Research on Bidirectional Matching Algorithm of Variable Threshold SIFT Based on DBSCAN

Abstract

The safe distance measurement of stacking in hazardous chemicals warehouses is an important factor to ensure the safety of hazardous chemicals storage. Binocular vision distance measurement is a very promising technology. Aiming at the key problems in that feature points matching in distance measurement is easily affected by illumination and feature similarity, and based on the degree of density, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm is proposed to realize Scale-invariant feature transform (SIFT) matching algorithm of variable thresholds in different regions of the image, and then adaptive neighborhood and bidirectional matching algorithms are used to further eliminate the mismatch points. The experimental results show that the matching rate of binocular distance measurement in hazardous chemicals warehouse is 68% and the accuracy of feature points is 98%.