Unsupervised Shape Classification of Convexly Touching Coated Parts with Different Geometries

Abstract

With the development of Rapid manufacturing techniques, a heterogeneous object can be manufactured by processing geometric and material information in each layer while slicing the object. The functionality and surface finish of heterogeneous object depends upon precise slicing and accurate information processing i.e. geometric and material during fabrication. The current work focuses on both the issues. Layer thickness is varied using adaptive thickness method. Horizontal and vertical contours are approximated to reduce the error and minimize the effect of geometry and material stair-step effects. Reference based boundary mesh generation approach is used for defining material composition at each point in the slice domain. An algorithm is proposed for scanning layers to retrieve geometric and material information at each point in the object domain. A database system is developed for faster processing of information and to avoid the data redundancy during operation. Reference based approach, proposed algorithm and adaptive thickness method are implemented and some cases are discussed to show the validity of the work.

An Economical Design of Automatic Rice Grading using Image Processing

“…Researcher Hobson et al [15] used unsupervised clustering method for the classification using various features. Rajalingappaa et al [16] utilized morphological operations to identify touching components. For the separation of touching grains, Fourier analysis [17] and the method on convexity of boundary of two touching grains is used but in this method there is the limitation of amount of grains.…”