Tracking method and device for special shooted objects and tracking method and device for aspect parts

Abstract

The present invention provides a specific shot body tracking method based on characteristic region detection and equipment, and characteristic region tracking method and its equipment. The specific shot body tracking method based on characteristic region detection is characterized by that it utilizes particle filter to divide multi-frame image of video sequence by means of sub window and make calculation treatment so as to define the position and size of shot body. The described characteristic region is one portion of specific shot body. The observation model of the described particle filter includes characteristic region observation model.

Description

technical field [0001] The present invention relates to a tracking method and device for a specific subject based on feature part detection, and a feature part tracking method and device. More specifically, the present invention relates to a head tracking method and device based on face detection, and a face tracking method and device. Background technique [0002] Object (such as face, head, other human body parts, cars, pedestrians, etc.) tracking is a research topic that is generally concerned in the field of computer vision, and head tracking, as one of its sub-problems, has a particularly broad application prospect. Motion analysis, camera autofocus, intelligent broadcasting and human-computer interaction all play an important role. [0003] Due to the complexity of the tracking problem itself, in the field of computer vision, there are a large number of literatures that study various forms of tracking problems from various perspectives. Not only the application envir...

AI Technical Summary

Problems solved by technology

Part of the reason is that, on the one hand, tracking is faced with video sequences. Once the tracked target is initialized, there is a large amount of inter-frame similarity information that can be used, and it does not require an understanding of the special pattern of the face; relatively In terms of face detection, the problem of face detection is much more complicated. The features used are all learned from a large number of samples, and the classifier used for detection is also quite complicated.

On the other hand, tracking problems generally require real-time performance, that is, the requirements for speed are relatively high, so detection is generally only used as the initialization of the tracking process

Document [5] is a document that is relatively close to the method of the present invention, but document [5] uses the detector of document [9] for face tracking, and its main defect is that the processing speed is relatively slow (26s per frame (352× 288 pixels))

Images