Fishing boat illegal behavior detection method based on video analysis

Abstract

The invention discloses a method for detecting illegal behaviors of a fishing boat based on video analysis. The method comprises the following steps: S1, establishing a detection model; s2, data acquisition: acquiring a monitoring video frame of a monitored water area in real time through a camera or an unmanned aerial vehicle and inputting the monitoring video frame into the detection model; s3, target detection: the detection model performs target detection on the monitoring video frame by using a YOLO algorithm, and identifies elements in the fishing boat and the surrounding environment; s4, behavior analysis: the detection model analyzes the behavior pattern of the fishing boat by using a deep learning algorithm, and identifies potential illegal behaviors; s5, data verification: the detection model carries out secondary verification on the suspected illegal behaviors through a set rule or algorithm; and S6, early warning notification: the detection model sends early warning information to a supervision mechanism. Fishing boat illegal behaviors in complex water areas can be recognized in real time, intelligent alarm is achieved, and the recognition accuracy of the fishing boat illegal behaviors is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fishing vessel supervision, and in particular to a method for detecting illegal fishing vessel behaviors based on video analysis. Background Art

[0002] Strengthening the monitoring of marine fishing vessel behavior is an important measure to adapt to the rapid development of my country's marine fisheries, and is also a necessary means to promote the standardized development of my country's marine fisheries.

[0003] With the continuous development of big data, the concept of intelligent fisheries has been continuously proposed, and deep learning algorithms have also played an important role in the field of fisheries. Traditional fishing vessel behavior recognition is mostly based on fishing vessel VMS data or satellite radar and GPS data, and fishing vessel behavior recognition models are established through various algorithms. However, there are deficiencies in the depth of behavior analysis, learning mechanism, environmental adaptability, and decision-making assistance. It is difficult to adapt to complex scenarios, and the accuracy of identifying illegal fishing vessel behaviors is not high. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for detecting illegal fishing vessel behaviors based on video analysis, which can identify illegal fishing vessel behaviors in complex waters in real time, realize intelligent alarm, and improve the recognition accuracy of illegal fishing vessel behaviors.

[0005] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0006] A method for detecting illegal fishing vessel behaviors based on video analysis, comprising the following steps: S1. Establishing a detection model; S2. Data collection: Obtaining surveillance video frames of the monitored waters in real time through a camera or an unmanned aerial vehicle and inputting them into the detection model; S3. Target detection: The detection model uses the YOLO algorithm to perform target detection on the surveillance video frames and identify elements in the fishing boat and the surrounding environment; S4. Behavior analysis: The detection model uses a deep learning algorithm to analyze the behavior patterns of fishing boats and identify potential illegal behaviors; S5. Data verification: The detection model performs secondary verification on suspected illegal behaviors through set rules or algorithms; S6. Early warning notification: The detection model sends early warning information to the regulatory agency.

[0007] Preferably, the detection model is trained based on the collected surveillance video frames and derived surveillance video frames to obtain and construct an illegal behavior feature library; the detection model adopts an online learning or transfer learning mechanism to achieve continuous optimization according to the real-time acquisition of surveillance video frames of the monitored waters during operation, and automatically adapt to different sea areas, different types of fishing boat behavior patterns and newly emerging illegal behavior patterns.

[0008] Preferably, the detection model in step S3 also uses night vision enhancement and severe weather filtering technology to process the surveillance video frames before performing target detection.

[0009] Preferably, in step S4, the detection model uses a deep learning algorithm to analyze the behavior pattern of the fishing boat based on the identified key elements of the fishing boat and its surrounding environment, combined with the location information and historical trajectory of the fishing boat, and compares it with the illegal behavior feature library to identify potential illegal behaviors.

[0010] Preferably, in step S5, comprehensive environmental factors are taken into consideration, the suspected illegal behavior is secondary verified through multimodal data fusion, and a confidence score is output.

[0011] Preferably, in step S6, the detection model evaluates the severity of the illegal behavior according to the confidence score and sends a warning message, which includes specific evidence of the illegal behavior and preliminary response suggestions.

[0012] Due to the adoption of the above technical scheme, the technical progress achieved by the present invention is as follows.

[0013] The present invention can effectively determine potential illegal behaviors by combining night vision enhancement and severe weather filtering technology, the YOLO algorithm and the deep learning algorithm; by considering comprehensive environmental factors, potential illegal behaviors can be verified secondary, thereby meeting the detection needs of fishing vessel violations in complex scenarios, achieving effective monitoring and timely early warning of fishing vessel violations, and improving the efficiency and accuracy of marine resource protection and fishery supervision. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0015] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] A method for detecting illegal fishing boat behaviors based on video analysis mainly detects the behaviors of fishing boats at sea through real-time monitoring and intelligent recognition technology to determine whether there are illegal behaviors such as illegal fishing and cross-border operations. Figure 1 As shown, the following steps are included:

[0017] S1. Establish a detection model.

