Cam-shift-based prediction tracking method

Abstract

The invention discloses a Cam-shift-based prediction tracking method. A Kalman filtering and Cam-shift algorithm combination method is proposed; and a linear prediction and Cam-shift algorithm combination-based target prediction tracking method is proposed, namely, a linear prediction method replaces Kalman filtering to finish prediction estimation, and a prediction estimation result is substituted into a Cam-shift algorithm to perform tracking. In order to well finish position prediction tracking of a moving target in a complex background with interference of factors such as shielding and the like, the Kalman filtering and Cam-shift algorithm combination method is researched. The Kalman filtering can finish position prediction of the moving target more accurately; and the tracking can be well finished by combining the Kalman filtering with the Cam-shift algorithm. Based on this, the linear prediction and Cam-shift algorithm combination-based target prediction tracking method is proposed, namely, the linear prediction method replaces the Kalman filtering to finish the prediction estimation, and the prediction estimation result is substituted into the Cam-shift algorithm to perform the tracking. An experiment shows that the method can not only ensure the accuracy of tracking in a shielding process but also shorten the iterative time, and can better meet the requirement of timeliness.

Description

Technical field: [0001] The invention relates to a prediction tracking method based on Cam-shift. Background technique: [0002] The tracking and prediction of moving target trajectories is an indispensable part of the field of artificial intelligence research, and is a key technology for machine vision and target tracking and detection. Different application backgrounds have different tracking methods: L-K (Lucas & Kanade) optical flow method and H-S (Horn & Schunck) optical flow method do not need to consider background information, by assigning velocity vectors to pixels, tracking the target according to the difference between the target and the background velocity vector , but the brightness between adjacent frames is required to be constant; the graph cut algorithm maps the target tracking problem to the energy function, and optimizes and solves the target area, which is computationally complex; the Snake model algorithm proposed by Kass [3-4] It is a tracking method b...

AI Technical Summary

Problems solved by technology

Different application backgrounds have different tracking methods: L-K (Lucas&Kanade) optical flow method and H-S (Horn&Schunck) optical flow method do not need to consider background information, by assigning velocity vectors to pixels, and tracking the target according to the difference between the target and the background velocity vector , but the brightness between adjacent frames is required to be constant; the graph cut algorithm is to map the target tracking problem into an energy function, optimize and solve the target area, and the calculation is complicated; the Snake model algorithm proposed by Kass [3-4] It is a tracking method based on contour edge features. Its energy utilization is the smallest, but the contour initialization is complicated; Dong Chunli et al. combined particle filter and GVF-Snake, and proposed an adaptive nonlinear filtering algorithm to track moving and deformed targets. But the efficiency is low; and Yizong Cheng [6] The proposed Mean-shift algorithm is a density gradient non-parametric estimation algorithm based on color features. It is simple in form and easy to operate, but it does not have the function of updating the search window in real time, and the scale of moving targets changes rapidly, which easily leads to loss of target tracking; Bradski in Mean- The Cam-shift algorithm proposed on the basis of the shift algorithm can automatically adjust the size of the search window in real time, and retains the advantages of the Mean-shift algorithm, but when the color characteristics of the moving target are not obvious or there are factors such as occlusion, it will cause Tracking failed

Images