Abstract:This paper addresses the problem of recognizing threedimensional(3D) moving object from multiple views. It is based on the 2D processed frames of a video sequence which are clustered into view categories called feature aspects of the object and their transitions. Logpolar mapping(LMP) and Discrete Fourier Transformation(DFT) are used for getting the position, scale and rotation invariant feature vectors of 2D characteristic views. ART-2 model is used as memory and classifier of the feature information of the object. Improved ART-2 neural network is used in experiment, and the results are satisfactory.
1 Carpenter G A, Grossberg S. A RT-2:Self-Organization of Stable Category Recognition Codes for Analog Input Patterns, Applied Optics,1987, 26(25):4919-4930 2 Chin R, Dyer C. Model-Based Recognition in Robot Vision. Computer Surv,1986,18:67-108 3 Lowe D G.3DObject Recognition from 2D Image, Artificial Intel ligence,1987, 31 4 Seibert M, Waxman A M. Adaptive3 D Object Recognition from Multiple Views. IEEE Trans, Pattern Analys is and Machine Intell igence, 1992, 14(7):107-124 5 Liu C H, Tsai W H. 3D Curved Object Recognition from Multiple 2D Camera Views.CVGIP,1990,50:177-187 6 Steven F Kimball. Aircraft Silhouette Identification Usinga Neural Network Classifier with Scale and Rotati on Invariant Features,Master Th esis. University of NHUSA,1986 7 Mundy J L,ZissermanA.Geometric Invariance in Computer Vision.the MIT press, 1992 8 Peter J Van Otterloo. A Contour-oriented Approach to Shape Analysis.Prentice Hall, 1991 9 Messner R A.An Image Processing Architecture for Real Time Generation of Scale and Rotation Invariant Patterns.1985, CVGIP(3):50-60