一种基于神经网络特征识别方法的研究

Research on Automatic Feature Recognition Using Neural-Network-Based Techniques

提出一种新型基于人工种经网络的特征识别方法，可大大加快特征识别的速度，降低交叉特征识别的难度。同时，提出的LAG图克服了FAG因特征信息不完备的缺陷，使网络的输入形式简洁统一，改善了现有用神经网络识别特征时由于输入的不确定性和复杂性而导致效率不高、特征识别域小和使用困难的不足。此方法通用性较强，计算量较小。

A new technique for performing feature recognition using neural network is proposed. The feature description is the kernel in feature recogniti0n, which,in this case,is represented firstly with loop adjacency graph (LAG) and loop adjacency vector, which contain more useful in formation for recognizing features than traditional face adjacency graph (FAG) and are suitable for the parallel input of data in neural network. The algorithm is described and, for illustruction, sevelal parts including some basic and interactive generic features are computed which highlight the strengths of the recognition method in the end.