Research on infrared small target detection based on improved CenterNet
With the continuous development of machine learning technology,the research on object detection technolo-gy is becoming increasingly popular.To address the issues of low accuracy and poor real-time performance in target detection,a single stage object detection algorithm CenterNet is adopted to achieve rapid recognition of targets.A CBAM attention mechanism is added to resnet50,the backbone network of the algorithm,to improve the recognition accuracy of the network on the target.In the output module of the network,a new GSConv convolution module is used to improve the detection speed without loss of accuracy.The improved algorithm is validated on the infrared dataset-and its detection accuracy reaches 82.91％.The results show that that the improved CenterNet algorithm can accu-rately and efficiently accomplish the recognition of small infrared targets.
万方数据
维普
