Method of selecting matching template in image matching tracking

Abstract

The invention belongs to the technical field of computer vision and image matching, and particularly relates to a method of selecting a matching template in image matching tracking. The method comprises two steps including a step of self-adaptively determining the size of the matching template and a step of judging the locating capability of the matching template; (1) the method flow of the step of self-adaptively determining the size of the matching template comprises the steps of taking the position of a given coordinate point as a center, and setting a split area; carrying out binaryzation on a segmentation area by using an otsu threshold segmentation algorithm and an automatic threshold segmentation algorithm, carrying out statistics on the number of pixels in a foreground region and a background region, and calculating the specific value of the number of the pixels in the foreground region and the number of the pixels in the background region; respectively taking a central point of the segmentation region as a center of pixels in 8 neighbourhood pixel points, and extracting 8 image blocks according to the size of an existing segmentation area; judging the locating capacity of the matching template according to the relationship between a minimal value of a gray difference between 9 image blocks and the segmentation areas and a set value.

Description

technical field [0001] The invention belongs to the technical field of computer vision and image matching, and in particular relates to a method for selecting a matching template in image matching and tracking. Background technique [0002] Image matching is a method of pixel-by-pixel comparison. According to the known pattern, the corresponding pattern is searched in another image, and a certain amount of data is used to measure the similarity between images, so as to obtain the best matching position. This process is called image matching. We call the images to be matched real-time images, and the known images are called matching templates. [0003] Traditional methods for selecting matching templates can be roughly divided into the following three types: The first method is to obtain the entire target area by segmentation and use the entire target area as a matching template. This method of selecting a matching template does not take into account the internal structure a...

AI Technical Summary

Problems solved by technology

[0003] Traditional methods for selecting matching templates can be roughly divided into the following three types: The first method is to obtain the entire target area by segmentation and use the entire target area as a matching template. This method of selecting a matching template does not take into account the internal structure and Texture information, when the complex background or the internal texture and structure information of the target is rich, it is difficult to obtain a complete target area by segmentation; the second is to use a local area with certain texture and structure information as a matching template. When it is a point target or a patchy target, the target lacks texture and structure information, which makes it difficult to extract a reasonable local area as a matching template by using the local area of ​​the target as a matching template.

The third is to obtain matching templates based on prior information. The method of obtaining matching templates based on prior information has poor adaptive ability, and there is a certain difference between the target information and the obtained prior information in the absence of target prior information or in real application scenarios. In the case of , it will cause the matching template based on the prior information to be unable to adaptively select the matching template according to the target information in the real-time scene

However, the shortcoming of this method is that the matching method based on feature points cannot adjust the size of the matching template accordingly when the target size changes.

Images