AAM-based object location method

Abstract

This invention discloses one object position method based on AAM, which is based on the test and trace result to get position information and comprises the following steps: a, through object sample training to get different positions AAM modes, wherein only for object grey characteristics PCA mode to get the AAM mode composed of grey module as even value object; b, getting the current object image according to the object test and trace process to center point position, size, front frame object rotation angle; c, according to the even object to get each input object image relative to train sample affine conversion parameters; d, getting current object position information according to affine conversion parameters.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an AAM-based object positioning method in the process of face detection and tracking. Background technique [0002] Human face is one of the most convenient ways of human-computer interaction in computer vision system. Face detection is to determine the position, size and other information of all faces in an image or image sequence, while face tracking is to continuously track one or more detected faces in a video sequence. Face detection and tracking technology is not only a necessary prerequisite for face recognition, expression recognition, face synthesis and other technologies, but also has a wide range of application values ​​in intelligent human-computer interaction, video conferencing, intelligent monitoring, video retrieval and other fields. [0003] The image targeted by this system is the video sequence input by the video camera. Previously, the present applicant has p...

AI Technical Summary

Problems solved by technology

[0004] However, the algorithm in the above patent application also has certain limitations

First of all, the algorithm only trains the detection model of the frontal face, and cannot detect the profile face, which means that the detection and verification of this face can only be aimed at the frontal face, which greatly limits the application range of the algorithm; secondly, the The algorithm only tracks the face through the skin color histogram, and the skin color features of the face are very susceptible to interference from other skin color areas such as the neck, hands, or similar skin color areas such as yellow clothes. It is reflected in the tracking results that the tracking area sometimes jumps to the hands and neck or yellow clothes; again, the size and position of the tracking area obtained by the original algorithm change drastically, and even if the face remains still, the tracking result will vibrate; and the algorithm cannot obtain further details of the face. Posture information, such as the rotation angle of the face, the current general posture, etc.

Images