Moving target tracking system for maritime surveillance

With the fixed camera mounted on the beach, it is necessary to make a successful target detection for the moving target tracking system in the maritime surveillance. For this reason, at the research made for taget detection, it is seen that the best method that can be used for the sea, which is a dynamic environment due to the factors such as wave foam, wave, water, wake, weather, sunlight reflected on the sea, etc., is the background method. The background extraction method is used to determine objects in images by classifying whether pixels are foreground or background by making a comparison between a background model created in images and instant frames. When we examine the best results obtained using the background extraction method, it was found that the ViBE algorithm made for motion detection was suitable. However, it has been observed that sometimes there are erroneous evaluations in the foreground and background classification when we run the ViBE algorithm on maritime surveillance, there may be noise in the image, and the target may have a foreground which includes a small part of the target but not the whole. Before designing the moving target tracking system, in order to handle these conditions, some developments have been made on the algorithm and appropriate results have been obtained. After receiving satisfactory results in the ViBE algorithm that we developed and comparing the results with the numerical values in the studies we have reviewed in the literature, the target detection part, the most important step of the target tracking system, was completed. After detecting target, it is important to track this target successfully in the transition between the frames. Moreover, sometimes the target-related detections are lost momentarily, in such cases important thing is being able to track the target from where it is left when the target reappears within a short period of time by keeping the information related to the target. For this reason, it is aimed to provide target tracking by using Blob analysis and Kalman filter for target tracking system. Using Blob analysis, it was ensured that the foreground of the noise which may occur in small dimensions outside the target foreground in the transitions between the frames is not taken into consideration. Thus, Blob analysis formed a significant part of the target tracking system by taking the foregrounds of only the target into the rectangle. When the target could not be detected instantly due to the reasons such as noise, rapid change in image during the target tracking process, the Kalman filter is used to continue for tracking target. Thus, when the target is detected again, the tracking is continued. In the study conducted in MATLAB with the RGB images obtained at 480x856 resolution using the Canon EF 70-300mm f/4-5.6 IS USM camera in the Singapore seas, at the source named "Singapore Maritime Dataset", it was found that the target tracking was achieved.