Gating image-based target detection method

A target detection method based on a gating image comprises the following steps: 1) acquiring the gating image in a dark light environment through a gating camera, and enhancing the image through histogram equalization; 2) inputting an image into the improved Yolov5 network, and detecting and identifying a target; the improved Yolov5 network is characterized in that a Focus layer of the Yolov5 network is replaced by a multi-scale layer; adding a cavity space convolution pooling pyramid on a PAN layer in a Neck layer of the Yolov5 network to realize multi-scale feature extraction, and adding a CBAM attention module in front of a detection head; the image processing method comprises the following steps of: 2.1) inputting a gated image into a Backbone to carry out feature extraction; 2.2) after feature extraction of Backbone, the extracted features are input into a Neck network for multi-scale fusion; and 2.3) finally, inputting the three features with different sizes into a Head network to generate a prediction