| The ship is an important monitoring target at sea.The remote sensing satellite technology to interpret the information of the ship target has a wide application prospect in both Marine surveillance and military reconnaissance fields.With the change of application requirements,attention should be paid not only to the static position of the ship target,but also to the state and movement trend information of the target.As an important characteristic of a moving ship,wake can be used to estimate the information of the ship’s motion parameters.It is of great significance to carry out the research of ship’s wake detection and inversion of ship’s motion information.In this paper,a ship wake region extraction method based on high-resolution optical remote sensing image is studied,and the ship motion parameters are estimated by making full use of the detailed information extracted from the wake region.The preliminary simulation experiment verifies the feasibility of the method,and the research results lay a theoretical foundation for the reconnaissance and surveillance of Marine ship targets.The main work of this paper is as follows.In order to meet the requirements of ship wake detection and ship motion parameter estimation from remote sensing images,the types and causes of ship wakes were studied,the various characteristics of ship wakes were analyzed,and a high resolution optical remote sensing ship wakes data set was constructed and annotated.In order to ensure the validity and availability of the remote sensing ship wake data set,image screening and labeling are carried out from different interference conditions,different wake intensifies and different wake directions,so as to make it more scientifically and reasonably conform to the actual ship wake.Thus the performance evaluation of the subsequent wake detection algorithm is more objective and effective.To solve the problem that conventional wakes detection methods are difficult to extract ship wakes effectively,a ship wakes detection method based on Gabor feature and SVM was proposed.Based on the difference of texture characteristics between ship wakes and sea clutter,a classifier based on SVM is constructed.A usable classifier can be trained by using small piece of data.The experimental results show that the detection performance is better compared with the conventional wake detection methods.An improved Mask R-CNN ship wake detection method was proposed to solve the problem that it was difficult to detect optical ship wakes due to the interference of sea clutter and clouds.Different from the traditional wake detection algorithm,the Mask R-CNN model is a multi-target multi-task model.Based on the detection of turbulence wakes and Kelvin wakes,the ship target detection is added and the example targets are segmenting.In the experimental process,the influence of different depth of backbone network on the experimental results is explored,and the experimental results are compared with the traditional ship wake algorithm.The experimental results show that the deeper backbone network can effectively improve the performance index,and the Mask R-CNN model shows good performance in the segmentation and extraction of ship targets,turbulent wake targets and Kelvin wake targets.Compared with the traditional method,the Mask R-CNN model has great performance advantages and shows better anti-jamming ability in various complex situations.To solve the problem of how to accurately invert ship motion information from a single static image,a method to estimate ship motion parameters based on ship wake characteristics is studied.By making full use of the information extracted from the ship wake region,the course parameters of the ship are estimated by the turbulence wakes and the ship speed parameters are estimated by the shear wave wavelength of the Kelvin wakes.An end-to-end high resolution optical remote sensing image ship motion information interpretation scheme is provided.The preliminary experimental results prove the feasibility of the proposed method. |
PDF Full Text Request