A Visual Tracking Method Based on Optimizing Scale and Angle Based on Foreground Conditional Probability

Abstract

The invention discloses a visual tracking method based on foreground condition probability optimization scale and angle, which belongs to the field of computer vision. The implementation steps are as follows: 1) read the video frame sequence, and use the SiamMask method to calculate the frame regression frame, segmentation mask (foreground) and minimum bounding frame; 2) calculate the proportion of the foreground area in the minimum bounding frame; 3) when the ratio When it is less than the set threshold, calculate the reliability of the minimum bounding box of the frame; 4) Select different strategies to optimize the scale of the minimum bounding box according to the reliability; 5) Set the offset for the angle of the tracked frame after scale optimization; 6) Calculate the IoU value of the rotation frame and the foreground at each offset angle; 7) The tracker adaptively outputs the rotation frame with the largest IoU value of the foreground. The visual tracking method effectively improves the overall performance of target tracking under complex conditions such as target motion, rotation, and scale change.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a visual tracking method for optimizing scale and angle based on foreground conditional probability. Background technique [0002] Object tracking is a hot issue in computer vision, and it is the premise and basis for higher-level image understanding. It is widely used in intelligent video surveillance, human-computer interaction, visual navigation, medical diagnosis and other fields. In the process of tracking the target, it often encounters interference from the background (occlusion, illumination changes, etc.) Object tracking is an extremely challenging problem in the field of computer vision. [0003] In recent years, tracking methods have gradually shifted from generative to discriminative, and discriminative tracking methods are represented by correlation filtering and deep learning. At present, the Siamese network series based on deep learning combines the advantages of c...

AI Technical Summary

Problems solved by technology

[0006] The present invention is a visual tracking method based on the foreground conditional probability optimization scale and angle aiming at the inaccurate scale and angle of the tracking frame, aiming to solve the problem of the scale and angle of the tracking frame caused by complex scenes such as target movement, rotation, and scale change. Instability and Inaccuracy

Images