Study On Fuzzy Cascade Classifier And ORB Algorithm For Detecting Covered Face Before Bank’s ATM Machine

Detecting covered face for ATM machine is one of the difficult problems of the intelligent video surveillance. Due to the reduction of the covered face’s features and the complex environment, the study on the algorithm for detecting covered face is imperative, which has higher precision and better real-time effect.Firstly, several algorithms are compared in some aspects, such as detection speed. The algorithm for detecting covered face is proposed, which is based on fuzzy cascade classifier and ORB algorithm. The algorithm of light compensation is applied for light balance in the whole image because of the light and shadow’s interruption in the complex environment. The Gaussian pyramid is used for reducing resolution, which can remove the special characteristics of the same obstructions. The preprocessing is excellent.Secondly, the black and dark obstructions are selected for fuzzy cascade classifier to construct the sets of samples. On the one hand, the fuzzy cascade classifier with higher precision is applied for the common detection. On other hand, other types of obstructions are dealt with the ORB algorithm. The algorithm’s effect is evaluated by comparison with other common algorithms in the aspects of detection rate, error rate, loss rate and detection speed. The results show that the algorithm can fit in the complex environment and the problems for detecting covered face can be solved. Not only its precision is better than that of other algorithms, but also it has a better real-time effect.At last, the intelligent video surveillance system is presented with good interface, consisting of initialized module, waiting-operating module, detecting module, warning module and data updating module. The results of videos’test show that the system can detect the covered face in the complex environment. It lays the key foundation for the real application and profound development.