WITHDRAWN: Simulation of physical education teaching video recognition based on FPGA and Sobel algorithm

Abstract

Abstract Bullying in school is a serious problem among teenagers. School violence is the type of bullying in the school, which is considered the most harmful. In developing (artificial intelligence) technology, a new method for detecting school violence is present. It proposed a video-based detection algorithm for the power on the campus. Foreground (K Nearest Neighbor Method) for moving the target through the first KNN detected a pre-process detection target by morphological processing method. The circumscribed rectangular frame integration method has been proposed to optimize the moving target's circumscribed rectangular frame. The function's characteristics and the rectangular frame's optical flow are extracted to explain the difference between campus violence and daily living activities. It was using the relief -F and packing algorithm to reduce the feature size. It is applied as an SVM (Support Vector Machine) classification five times to do the cross-validation. Furthermore, to improve recognition performance, DT-SVM (Decision Tree SVM) has developed the two layers' classifier. Typical physical violence and box plot to determine the activities and daily life. The regular activities' specific features can be distinguished from the layer of physical violence of everyday life. For the remaining part of the action, the SVM layer, do the classification. For this DT-SVM classifier, it reached a better accuracy rate, and accuracy is thus shown a significant improvement is achieved. Measurement and high spatial frequency region of the two-dimensional spatial gradient performed by the Sobel operator highlights that correspond to the edge image typically used to find the magnitude of the absolute approximation slope at each point of the input gradation image.