基于人工神经网络的立体视觉系统的研究

立体视觉系统已广泛应用于三维形体检测等诸多方面，针对该系统物点到像点非线性关系的确定技术，提出了在于人工神经网络的标定、测量方法，同时提出了加快网络训练速度、提高网络训练精度的新方法，并给出了学习和测量结果，验证了此方法的正确性。这种方法降低了对系统本身精度的要求，且不同考虑透镜畸变的影响。是对立体视觉系统数学建模的有益探索，具有一定的理论意义和现实意义。

A 3-D STEREO VISION SYSTEM USING AN ARTIFICIAL NEURAL NETWORK

D stereo vision is widely used in various fields,especially in shape inspections of workpiece.3 D noncontact measurement method is introduced by using an artificial neural network.A new way to raise the training precision of BP neural network has been proposed .Some experiments have been completed and the results proved the feasibility of the method.Finally,this method reduces the requirement for system accuracy,and needs not to consider the effect of lens distortion.It is a beneficial exploration to stereo vision system and of theoretical and practical significance.