Detection and Tracking of Moving Objects Optical Flow Based

The problem of detecting and tracking moving objects on video is considered. The goal of the work is to develop optical flow algorithms for detecting moving objects in the presence of camera movement. Optical flow-based methods are computationally expensive and therefore require fast hardware and software solutions for implementation. Solutions are proposed based on the calculation of a sparse map (only for a few pixels), which can significantly speed up the calculation. The moving object detection algorithm includes: selecting a moving object that the operator is interested in tracking; identification of moving objects; detection and selection of objects of interest, analysis of the velocity vectors of selected objects, as well as localization while preserving the coordinates of objects and their characteristics. The selection of an object for tracking can be carried out by the operator by selecting a rectangular area of interest, or in automatic mode. The method includes: selecting a moving object that the operator is interested in tracking; identification of moving objects; detection and selection of objects of interest, analysis of the velocity vectors of selected objects, localization while preserving the coordinates of objects and their characteristics.