Video Face Recognition of Virtual Currency Trading System Based on Deep Learning Algorithms

Virtual currency trading develops rapidly, taking up a large proportion in the whole economy, and has become an important part of people’s daily life. Firstly, a reliable virtual currency trading system model is proposed, which is composed of four participant platforms, providers and payers, mainly including the purchase of virtual currency, purchase of virtual goods or services, exchange of virtual currency and other trading activities. In order to monitor possible collusion fraud in virtual currency transactions, payers are introduced to prevent collusion attacks. Video face detection algorithm based on multi-task Cascaded Convolutional Networks is designed and implemented to solve the problems of pose, light and occlusion in the virtual currency trading system. The algorithm utilizes the inherent correlation between detection and calibration to improve the detection performance under the framework of deep cascading tasks. Furthermore, the paper utilizes the three-tiered architecture combined with the well-designed VNN algorithm to realize the face detection and the rough location of key points. Meanwhile, on the basis of in-depth study and analysis of traditional face detection technology, a face detection algorithm based on Harr features and AdaBoost is implemented for comparison and analysis with the aforementioned algorithm. The experimental results show that the method based on MTCNN deep learning network can better extract effective features about faces in videos and achieve more accurate detection. Compared with the traditional method based on AdaBoost and cascade structure, the positive detection rate is much improved.