[0018] The detection model is built based on YOLO algorithm, deep learning algorithm, night vision enhancement and bad weather filtering technology. The detection model is trained based on the collected surveillance video frames and derived surveillance video frames to build a feature library of illegal behavior. The detection model adopts online learning or transfer learning mechanism to achieve continuous optimization based on the real-time acquisition of surveillance video frames in the monitored waters during operation, automatically adapting to different sea areas, different types of fishing boat behavior patterns and emerging illegal behavior patterns, improving the timeliness and flexibility of the model.

[0019] Specifically, the following technologies are mainly used for the training of the detection model:

[0020] ① Generate a private convolution layer to realize feature extraction. At the same time, the filter performs convolution operation on the input surveillance video frame to extract the features in the surveillance video frame, and use the fully connected layer to obtain the prediction value.

[0021] ② A method for generating multi-scale feature extraction. This method captures targets of different sizes through feature extractors of different scales, thereby improving the model's detection capabilities in complex scenarios.

[0022] ③ Generate a private classifier, which is responsible for mapping the extracted features to specific categories and mapping the distributed feature representation to sample labels to complete the classification task.

[0023] ④Parameter settings, including learning rate, batch size, optimizer selection, etc. These parameters directly affect the training effect and generalization ability of the model.

[0024] ⑤Increase the diversity of training data so that the model can better cope with various complex scenes and lighting conditions.

[0025] ⑥ Train the model through regularization technology, use L1 or L2 regularization technology to reduce overfitting and improve the generalization ability of the model on unknown data.

[0026] ⑦ Through the dynamic threshold adjustment model, the confidence threshold of the detection result is dynamically adjusted according to the actual situation to adapt to different application scenarios.

[0027] S2. Data collection: Obtain surveillance video frames of the monitored waters in real time through cameras or drones and input them into the detection model.

[0028] Use cameras or drone aerial photography equipment to capture real-time sea surface images and video streams, and input the acquired surveillance video frames into the detection model.

[0029] S3. Object detection: The detection model uses the YOLO algorithm to perform object detection on the surveillance video frames and identify the fishing boat and elements in the surrounding environment.

[0030] Before performing target detection, the detection model also uses night vision enhancement and severe weather filtering technology to process the surveillance video frames, so that the detection model can maintain high efficiency in the complex and changeable marine environment and the detection results are not affected by factors such as light and fog.

[0031] The YOLO algorithm can directly predict object bounding boxes and category probabilities from surveillance video frames. It has the characteristics of fast speed and high accuracy and is suitable for real-time monitoring scenarios.

[0032] S4. Behavioral Analysis: The detection model uses deep learning algorithms to analyze the behavioral patterns of fishing vessels and identify potential violations.

[0033] The detection model uses deep learning algorithms to analyze the behavior patterns of fishing boats, such as fishing movements and navigation tracks, based on the identified fishing boats and key elements in their surroundings, such as other ships, fishing nets, buoys, etc., combined with the location information and historical tracks of the fishing boats, and compares them with the illegal behavior feature library to identify potential illegal behaviors. For example, if a fishing boat frequently appears in waters where fishing is prohibited, or operates at unauthorized times, it can be judged as a potential illegal behavior.

[0034] S5. Data verification: The detection model conducts secondary verification of suspected illegal activities through set rules or algorithms.

[0035] The potential illegal behavior data will be further analyzed and reconfirmed through the set rule engine or more complex deep learning algorithm to improve the accuracy of illegal behavior judgment. For example: comprehensive environmental factors are considered, more environmental factors are considered, such as the impact of weather and ocean currents on ship behavior, and suspected illegal behaviors are re-verified through multi-modal data fusion (such as satellite images and radar data) to output confidence scores.

[0036] S6. Early warning notification: The detection model sends early warning information to the regulatory agency.

[0037] In addition to basic detection and alarm, the detection model integrates a more advanced decision support system, which can evaluate the severity of illegal behavior based on confidence scores and put forward preliminary response suggestions. Confirmed illegal behavior will trigger the early warning mechanism, which will automatically send early warning information to the regulatory agency. The early warning information not only contains specific evidence of illegal behavior, provides a detailed chain of evidence, including timestamps, location coordinates, fishing vessel identification and other information, but also includes preliminary response suggestions.

[0038] S7. Subsequent processing: The regulatory agency takes corresponding law enforcement actions based on the early warning information.

[0039] A system for detecting illegal fishing vessel behaviors based on video analysis includes a front-end data acquisition module, a back-end processing center and an early warning module.

[0040] Front-end data acquisition module: including high-definition cameras and drones, responsible for collecting real-time monitoring video frames.

[0041] The back-end processing center includes technologies such as YOLO algorithm, deep learning algorithm, night vision enhancement and severe weather filtering technology, which are used for image processing and illegal behavior identification; it also includes a data storage and management module for storing original image data, analysis results and warning records.

[0042] The early warning module is used to automatically send early warning information to regulatory agencies and provide a visual display and management interface for illegal activities to facilitate operations by law enforcement personnel.

