Feature extraction based on learning for feature list object matching

The appropriate choice of feature extraction offers possibilities for reducing calculation complexity in machine vision applications, which also has a strong influence on the results of the feature list object matching. But the requirements for reasonable feature extraction are sophisticated and depend on different applications. Based on machine learning, an approach to gradient feature extraction using double thresholds is provided for feature list object matching in this paper. By training, the double thresholds adapted to the special application can be automatically estimated, where an unsupervised learning means is used. Then, the estimated double thresholds are used to the extraction of gradient feature points for the features list matching. The proposed method has been verified by the experiments.