Worm, Address Resolution Protocol (ARP) broadcast and other abnormal behaviorS which attack the cloud computing platform from the virtual machines cannot be detected by traditional network security components. In order to solve the problem, abnormal behavior detection technology architecture for cloud computing platform was designed, abnormal behavior detection for worms which brought signature and non-signature behaviors based on mutation theory and "Detection-Isolation-Cure-Restore" intelligent processing for cloud security was proposed. Abnormal detection, management of event and defense, and ARP broadcast detection for cloud computing platform were merged in the system. The experimental results show that the abnormal behavior inside the cloud computing platform can be detected and defensed with the system, the collection and analysis of the abnormal behavior inside cloud computing platform can be provided by this system in real-time, the traffic information can be refreshed automatically every 5 seconds, the system throughput can reach to 640 Gb and the bandwith occupied by abnormal flow can be reduced to less than 5% of the total bandwith in protected link.

Equipment failure often brings some defects on the surface of batteries in the battery production process. The traditional artificial detection has weakness on the timeliness and durability. But there has not been any efficient automatic detection means for the ordinary battery by now. Concerning the distribution and morphological characteristics of the defects, a new automatic optical detection method based on computer vision was proposed. The proposed method used Canny operator and virtual granule collision method with the minimum value searching method to determine the area to be detected based on the battery anode surface morphology features. Considering the sharpness of the defect, Harris corner points were used to mark the defects as mark points. False mark points were filtered by the degree of aggregation of the points. The defect region would be extracted at last according to the location of mark points. The experimental results illustrate the detection success rate of the proposed method is over 90% and the method can work more efficiently than the popular wavelet analytical method. The study achievement provides a reference for product quality automatic detection on battery production.