Three-dimensional Multi-sensor Detection Technology For Highly Reflective Surfaces

Three-dimensional optical detection have been widely used in industrial inspection, agriculture,aerospace, meteorology and other fields for its non-contact, high efficiency, high precision, etc. Withthe development of pattern recognition, image processing technology, the application areas ofthree-dimensional optical detection technology are expanding. In the process of industrial inspection,we need to develop different types of three-dimensional optical detection system to meet differenttesting requirements. Based on three-dimensional optical detection principle, we carried out a detailedanalysis and solution about some of key issues in multi-sensor dimensional detection system. Thisthesis mainly includes:(1) We design a multi-sensor detection system for the limitations of single vision sensordetection. The system is characterized by multiple cameras for multiple testing, and it is agood solution for the problems and inefficiencies of single vision sensor.(2) According to the positional relationship between each vision sensor in the three-dimensionalmulti-sensor detection system, we proposed a global calibration method by unifyingcoordinate systems. By converting coordinate systems, we unify dimensional inspection datainto a single coordinate system. At last, we optimize the calibration results.(3) Focusing on the interference problem of highlight that frequently happens on highlyreflective metal surfaces in the process of3D laser detection, a peak continuity method isproposed. Finally, we verify the proposed algorithm. Experiments show that50%increasement of reconstruction accuracy is gained over the existing methods.(4) This paper presents a global pixel-level center extraction method and sub-pixel levelextraction method based on the direction of images. First, we extract light stripe fromvertical and horizontal directions; Then by calculating the precise direction information, wehave accurate extraction; Finally, we introduce the common characteristics and detectionmethods(5) We build a system experiment platform. Using VC++2008to achieve algorithm and haveexperimental verification.