A har-surf-ran feature point matching method for stereo vision 3D reconstruction

Abstract

The invention discloses a Har-SURF-RAN feature point matching method for stereoscopic three-dimensional reconstruction, which belongs to the field of digital image processing and comprises the following steps: Step (1): performing Gaussian difference preprocessing on pixels in the image, screening Candidate feature points are selected, and then filtered by weight coefficients to obtain feature points. Step (2): Process the feature point data set to obtain the feature descriptor of the feature point; Step (3): Use the nearest neighbor feature matching to match the detected feature points; Step (4): Match the feature point pair Preliminary elimination of misjudgment points, using the idea of ​​PROSAC to sort the data sets, improving the selection model and evaluation part of RANSAC, using the idea of ​​local optimization to improve RANSAC, and introducing the optimization parameter K to further optimize the model part. The invention reduces the amount of calculation, improves the speed, and the improved algorithm produces less redundant points and higher precision.

Description

technical field [0001] The invention belongs to the field of digital image processing, in particular to a Har-SURF-RAN feature point matching method for three-dimensional reconstruction of stereo vision. Background technique [0002] In recent years, computer hardware and software have kept pace with each other, and great progress has been made in terms of operating efficiency and information processing capabilities. People have also raised higher expectations for the information presented and processed by computers. As a result, 3D reconstruction technology has been continuously developed and improved. [0003] 3D reconstruction is a common scientific problem and core technology in the fields of computer-aided geometric design, computer graphics, computer animation, computer vision, medical image processing, scientific computing and virtual reality, and digital media creation. recover the three-dimensional structure. At present, it has been widely used in various fields s...

AI Technical Summary

Problems solved by technology

[0007] (1) Although the existing feature point detection algorithms are becoming mature, there are still some shortcomings such as many redundant points and weak anti-interference ability, and the accuracy and efficiency still need to be further improved

Among the common centralized feature point detection methods, Harris feature point detection method has strong stability but slow speed, SURF feature detection speed is fast but poor stability, both have certain limitations

[0008] (2) The 3D reconstruction based on stereo vision does not have high requirements on the scale, but the pyramid structure proposed by the SURF algorithm to ensure the invariance of the scale makes the algorithm slow and computationally intensive

[0009] (3) The efficiency of the existing mismatch elimination algorithm is related to the size of the subset, the proportion of intra-class points and the size of the data set, and the efficiency is low in some scenarios; in terms of accuracy, considering the efficiency factor, the obtained parameters are often Whether it is the optimal parameter, it cannot be directly applied, and needs to be refined

Images