Analysis And Processing Method Based On Point Cloud Data In Non-Contact Size Detection Of Multi-Feature Parts

With the rapid development of industrial products, structure of parts becomes diversified and complicated increasingly, for these multi-feature parts with many complex internal structure, it's difficult to achieve rapid and comprehensive detection by the conventional measuring instrument, therefore, the non-contact detection of digital technology has developed rapidly in recent years. But there are still some problems in the process of non-contact detection, this paper has researched on denoising and smoothing, data reduction, alignment of point cloud model based on the non-contact detection of digital technology deeply.Based on the smoothing and denoising method of point cloud model, determine the noise point for the point cloud model of multi-feature parts in this paper, after deleting the noise point, redefine the position of remaining points through weighted fuzzy c-means clustering algorithm, then smooth the clustering point cloud model by bilateral filtering. The experimental result shows that the method of this paper can remove the noise points effectively, much better than using bilateral filtering only.After smoothing and denoising the piont cloud, reduct the piont cloud in this paper. First redefine a weighted feature parameters, then use octree principle to estimate density of point cloud model and set threshold value, through comparing the characteristic value with threshold value to determine the feature points. Finally, after computing and comparing all the data points, remain feature points motionless, simplify the non feature points by improved curvature only. The experimental results show that the method of his paper can preserve characteristics of parts effectively on the basis of simplifing data.After simplifying the piont cloud, compare the point cloud model and design model in this paper. Analyze the reverse problem existing in the principal component analysis, and revise the problem. research on the recent iteration algorithm, find corresponding points quickly by k-d tree method and calculate the transformation matrix by singular value decomposition, make the two models compare. The experimental results show that the method of this paper raise the computational efficiency in the case of guaranteeing the alignment precision.Finally to verify, the corresponding deviation difference data report between two models canbe gotten, the detection method of this paper not only can be used as the size detection method of the new products before they leave the factory, but also can be used as a damage detection method of the parts, and provide reference for subsequent processing. Because of the strong practicability and good generality this detection method has certain application value in practical engineering.