Research On Detection And Identification Of Target Ships In Boundary River Waters

With the continuous development of new computer information technology and artificial intelligence,the power of science and technology can be seen everywhere,from daily life shopping to aviation,aerospace and military wars,all of which are supported by powerful information technology.In China’s most important territorial sovereignty,scientific and technological information is also essential.Based on the analysis and research of border monitoring and identification,this paper focuses on the boundary river area,and finds that the monitoring and identification technology in the boundary river area still has defects,and many places still need a lot of manpower and material resources for supervision.To solve this problem,this paper designs and improves the R-CNN algorithm and feature extraction method by using the ship images collected at the boundary river to complete the detection and recognition of ship targets.On the one hand,it can change the traditional border defense mode,liberate manpower and material resources,realize the real-time border information,and guarantee the border security and social stability.On the other hand,the target detection and tracking technology on the border can be applied to civil or military fields to promote the development of detection and tracking technology and the improvement of theoretical knowledge.Therefore,it is of great theoretical and practical significance to study the target detection and tracking algorithm in the boundary river area.The main work completed in this paper includes the following aspects:(1)The first analysis boundary river detection of deficiency,according to the interference of its large number of complex background and R-design and improvement of CNN algorithm,by mixing channel image segmentation are extracted,and then use selection algorithm and greedy strategy to make the candidate set generation,want to get the image,image preprocessing.(2)Before the experiment,this paper due to the particularity of Jie Jiang border region,is not suitable for training and testing data sets,through the network search and Jie Jiang forms a video capture,get the data you need atlas 1259 pieces,and then 1159 as the training set,and rotate operations,such as increase the number of training,the other 100 pieces as a test set,For testing and comparison in subsequent experiments.(3)This paper introduces a number of feature extraction methods and algorithms,and chooses to use geometric features and moment invariant features to extract features of ships,through the shape of the appearance of the features for recognition,and simulation experiments.The experimental results show that this feature can realize the detection and recognition of ships.(4)MATLAB was used to classify the test atlas and conduct experiments.Different number of test sets were selected,and the results were compared with the number of correct,error and missed detection of other common algorithms to verify the accuracy and robustness of the algorithm.