Tire imprinting character dynamic identification method based on collaborative network

The invention provides a tire imprint character dynamic identification method based on a collaborative network, which comprises the following steps: firstly, identifying a tire in a captured image by using a YOLO-V5s network, and cutting the image according to the contour of a detection frame; carrying out four partitions on the cut tire image, positioning character regions in the partitions by using ENet, and filtering out noise regions with too small areas; extracting a complete character region based on a region growing algorithm; making a circumscribed rectangle for the irregular mask area, and splicing the four segmented images; calculating the areas of the spliced character regions, sorting the spliced character regions, and filtering non-target character regions; making a circumscribed rectangle again, rotating the image and expanding the projection to make the character in a normal position, and straightening the character; and identifying the characters in the image blocks by using a YOLO-V5s network