A ship detection method in a bridge anti-collision system

Abstract

The invention discloses a ship detection method in a bridge anti-collision system, comprising the following steps: adopting Zhang Zhengyou checkerboard calibration method to perform camera calibration; image acquisition and preprocessing; drawing a virtual channel; combining a mixed Gaussian background method and a three-frame difference method Detect the moving foreground; extract the feature value of the moving foreground target as the sample data for training the deep neural network DNN; build, train and test the deep neural network DNN; identify the ship features in the real-time video stream according to the trained deep neural network DNN , to mark the moving ship; according to the calibration results, regularly monitor the navigation status of the ship. The ship detection method based on GMM and three-frame difference method designed by the present invention overcomes the problems of broken foreground contours and incomplete targets in the traditional method, and uses the characteristics of foreground ships to train the deep neural network DNN, which can accurately and intelligently identify ships , improving the accuracy, efficiency and real-time performance of ship detection.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a ship detection method in a bridge anti-collision system. Background technique [0002] In recent years, the number of ships in inland waterways has been increasing day by day, and inland waterways are crowded. Hundreds of water traffic safety accidents inevitably occur every year, causing immeasurable property and personnel losses. The bridge anti-collision warning system came into being. The moving ship recognition algorithm plays a very important role in the bridge anti-collision warning system. Early warning to ensure that ships pass through the bridge navigation port smoothly and avoid accidents such as ship-bridge collisions. Therefore, a reliable moving ship recognition algorithm is of great significance for ship navigation monitoring. [0003] At present, there are mainly three types of methods for detecting moving targets in static backgrounds at home and abr...

AI Technical Summary

Problems solved by technology

[0003] At present, there are mainly three types of methods for detecting moving targets in static backgrounds at home and abroad: target visual detection algorithm based on background difference method, frame difference method, optical flow method, etc., but only these three types of methods are used to detect ships. There are problems such as incomplete targets and broken contours

When observing a ship from a distance, due to the perspective effect, its motion speed appears to be slow, and the difference between adjacent frames of the video sequence is small, resulting in missed detection by the system

Images