Research Of Ship Recognition Algorithm In VTS Based On Machine Vision

With the rapid development of the shipping industry, Vessel Traffic Services(VTS) integrated the technology began to improve their own development, but there are many deficiencies, such as ship type of identification, radar blind spot monitoring and other issues still to be resolved.Currently, machine vision has been widely used in a variety of identification works which accuracy and reliability are highly important. If machine vision technology used in the VTS which will enable the ship recognition, will be able to solve many current problems. The advantages of improved machine vision systems for VTS include comprehensive concept and complete solution, and have high measuring accuracy, short processing time. For ship recognition, the machine vision system compared with other surveillance techniques, it should be one of the best choices. Machine vision system can detect and classify vessels in the port area, and help improve the efficiency of VTS; surveillance of water traffic area, improve the safety, and reduce marine traffic accidents in port and near-shore area, monitor accidental situation and to improve efficiency and operational vessel traffic flow in water sports.This paper's core goal is to approach the ship recognition via machine vision and image processing techniques. In our work, first do gray-scale processing and image pre-processing works on image after we acquire the ship image via video camera, aimed at enhancing the useful information of the image and remove the image noise,is smoothed using median filtering method to achieve the noise. Second, begin to locate the ship position in the image, and extract the ship segmentation from the image, the background is relatively fixed, after background removal, we can highlight the ship outline in the image, and then use the binary image processing, the ship area can be easily spitted. Then, extract feature from the ship area by using the edge detection and morphological erosion and dilation operations. Finally, we start to identify the ship, this paper presents a recognition algorithm combined the template matching with neural network model, first matching the feature extraction against various types of ship models in ship model ship library, select the high similarity model, then use BP neural network for final identification.