Design of a Computer Vision Algorithm for Surface Vessel Position Detection in a Laboratory Pool

The growing availability of cameras and advancements in computers hardware that led to increase in computer performance make systems based on computer vision attractive because of its ease of setup and use but at the same time low economic expenses. There are many methods for object detection and recognition, image segmentation that are already in use in many practical applications such as face recognition, industrial production etc. Many of these use cases do not require the knowledge of the exact position of the object as detection methods usually provide axis-aligned bounding box. In order to get more precise borders of the object image segmentation methods could be used, but in some applications such as control systems based on computer vision it is necessary to not only know the exact position of the object, but also its direction. In this study the objective is to find the linear and angular direction of surface vessel using computer vision approach. The proposed methodology is capable of obtaining the center coordinates of the surface vessel as well as the direction the surface vessel is heading. Out findings indicate that this could be done based on just image processing and some transformations without the use of deep learning approaches thus avoiding the need to collect, annotate and clean data etc. The approach considered in this study is based on filtering, edge detection, dilation, contour tracing and Hough Transform.