Design And Realization Of Three - Dimensional Inspection System For Flight State Ore

Granularity information is the main judging of ore sorting. In mining and metallurgy operations, ores are usually sorted for the optimal processing by granularity information, it can save energy and increase economic benefits in this way. Machine vision technology, with the characteristics of non-contact, high precision, flexible, and facile constructibility, has become a significant research direction in the automatic ore-sorting field. However, the current optical ore sorting technology is limited to the two-dimensional granularity information of ore, compared with the two-dimensional information, the 3D information of targets gains richer content and contains higher confidence degree as the basis of ore sorting. To solve the time-consuming issue in the whole 3D ore information detection utilizing the line-structured light and to meet the requirement of high-accuracy and high-speed measurement in actual processing operation, a novel system with line-structured light for fast and full-surface ore detection is designed. This system uses distributed multi-cameras to capture the ores in flight state, and processes the captured video subsequence in real-time, then reconstructs the whole 3D information of targets with series of algorithms. The main researches and implementations in this paper are as follows:(1) A scheme for 3D detection of flight-state ore in real-time is presented in this paper, in which we design and build up a four-camera optical system for 3D detection based on line-structured light.(2) Two-step calibration method based on radial alignment constraint is mainly studied, based on which we also combine the image processing toolbox of Mil to implement the system global calibration with four cameras, then the accuracy and stability of calibration results are analyzed.(3) To extract the 3D information of ore targets, the technologies of 3D target reconstruction, image mosaic, connected component analysis, laser defect compensation are studied, and their programming implementation are explored in this paper.By measuring the standard pieces of known specifications, the performance of the system has been evaluated in two aspects of accuracy and time. The experiment results indicate that the accuracy error of system is very small, and the processing time is quite short and satisfies the real-time requirement.