Stereo matching method based on neighborhood information constraint and adaptive window

Abstract

The invention discloses a stereo matching method based on neighborhood information constraint and an adaptive window, which relates to the field of machine vision, and specifically comprises the following steps: inputting an image, performing adaptive weight cost calculation and cost fusion, performing adaptive window cost aggregation, performing parallax calculation and optimization, and obtaining a final parallax image. Aiming at the problem that the traditional Census algorithm has high dependence on a central pixel, a weighted average sum mode of neighborhood windows is adopted to assign values to the central pixel. Then, through setting an adaptive threshold value, similarity secondary cost calculation is performed on neighborhood pixels and central pixels of the support window, the neighborhood pixels and the central pixels are combined with the initial cost, and further constraining is performed on a matching result; in the cost aggregation stage, a four-constraint method with acontinuously changing color threshold is adopted to construct a window, and gradient information is combined to further improve the matching precision; and in the parallax optimization stage, the parallax graph is further optimized by adopting a left-right consistency detection and WTA combined method. According to the method, the precision of three-dimensional reconstruction is effectively improved.

Description

technical field [0001] The invention relates to the field of machine vision, in particular to a stereo matching method based on neighborhood information constraints and adaptive windows. Background technique [0002] Stereo matching (Stereo Matching) is one of the research difficulties and hot spots in the field of computer vision. It is a technology to obtain the disparity value from the two-dimensional plane image pair and then restore the depth information. The accuracy of stereo matching directly affects the depth calculation. The accuracy and effect of 3D reconstruction of the scene. Its uses involve 3D environment perception and modeling, robot navigation, driverless cars, object tracking and detection, etc. [0003] According to different optimization theory methods, there are mainly two types of matching algorithms, local and global. The global algorithm obtains the matching result by minimizing the energy function, which has high precision and low efficiency. The...

AI Technical Summary

Problems solved by technology

[0005] In view of the deficiencies in the prior art, the present invention provides a stereo matching method based on neighborhood information constraints and self-adaptive windows, by improving the traditional Census algorithm in the matching cost calculation stage, aiming at the dependence of the traditional Census algorithm on the central pixel For high-level problems, the self-adaptive weighting template is used to re-assign values ​​to the central pixels, and the cost fusion model is established according to the smooth area and the non-smooth area to obtain the matching cost value; in the cost aggregation stage, a four-constraint adaptive window with constantly changing color thresholds is used to construct method to further improve the matching accuracy; use the WTA algorithm to calculate the optimal disparity; use the left-right consistency principle to optimize the disparity

Images