Research On Ship Detection Technology Of Unmanned Surface Vehicle Based On Deep Learning

With the development of land resources becoming saturated,many countries begin to turn their attention to marine resources.China has a long coastline,many islands and abundant biomineral resources in the adjacent sea area.The surrounding countries have coveted China’s marine resources for a long time,for which China’s maritime security has been threatened.As an intelligent platform,the Unmanned Surface Vehicle(USV)can be used to perform maritime monitoring tasks.It can detect and feedback real-time illegal ships violating our territorial waters.In this paper,a vision system based on USV platform is constructed.Video images are collected and target detection algorithm based on depth learning is used to detect and recognize various types of ships.This paper first introduces the existing ship detection technology and analyzes the problems existing in the traditional target detection method.Aiming at the problems of traditional methods and the advantages of AI big data in learning characteristics,this paper studies the target detection algorithm based on deep learning to realize ship target detection.After that,two kinds of target detection methods based on deep learning are introduced,and the SSD algorithm model and its advantages and disadvantages are emphatically analyzed.On this basis,an improved method is studied,which replaces VGG16 in the basic network of the original algorithm with ResNet50,and adds additional convolution layer and pooling layer,thus improving the detection accuracy and detection effect of the algorithm,and improving the robustness of the algorithm.Then design experiments to compare the improved algorithm with the original one.The results show that the improved SSD algorithm has improved the detection rate and detection effect compared with the original SSD algorithm.In addition,four kinds of ship images are collected to create ship data sets,and experiments are designed to test the performance of improved SSD algorithm in ship data sets.The data enhancement method is used in the ship data set,which effectively increases the number of samples.The transfer learning method is used to train the improved model,which improves the convergence speed of the model and reduces the training time.The experimental results show that the improved algorithm achieves the detection and recognition of four types of ships,and has greatly improved the detection effect and accuracy.Finally,the vision system of USV and the ship detection system based on improved SSD algorithm are constructed to realize the detection and recognition of ship targets on the river.