Research On Ship Target Detection And Tracking At Sea Based On Deep Learning

In order to ensure the safety of ships sailing at sea,and at the same time to better manage the sea,the technology of marine target detection and tracking plays an important role.Traditional target detection and tracking technology can’t meet the current actual needs.Therefore,this paper studies the target detection and tracking technology based on deep learning with higher accuracy and faster real-time,and on this basis,improves and optimizes it to achieve better working effect.Firstly,by means of field collection and online collection,using LabelImg tool,the marine ship detection data set is made.The target objects include passenger ships,cargo ships,speedboats,warships,sailboats,tugboats,kayaks,fighters,buoys and other ships in ten categories.Secondly,considering the actual foggy weather at sea,image defogging is used to make the detection algorithm better improve the detection accuracy under foggy sea conditions.In this paper,the defogging algorithm based on PMSNet is used for image processing.This algorithm can better compensate for the image color distortion caused by traditional defogging processing,and at the same time can adapt to more environmental conditions.By combining with YOLO v4 algorithm,the detection accuracy of the detection algorithm model in foggy conditions can be significantly improved.Then,aiming at the poor detection of multi-targets,small targets and cross targets at sea by the original detection algorithm,the residual network is used to make the model learn new features based on the original features,and three residual modules are added in front of the Prediction structure of the model.Using multiple SPP modules to improve the detection sensitivity of the model to image targets;The introduction of ECANet module can effectively improve the detection performance on the basis of reducing the amount of training parameters.The experimental results show that the detection accuracy of multi-targets,small targets and cross targets at sea has been significantly improved by the improved and optimized detection algorithm,and the real-time detection can also meet the actual needs.Finally,according to the idea of Tracking-by-Detection,the target tracking of ships at sea is carried out.According to the results of the detector,Kalman filter algorithm is used to predict and update the state of the moving target in the video frame,the combination of the target motion and features is completed by the deep neural network and cascade matching method,and the tracking and detection information of the target is combined by Hungary matching algorithm.The experimental results show that the tracking algorithm can achieve good tracking effect on marine ship targets,and can meet the tracking requirements under various conditions.