Ulva polifera Detection Method for High Resolution Remote Sensing Images Based on Dual-path Convolutional Neural Networks

Objectives: The green tide formed by Ulva prolifera (U. prolifera) is a harmful marine ecological disaster. The rapid and accurate detection of U. prolifera is of great significance for timely management of U. prolifera and the healthy development of the marine industry. The U. prolifera detection from remote sensing satellite images has become an important technical means for monitoring U. prolifera. With respect to the problem that the boundary of the U. prolifera area is difficult to be determined accurately in high spatial resolution remote sensing images (HSRIs), an U. prolifera detection method for HSRIs based on dual-path convolutional neural networks (CNN) is proposed in this paper. Method: First, a dual-path CNN semantic segmentation framework is designed based on the distribution characteristics of U. prolifera in HSRIs. The area and boundary of the U. prolifera in HSRIs can be extracted simultaneously using the proposed framework. Second, the strategy for optimizing the initial U. prolifera area detection results based on the U. prolifera boundary is proposed to improve the detection accuracy of U. prolifera area. Results: The experimental results show that the proposed method can extract the U. prolifera accurately, with F1- score of 88.25%, IOU of 78.97% and OA of 98.99%, which is better than other U. prolifera detection algorithms. Conclusions: The proposed method can obtain good results for the detection of different types of U. prolifera in HSRIs.