Method and device for filtering outliers in three-dimensional model tracking

Abstract

The invention provides a foreign point filtering method and device in three-dimensional model tracking. The method comprises the following steps of: extracting contours or feature points in a currentframe image; Obtaining a color model based on the contour sampling points or the feature points; Reading the next frame of image, and predicting the initial position of the contour or the feature point in the next frame according to the motion model; Optimizing the contour in the next frame or searching corresponding feature points according to the initial position; On the basis of the color model, calculating energy values of color probability distribution of the contour sampling points or the feature points, and filtering out outer points according to the calculated energy values; And solving the attitude of the next frame, and replacing the next frame with the current frame. According to the method, the problems of high light, complex environment, shielding and the like of three-dimensional model tracking in an actual scene can be solved, so that the success rate and robustness of three-dimensional tracking are improved.

Description

technical field [0001] The embodiments of the present invention relate to the field of computer vision, and in particular to a method and device for filtering outliers in 3D model tracking. Background technique [0002] Tracking technology was mainly used in plane tracking in the early days. Specifically, in a certain amount of video sequences, specific targets (such as vehicles, people, billboards, etc.) are tracked, and the position of the target in the video is marked in the video sequence. . Object tracking has a wide range of applications in the field of computer vision: such as video surveillance, traffic flow monitoring, unmanned driving, face recognition, augmented reality (augmented reality, AR), etc. For example, in face recognition, it is necessary to capture the position of the face in the video sequence in real time in order to extract face-related features for recognition; in the field of AR, it is necessary to capture the position of the mark in the video seq...

AI Technical Summary

Problems solved by technology

[0011] In view of the above problems, the purpose of the embodiments of the present invention is to provide a method and device for filtering outliers in 3D model tracking, which can solve the problems of highlights, complex environments and occlusions in 3D model tracking in actual scenes, thereby improving the success of 3D tracking. rate and robustness

Images