A Study of 3D CAD Model and Feature Analysis for Casting Object

Abstract

When dealing with computer vision inspection testing parts in production line, the appearance of noise such as dust and inconsistent light distribution should be consider for further analysis on the parts image. In this paper, shape representation model using feature vector and Fourier descriptor were presented on the 3D CAD model image with the aim to gain the shape feature analysis for casting object. By adding light and salt & pepper noise on the CAD model image, the predicted database was compared to its original CAD image. In feature vector method, calculation on its Similarity, Correlation, Matching black and white points was carried out. Results observation show similarity of feature vector method performs 68% accuracy for light noise appearance, while correlation method performs 98% accuracy on disturbance of salt & pepper noise. Fourier Descriptor used to present the pose estimation of images on CCW and CW direction. Result shows matching sets similarity is value high since the dissimilarity value keeps below 0.3 and achieve few similar points in certain position. Thus, it is sufficient for casting object by implementing feature vector method which were very useful in analyze the noise on the image while pose estimation position described by Fourier Descriptor function.