Blink Detection Algorithm Based on BP Neural Network.

Blinking is a very important behavior feature for fatigue driving detection. It is of great significance to detect the blink times regularly for reducing traffic accidents. Therefore, this paper proposes a blink detection algorithm based on BP neural network algorithm. The algorithm collects the driver image data by raspberry pie camera, uses AdaBoost detection algorithm with Haar feature to locate the position of face and eye, and extracts the information of human eye. The extracted eye information is used as the input neuron of BP neural network, and LM algorithm is used to train the neural network model to achieve the function of blinking detection. The experimental results show that the algorithm can recognize blink better, the accuracy rate can reach 98.34%, and the speed is about 20 times of the eye feature point aspect ratio (EAR) algorithm.