Multi-source camera moving target tracking method

Abstract

The invention discloses a multi-source camera moving target tracking method. The method comprises the following steps: acquiring a target detection total number episode; training the acquired episode of the total target detection number by using a yov5 detection model to obtain a model file; defining a tracking type variable of a global variable, and calling the model file to obtain a tracking target feature; using a feature acquisition function to acquire tracking target features, loading and calculating target depth features, transmitting calculated data features into a kalman algorithm and a matching algorithm for packaging, and performing filtering tracking processing and matching processing on a class object, namely a target; judging the authenticity of the target by using a linear distribution function and an intersection-to-union ratio function; updating, adding or deleting the target coordinate and the image information of the target according to the target judgment result; and after the updated targets are fused, target information is output.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a multi-source camera moving target tracking method. Background technique [0002] With the development of artificial intelligence and the improvement of hardware level, the field of visual image technology is also rapidly updating and developing. Various networks and target detection networks emerge in an endless stream. The application requirements of the target tracking direction are also increasing, so the tracking algorithms or models for tracking, tracking of multiple targets, multi-camera scenes, tracking of moving targets, etc. are also facing the test of practical application effects. Therefore, the technical research on moving target tracking in multi-camera scenes is of great research and practical significance. [0003] At present, the Chinese Patent Publication No. 113160283 discloses a target tracking method based on SIFT in a multi-camera scene; includ...

AI Technical Summary

Problems solved by technology

In this tracking method, the target is tracked by combining model training and deep network algorithm. However, when acquiring an image, a motion algorithm is required to splicing the captured video, which has the problems of complex calculation of the tracking object and lack of accuracy.

Images