.png){kind=icon}
.png){kind=icon}
Return
Two practical systems for classification of three-dimensional objects
.png){kind=icon}
PRE
AI
DOI：10.1016/S0164-1212(99)00045-X.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The main goals of practical vision systems are the fast and accurate recovery of the three-dimensional description of an object that produced a two-dimensional image. In this paper, we propose two practical systems (called VISION1 and VISION2 for the classification of three-dimensional objects using a set of their two-dimensional images that adhere to these goals. VISION1 is an online recognition system utilizing a colored TV camera and the necessary interface equipment to supply a two-dimensional image to a digital computer for processing and classification. It essentially relies on shape-from-X techniques to compute the depth. In addition it exploits automatic rotation of the object to generate a training sample set, and extraction of 12 two-dimensional invariant features to train an artificial neural network (ANN) with. Then VISION1 recalls its trained ANN for classification of objects. The proposed VISION1 is simple, fast, uses less man-power and provides storage space at lower cost than comparable vision systems (Sahibsingh, A.D., Kenneth, J.B., Robert, B.M., Aircraft identification by moment invariants. IEEE Transactions on computers, vol, c-26 (1), pp. 39-45). On the other hand, VISION, system follows the classical approach of vision theory developed by Marr, D., 1982 (Vision - a Computational Investigation into the Human Presentation and Processing of Visual Information. Freeman, San Francisco, CA). VISION2 depends upon image segmentation algorithms, shape-from-X techniques, and the extraction of 11 three-dimensional invariant features to train an ANN with. Thus VISION2 is very efficient even more accurate than VISION1. Both VISION1 and VISION2 may be used in different manufacturing environments for classification of objects. Experimental testing was conducted through a sample test of 112 real images for four different classes of objects. The reported results support our claim that the two proposed systems are practical, fast and accurate. (C) 1999 Published by Elsevier Science B.V. All rights reserved.
Keywords:
IMAGE
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
4.1
Papers:
5.4K
Citations:
8.4K
Organization
No organization information available