator respectively, and the cross-zone parameters are preset in the current predator. The predator tracking cost is calculated separately, and the tracking cost of the current predator and the next predator is compared. The predator with the lowest tracking cost is selected to track the target ship. According to the obtained distance, movement trajectory and cross-zone parameters, the cost of tracking the target ship for the current predator and the next predator is calculated separately. The calculation of the tracking cost will take into account multiple factors. For example, the longer the distance, the more energy and time consumed in the tracking process, and the higher the cost. ; The motion trajectory is complex, requiring more turning and acceleration operations, which will also increase the cost; the cross-zone parameters will be adjusted according to the actual area crossed; if the current predator needs the current detection area to track the target ship, the cross-zone cost will be obtained, and the cross-zone cost will be adjusted accordingly based on the model strength of the next area. If the signal in the next detection area is weak, then its cross-zone cost will increase accordingly. Comparing the tracking costs of the current predator and the next predator, the predator with the low tracking cost is selected to track the target ship, ensuring that during the relay matching process, the target ship can be continuously tracked at the lowest cost.

[0033] UAV abnormal intelligent inspection, identification and analysis system, including In the target search module, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel according to the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked; The predator docking module, when the discoverer tracks the target vessel, obtains the target vessel's location and the status of surrounding predators, selects the corresponding predator, and adjusts the distance between the discoverer and the target vessel and predator through the tracking strategy, adjusts the discoverer's flight path, and transmits the target vessel's target characteristics to the selected predator. The target characteristics include the vessel's location, anchor point, and abnormal conditions. In the evidence collection step, the predator monitors the target vessel to determine whether there are any abnormalities. If so, it obtains evidence frames. If not, it obtains key frames with target features collected by the discoverer and obtains detailed images of the target vessel collected by the predator. These images are cross-compared to form a complete chain of evidence.

[0034] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. A method for intelligent inspection, identification and analysis of drone anomalies, characterized by: include In the target search step, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel according to the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked; Predator docking step: When the discoverer tracks the target vessel, it obtains the target vessel's position and the status of the surrounding predators, selects the corresponding predator, and adjusts the distance between the discoverer and the target vessel and predator through the tracking strategy, adjusts the discoverer's flight path, and transmits the target vessel's target characteristics to the selected predator. The target characteristics include the vessel's position, anchor point, and abnormal conditions. Target vessel identification step: After the predator obtains the target features, it flies to the location of the target vessel. When the target vessel enters the predator's acquisition range, it locks the target vessel through feature recognition strategy based on the anchor point on the target vessel; In the evidence collection step, the predator monitors the target vessel to determine whether there are any abnormalities. If so, it obtains evidence frames. If not, it obtains key frames with target features collected by the discoverer and obtains detailed images of the target vessel collected by the predator. These images are cross-compared to form a complete chain of evidence.

2. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 1 is characterized in that: The tracking strategy includes obtaining the moving speed, moving trajectory and environmental visibility of the target vessel, and through the positional relationship between the target vessel, the predator and the discoverer, obtaining the corresponding docking weight and positioning weight according to the data index table, calculating the tracking value, and presetting a tracking threshold. When the tracking value is less than the tracking threshold, an adjustment strategy is generated to adjust the docking weight and positioning weight so that the tracking value is greater than the tracking threshold.

3. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 2 is characterized in that: The tracking strategy also includes a docking step, which includes The data acquisition sub-step is used to dynamically adjust the communication frequency and the transmission frequency according to the distance between the discoverer and the predator, the environmental conditions, and the flight speed of both, so as to obtain the real-time communication quality between the discoverer and the predator; The docking distance calculation sub-step is used to obtain the docking weight and real-time communication quality, and obtain the docking distance through the distance adjustment algorithm, and dynamically adjust the flight speed and flight path between the discoverer and the predator according to the docking distance.

4. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 2 is characterized in that: The tracking strategy also includes a positioning step, which is used to obtain a positioning weight, obtain a positioning distance based on the positioning weight, and dynamically adjust the flight speed and flight route of the discoverer based on the positioning distance.

5. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 1 is characterized in that: The predator docking strategy is provided with an anchor point acquisition step, which is used to obtain several frames of images collected by the finder as target images, and obtain the target ship and the image information of other ships in the target image, compare the target ship with the other ships, obtain several feature images of the target ship as anchor points, and obtain the anchor point weight according to the frequency of occurrence of the anchor point on the other ships.

6. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 5 is characterized in that: The feature recognition strategy includes obtaining anchor points and images collected by the predator as images to be analyzed, respectively calculating the similarity between the vessel images in the images to be analyzed and the anchor point images, and calculating the similarity based on the anchor point weights, and performing priority sorting based on the similarity, and selecting the vessel with the highest similarity as the target vessel.

7. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 1 is characterized in that: The predator docking step also includes a relay matching sub-step. When the predator has not entered the predation range during the tracking process of the current discoverer and the target vessel is about to leave the monitoring range of the current discoverer, the current discoverer searches for the remaining discoverers and passes the obtained target features to the next discoverer. The next discoverer docks with the next predator, and the current discoverer continues to track the target vessel until the next discoverer or predator tracks the target vessel.

8. The method for intelligent inspection, identification and analysis of drone anomalies according to claim 7, characterized in that: The relay matching sub-step also includes a predator value assessment strategy, which obtains the distance and movement trajectory between the current predator and the target predator respectively, and the cross-zone parameters are preset in the current predator. The predator tracking cost is calculated respectively, and the tracking cost of the current predator and the next predator is compared. The predator with the lowest tracking cost is selected to track the target vessel.

9. UAV abnormal intelligent inspection, identification and analysis system, characterized by: include In the target search module, the discoverer searches for the vessel and determines whether there is any abnormality in the vessel according to the abnormality identification strategy. If there is any abnormality, the vessel is marked as the target vessel and tracked; The predator docking module, when the discoverer tracks the target vessel, obtains the target vessel's location and the status of surrounding predators, selects the corresponding predator, and adjusts the distance between the discoverer and the target vessel and predator through the tracking strategy, adjusts the discoverer's flight path, and transmits the target vessel's target characteristics to the selected predator. The target characteristics include the vessel's location, anchor point, and abnormal conditions. In the evidence collection step, the predator monitors the target vessel to determine whether there are any abnormalities. If so, it obtains evidence frames. If not, it obtains key frames with target features collected by the discoverer and obtains detailed images of the target vessel collected by the predator. These images are cross-compared to form a complete chain of evidence.

Images