[0043] The front-end data acquisition module, the back-end processing center and the early warning module are connected through the network in turn for data transmission. The front-end data acquisition module transmits the monitoring video frame data to the back-end processing center. The processed results are stored in the data storage and management module and sent to the supervision platform through the early warning module.

[0044] In order to efficiently process large amounts of video data while maintaining the integrity of key information, the system mainly uses the following data processing and compression technologies:

[0045] ① Reduce storage and transmission costs through efficient video coding (VVC) methods. At the same time, pre-processing of video data, including video shot segmentation, key frame extraction and feature extraction, helps to more effectively identify and analyze video content in subsequent processing.

[0046] ②Convert unstructured video data into structured data so that key information in the video can be effectively identified and processed by computers. Deep learning technology also plays an important role in video processing. It uses algorithms to efficiently process massive amounts of unstructured video image data, achieving rapid retrieval, intelligent identification, and understanding of data.

[0047] In order to ensure that personal privacy on fishing vessels is not violated while monitoring illegal activities, the system mainly adopts the following privacy protection mechanisms:

[0048] ①Ensure the security of data transmission and storage through AES algorithm encryption.

[0049] ② Through anonymization and data desensitization technology, sensitive information is processed so that it cannot directly identify personal identities, effectively preventing personal privacy information from being leaked and abused.

[0050] ③ Use differential privacy technology to define privacy when sharing data and provide privacy protection while ensuring data availability.

[0051] ④ The use of homomorphic encryption technology allows operations on encrypted data without decryption, thereby performing data analysis and processing without exposing the original data.

[0052] ⑤ During data transmission and storage, the model is encrypted using private password encryption to ensure the security of the model during transmission. At the same time, hash functions and digital signatures are used to verify the integrity and source of the model to prevent the model from being tampered with or replaced.

[0053] The present invention has the following advantages when used:

[0054] (1) Improving detection accuracy: Using advanced deep learning algorithms, such as convolutional neural networks (CNN) and recurrent neural networks (RNN), to determine potential illegal behaviors, and conducting secondary verification of potential illegal behaviors by considering comprehensive environmental factors, it can more accurately identify fishing vessel behaviors, including subtle illegal operations, and reduce false positives and missed positives.

[0055] (2) Enhanced real-time performance: Optimized video processing technology and efficient data transmission protocols ensure that the system can analyze video streams in real time, respond quickly, and provide instant intelligence for law enforcement operations.

[0056] (3) Improved environmental adaptability: By introducing technologies such as night vision enhancement and severe weather filtering, the system can maintain high-efficiency operation in complex and changeable marine environments and is not affected by factors such as light and fog.

[0057] (4) Intelligent learning and adaptation: It has the ability of self-learning, continuously optimizes according to the newly collected data, automatically adapts to the behavior patterns of fishing vessels in different sea areas and types, and improves recognition capabilities.

Claims

1. A method for detecting illegal fishing vessel behaviors based on video analysis, characterized in that: The following steps are involved: S1. Establish a detection model; S2. Data collection: obtain the monitoring video frames of the monitored water area in real time through cameras or drones and input them into the detection model; S3. Target detection: The detection model uses the YOLO algorithm to detect targets in surveillance video frames and identify fishing boats and elements in the surrounding environment; S4. Behavior analysis: The detection model uses deep learning algorithms to analyze the behavior patterns of fishing vessels and identify potential violations; S5. Data verification: The detection model conducts secondary verification of suspected illegal behaviors through set rules or algorithms; S6. Early warning notification: The detection model sends early warning information to the regulatory agency.

2. The method for detecting illegal fishing vessel behaviors based on video analysis according to claim 1 is characterized in that: The detection model is trained based on the collected surveillance video frames and derived surveillance video frames to obtain and construct an illegal behavior feature library; the detection model adopts an online learning or transfer learning mechanism to achieve continuous optimization during operation based on the real-time acquisition of surveillance video frames in the monitored waters, and automatically adapt to different sea areas, different types of fishing boat behavior patterns and newly emerging illegal behavior patterns.

3. The method for detecting illegal fishing vessel behaviors based on video analysis according to claim 1 is characterized in that: In step S3, the detection model also uses night vision enhancement and severe weather filtering technology to process the surveillance video frames before performing target detection.

4. The method for detecting illegal fishing vessel behaviors based on video analysis according to claim 1, characterized in that: In step S4, the detection model uses a deep learning algorithm to analyze the behavior pattern of the fishing boat based on the key elements of the identified fishing boat and its surrounding environment, combined with the location information and historical trajectory of the fishing boat, and compares it with the illegal behavior feature library to identify potential illegal behaviors.

5. The method for detecting illegal fishing vessel behaviors based on video analysis according to claim 1 is characterized in that: In step S5, comprehensive environmental factors are taken into consideration, and the suspected illegal behavior is secondary verified through multimodal data fusion, and a confidence score is output.

6. The method for detecting illegal fishing vessel behaviors based on video analysis according to claim 5, characterized in that: In step S6, the detection model evaluates the severity of the illegal behavior according to the confidence score and sends a warning message, which contains specific evidence of the illegal behavior and preliminary response suggestions.

Images