Research on Parallel Detection Technology of Remote Sensing Object Based on Deep Learning

Abstract

In recent years, methods based on deep learning have become a hot spot and trend in the field of remote sensing object detection, and a series of encouraging results have been achieved. With the development of science and technology, people's technology and ability to acquire remote sensing data have been comprehensively improved, and high-resolution large-scale remote sensing image data has increased dramatically. However, the current mainstream object detection models cannot directly input large-scale high-resolution images for prediction. This paper proposes a remote sensing object parallel detection algorithm, which uses the mpi4py module to realize multi-CPU+GPU distributed parallel processing. Based on the yolov4 object detection model, it can detect remote sensing images of any scale without reducing the prediction accuracy. And shorten the object detection time according to the number of distributed nodes. The experimental results show that the parallel algorithm has a high speedup ratio, and the parallel detection technology has a good development prospect in the field of remote sensing image object detection.