Research on feature extraction and matching algorithm based on speckle structured light binocular vision system

Binocular stereo vision combined with structured light technology has developed rapidly in recent years. We have studied the blob extraction and matching algorithm of randomly distributed infrared speckle structured light, proposed an improved blob detection algorithm based on Laplacian of Gaussian (LoG) operator, which combines the contrast of the blob, the intensity difference and statistics between the center and edge pixels of the blob. The improved blob detection algorithm can extract the blobs of various scales and brightness effectively at one time, and restrain the influence of edges, bright strips and corners. According to the randomness of projected dot distribution, a matching method based on the characteristics of surrounding blob distribution is proposed. In order to solve the correspondence problem, we attach a descriptor to each blob. The descriptor at a reference blob center captures the angle and distance of the remaining blobs relative to it. In a small local region, the corresponding point sets in the two images are approximately affine transformed, and the transformed point sets have similar descriptive values. Experiments show the feasibility and correctness of this improved blob detection algorithm and matching method.