An Improved SSD Network-Based Target Detection Method for Penguin Populations

Global warming is seriously threatening the natural environment of the Antarctic region, and penguins, as the iconic Antarctic marine life, are very responsive to climate change. Therefore, the changing pattern of the penguin population can provide a scientific basis for predicting the response of the Antarctic ecosystem to climate change. In this paper, we propose an improved penguin population target detection algorithm based on the SSD (Single Shot MultiBox Detector) network to address the practical problems of the harsh Antarctic climate. The algorithm adds the ECANet attention module to enhance the detection capability of the model. Experiments show that the detection capability is improved by 13.5%, and the ap value is improved by 26% compared with the traditional SSD algorithm.