Application And Research Of Seat Belt Detection In Bayonet Surveillance

With the rapid development of society,all major traffic sections in the country are equipped with bayonet surveillance video to monitor whether the drivers regulate driving,such as monitoring whether drivers wear seat belts.As for the huge bayonet surveillance video data,using manual monitoring to determine whether the driver wears a seat belt will consume huge human and financial resources.This paper studies ways to detect seat belts based on computer vision which study seat belt detection in bayonet surveillance from two aspects:Firstly,this paper propose a special feature extraction method to construct a salient gradient map to detect seat belts.The traditional seat belt detection utilizes Hough Transform to extract the edge straight line of seat belts to detect seat belts.The paper proposes gradient mutation straight line detection to extract the edge of seat belt.The method used in this paper can exclude some interference such as the edge of the clothes collar,stripe and behavior of placing seat belt on chest to pretend wearing seat belt.This paper compares seat belt detection based on Hough Transform with the proposed method in experiment,and result shows that the proposed method is superior to the traditional Hough transform method in seat belt detection.Secondly,this paper proposes integral projection detection based on Fully Convolutional Network which has excellent application in the field of image recognition to detect seat belt.Compared with the traditional machine learning,the biggest highlight of deep learning is that it can automatically learn features from a large number of data samples,minimizing the complexity of manual design features and human intervention.The paper studies the Fully Convolution Network in deep learning.Using Fully Convolution Network model trains the sample picture set to get the model used to detect the seat belt.After that,the model is used to detect the seat belt.Then,the paper uses the integral projection method to find the seat belt end point and uses the slope of the line segment of the two end points to judge the seat belt.Finally,compared with other methods,the algorithm has good ability to detect the seat belt.