Method for realizing fishing behavior detection processing based on codec structure and corresponding semantic segmentation network system

Abstract

The invention relates to a method for realizing fishing behavior detection processing based on a codec structure. The method comprises the following steps: capturing a high-definition image from a video monitoring system at regular intervals, and carrying out nine-block segmentation and size normalization processing; inputting each block image into a target detection neural network to detect whether a portrait is contained or not; inputting the block image of the detected portrait into a semantic segmentation network of a coding and decoding structure to perform semantic segmentation detectionof the fishing rod; judging whether a block image detects a fishing rod subjected to pixel-level segmentation and is overlapped with a portrait detection frame or not; determining whether fishing behavior exists. The invention further relates to a semantic segmentation network system based on the codec structure. The method for realizing fishing behavior detection processing based on the codec structure and the corresponding semantic segmentation network system are adopted. According to the method, strong prior space information with a long-distance continuous shape can be propagated on neurons on the same layer of a high-level semantic network layer, the receptive field of a feature map is increased, and the capture of long-distance context information is further enhanced.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to the technical field of deep neural network, and specifically refers to a method for detecting and processing fishing behaviors based on a codec structure and a corresponding semantic segmentation network system. Background technique [0002] The power supply department usually installs video surveillance equipment near the pond near the power line, and then intelligently analyzes the video surveillance images, and outputs an alarm signal when the image intelligent analysis system detects fishing behavior. [0003] Due to factors such as the monitoring depth of field of the monitoring camera, lighting conditions, and the complex background of the monitoring image, the detection accuracy of the fishing behavior by the intelligent video analysis system based on conventional image processing technology is often not high. [0004] The application of target detection t...

AI Technical Summary

Problems solved by technology

[0003] Due to factors such as the monitoring depth of field of the monitoring camera, lighting conditions, and the complex background of the monitoring image, the detection accuracy of the fishing behavior by the intelligent video analysis system based on conventional image processing technology is often not high

[0004] The application of target detection technology based on deep neural network to phishing behavior detection is a research hotspot at present, but target detection networks based on rectangular bounding boxes (Bounding Box) such as Fast-RCNN, YOLO and other deep network structures are more suitable for detecting compact targets ( Such as people, cars, etc.), not suitable for detecting linear objects with low fill rate in the bounding box

Since the fishing rod is relatively slender and occupies a very small area in the Bounding Box, when using deep neural networks such as Fast-RCNN and YOLO for target detection, it still cannot handle this kind of strong prior structure with few appearance clues. target, its detection accuracy for fishing rods is still not high, and it is easy to cause missed and false detections

Images