A Moving Track Identification System Based On Neural Networks

The intelligent identification technology has been widely using in surveillance systems. The existing surveillance systems have been able to automatically identify some people's activities and alarm for risk activities, but a few systems can specifically deal with identification and detection from online videos. Performance of the systems has not yet reached the practical application requirements. Since artificial neural network is the most popular research direction in the intelligent recognition technology, the intelligent surveillance systems using neural network technology will be a hot research topic in future.This paper analyzes development and application prospect in intelligent surveillance systems, and proposes a implementation scheme of a moving track identification system based on neural networks, in which an algorithm is proposed for moving object detection and tracking using the spiking neural network based on the principle of axonal delay and image processing. An image processing platform is designed to capture the moving objectcentroid track. A new algorithm for fast feature extraction of centroidal tracks and neural networks are used to realize the design of the moving track identification system. Through capturing the track of moving objects the system can judge whether there is potential threat. The advantages are that neural network algorithms are integrated to the traditional image recognition approaches to improve recognition accuracy and achieve real-time processing for requirements of remote network real-time surveillance.The experiments showed that the system could quickly obtain the correct result for the identification of the moving objects. The system can be implemented in the ARM development platform as a network terminal product, which has important practical significance to many important places such as video surveillance and intelligent transportation surveillance.