CenterNet-Based Target Detection Method for Remote Sensing Images

Abstract

With the wide application of remote sensing image detection in military and civil fields, fast and accurate identification of small and dense targets with low imaging quality in remote sensing images becomes the focus and difficulty of remote sensing target detection research based on deep learning techniques. A CenterNet-based target detection method combining attention mechanism is proposed, which utilizes ResNet-50 network for basic feature extraction; an improved channel attention module (ECA-NET) is introduced at the backbone output to weaken the expression of non-concern points while enhancing the information channel of concern points; and adjusts the learning strategy in stages to accelerate the model convergence. Experiments are conducted on remote sensing dataset, and the improved CenterNet algorithm increased 13 percentage points compared with the original, and the detection speed reached 52.61 frames per second. The experimental results show that CenterNet-based target detection method maximizes the remote sensing target representation capability of CenterNet under the condition of maintaining certain computational efficiency, effectively balances the accuracy and speed of remote sensing target detection, and is crucial in practical applications.