There are three common shape feature extraction methods based on machine vision including threshold processing method, coutour-based extraction method and region-based extraction method. Threshold processing method is one common method of shape segmentation and extraction and is easy to manipulate with high speed, but it is not suitable for complicated target shape features with high accuracy; Coutour-based extraction method could process with high speed but is more likely to bring about bias or error while dealing with complicated target shape features; Region-based extraction method requires enormous internal memory though it could be easily and accurately implemented to extract shape features. However, the limitation in extraction methods in application is still obvious and it is of significance in developing new extraction method.