Ship multi-target tracking method based on YOLO V5 algorithm

Abstract

The invention discloses a ship multi-target tracking method based on a YOLO V5 algorithm, and the method comprises the steps: carrying out the screening and marking of collected ship image data, building a data set, and dividing the data set into a training set, a verification set, and a test set; training the YOLO V5 network by using a training set and a verification set in the self-established data set to obtain a ship detection model based on the YOLO V5 network and a weight file; detecting the test set by using the trained YOLO V5 detection model, outputting a detection result, and evaluating the detection model; based on the trained YOLO V5 detection model, generating a tracking model through DeepSort algorithm processing; and carrying out real-time verification on the generated DeepSort tracking model. According to the invention, marine ship detection and multi-target tracking can be realized, the detection precision is high, the real-time performance is good, and the speed is fast.

Description

technical field [0001] The invention relates to the field of machine vision, in particular to a ship multi-target tracking method based on the YOLO V5 algorithm. Background technique [0002] Target tracking is to detect, extract, identify and track the moving target in the time series of continuous frame images, obtain the position and trajectory of the tracked target, realize the understanding of the behavior of the moving target, and complete higher-level detection tasks. According to the number of tracking targets, tracking algorithms can be divided into single target tracking and multi-target tracking. Compared with single-target tracking, the problem of multi-target tracking is more complex and difficult. The main task of multi-target tracking is to locate multiple targets of interest in a given video at the same time, maintain the ID of each target of interest and record their trajectories. [0003] For the specific environment at sea, firstly, the ship will be affe...

AI Technical Summary

Problems solved by technology

[0003] For the specific environment at sea, firstly, the ship will be affected by the wind, wave and surge during navigation, which will cause the camera to be unable to stably acquire image data with a fixed angle of view, resulting in difficulty in target recognition and large errors in the multi-target detection of the ship; secondly, at sea The changeable climate makes the camera produce different light and shadow pictures when acquiring image data, which makes the traditional target detection algorithm unable to detect the target ideally, and the traditional target detection algorithm is difficult to overcome the overlapping of the detection target when the ship is driving and occlusion; finally, the problem of ship multi-target tracking lies more in the accuracy of target detection on its basis. In the target tracking algorithm, the accuracy of target detection affects the tracking effect

[0004] The ship multi-target tracking method is the core of the offshore ship intelligent monitoring system. The existing offshore ship intelligent monitoring system can only realize ship detection, and the speed is slow, the precision is low, and the real-time performance is not good, so the target ship cannot be tracked.

The existing target tracking methods can only meet the requirements of land and other relatively stable detection environments, and the navigation environment of marine ships is complex and changeable, which cannot guarantee detection accuracy and real-time performance.

Images