An Object Detection Algorithm Based On Grid Matching Of Feature Points

In the field of computer vision,image feature point matching has been widely used in many fields due to its advantages such as cheap,efficient and effective,such as medical image diagnosis,remote sensing image mosaic,robot vision navigation,3D model reconstruction,object detection and tracking.And workpiece inspection.In real life,the images we obtain from natural conditions generally have some defects,such as uneven light in the picture caused by dark lighting,messy picture texture caused by complex scenes,and image sharpness caused by device pixel restrictions.Insufficient,flexible deformation or blurring problems in picture content,etc.,these problems have greatly limited the feature point matching algorithm of the image to distinguish true and false matches,making the problem of image feature point matching difficult.This paper conducts research on feature point matching algorithms.Based on the principle that adjacent pixels share similar motion,a mesh-based scoring model is proposed to distinguish true and false matches,which can maximize the feature point matching.Precision and accuracy.The work of this article is as follows:1)Analyze the principle and idea of the feature point matching algorithm,find out the deficiencies,and think based on the deficiencies,so that the subsequent algorithms can be improved.2)The smoothness of the movement is converted into statistical problems to discriminate false matches,which can distinguish true and false match pairs more accurately and faster.The grid matching matching is used to eliminate false matches and improve the accuracy of matching.3)From the perspective of probability theory,the number of pairs of matching points in the grid is regarded as a successful match.4)A complete target detection algorithm based on grid matching of feature points is proposed.First,the image is gray-scaled,and then a rough match is performed.Then use the grid matching model to perform fine matching on the pictures,and finally use the ransac homography transformation to obtain the position of the target in the template picture.Finally,this paper carried out experimental verification based on the designed algorithm,and compared the feature point matching algorithms with the same function,and compared and analyzed from the three aspects of accuracy,recall,and balance function combining precision and recall..It is confirmed that the feature point matching algorithm proposed in this paper can achieve effective,robust,fast and accurate matching.