Anti-interference contour tracing method based on prior model constraint

Abstract

The invention provides an anti-interference contour tracing method based on prior model constraint. The method comprises the following steps: establishing an initial prior model of a video image to be traced as a target model by using an SLIC algorithm and a mean shift clustering algorithm; carrying out level set evolution on a current frame of image by target feature distribution of the target model; determining a target region of the current frame of image by the SLIC algorithm and the mean shift clustering algorithm; and updating the prior model according to an occlusion condition of the current image. By the method, interference of non-target information in the image in an image tracing process is removed, so that description of the prior model to a target is accurate. Accuracy of the model is guaranteed.

Description

Technical field [0001] The invention belongs to the technical field of computer vision, and specifically relates to an anti-interference contour tracking method based on prior model constraints. Background technique [0002] Target tracking technology has been one of the hot issues in the field of computer vision research since it was proposed. It can be simply defined as a series of motion trajectories left on the image plane when the target is estimated to move regularly or irregularly around a specific scene. That is, through a series of extraction, recognition, and classification operations on the objects in the video sequence, information such as their shape, position, and contour can be obtained. Then use the registration technology for analysis and processing to complete the tracking of the object of interest. As a hot research field, target tracking incorporates some advanced technologies in multiple fields. Such as graphic image processing, target detection, behavior p...

AI Technical Summary

Problems solved by technology

[0005] However, the above five methods have the following problems: complex background problem: when the lighting changes in the video image, the static object in the image becomes a moving object, and the moving object becomes a constant object, using the background to do subtraction is easy to cause video damage. The foreground is indistinguishable from the background in an image sequence

Out-of-plane rotation means that when the rotation of the moving target occurs in a two-dimensional plane that is not parallel to the camera, the tracking is carried out in three-dimensional space, and the processing method using two-dimensional images often causes the local features of the target to be lost. Through three-dimensional The rest of the data is obtained by modeling, which increases the difficulty of algorithm implementation

Images