Automatic parking space detection system based on improved YOLO algorithm

With the rapid increase in the number of cars in society, the difficulty of parking has become one of the problems in the process of urban development. To solve this problem, a parking space detection system based on improved YOLO algorithm is proposed. The data sets in different occasions, different time periods and different lighting environments are obtained through three methods: network acquisition, camera shooting, and three-dimensional modeling. The original network was improved by six different measures including changing the YOLOv3 network input resolution, increasing momentum, increasing the weight attenuation value, increasing the batch size, reducing jitter, and choosing the appropriate ignore_thresh value. And further proposed the Attention-YOLO algorithm, the algorithm draws on the item-based attention mechanism, adds channel attention and spatial attention mechanism to the feature extraction network, and replaces the original feature vector with the selected feature vector. Perform residual fusion and add a second-order term to reduce the information loss in the fusion process and accelerate the model convergence. Experiments on the COCO and PASCALVOC datasets show that the algorithm effectively reduces the positioning error of the bounding box and improves the detection accuracy. The experimental results show that the new YOLO-
$288\times 288$
network is better than the original network. It has higher detection accuracy, recall and reliability, and can meet the parking space detection requirements of parking lots.