Ships Detection on Inland Waters Using Video Surveillance System

The video surveillance is used to monitor ships in order to ensure safety on waterways. The ships detection is a first step in a ship automatic identification process based on video streams. The paper presents a new algorithm for ships detection on inland waterways. The algorithm must detect moving ships of all kinds, including leisure craft, that are visible on a video stream and is designed to work for stationary cameras. Furthermore, it only requires an access to video streams from existing monitoring systems without any additional hardware or special configuration of cameras. The algorithm works in variable lightning conditions and with slight changes of background. In the paper, the test application implementing the algorithm is presented together with a series of experimental results showing the algorithm quality depending on different parameters’ sets. The main purpose of the tests was to find the optimal set of twelve parameters that will become the default setting. All moving ships, including small boats and kayaks, must be detected, which is the main difference from existing solutions that mostly focus on detection of only one vessel type. In the proposed algorithm, all objects that are moving on water are detected and then non-ships are eliminated by usage of some logic rules and excluding additional image processing methods.