Pier Anticollision System Based On Image Processing

Based on studying and organizing the relative materials, this thesis mainly involves following four research contents: image processing, background distilling methods, detecting and following moving vessel.The paper devotes to studying a kind of safer pier anti-collision system which don't need enormous project and have reduced the cost of researching and developing.On the research of the establishment of real-time video grabbing system, we researched the constitution of hardware and software of video grabbing system, and two key problems of the system is solved. Asynchronous double buffer grabbing is used to solve the problem of real-time grabbing. After converting the format of processed image data, the data may be saved as AVI format files.On the research of image processing, we mainly carry on image sharpen, noise removal according to the purpose of the paper.For noise removal, we researched several methods and finally we used midfilter to remove the noise. Because midfilter is not only simple, the final effection is very well. For image sharpen, we used Sobel Gradient Operator.The test indicate that using Sobel Gradient Operator to processing image can make the outline of picture in focus.On the research of background, we researched several kinds of daily background distilling method at first, through the test for the study on several kinds of methods and analyzed the merits and demerits of them, the results of experiments indicate that the improved algorithm based on the Surendra do good work.On the research of the vessel detection under a static background, firstly we researched three main algorithms of motion detection- continuous frame difference, light stream and background difference and analyzed their advantages and drawbacks. Secondly, we used those methods to do experiments, and compared the effection of each one. The results show us background difference is fit for this system, especially the effections and speed.On the research of following moving vessel and pre-alarm ,Firstly we researched several main object features and matching algorithms , Secondly we researched animproved algorithm.Finally we have compared the results of experiment of those methods, the experimenting results of vessel features and matching algorithms is better than others,So,we take it to follow the vessels and position them and then we can forecast the moving direction of the two frames border upon.when vessel come into the region we appointed, monitor the vessel real-time ,If the vessel is torwards pier, starting up the prealarm system ,warning the driver change their direction immediately, and then avoid the accident happening successfully.