Design and implementation of embedded vision-based tracking system using FPGA-SoC

Abstract

An implemented vision based tracking system has become an important application of embedded in the field of vision and surveillance systems. Object tracking is a fundamental component of embedded vision that is very beneficial in several applications such as unmanned vehicles, surveillance, automated traffic control, biomedical image analysis and intelligent robots. That is why many of researchers have been suggested different embedded vision tracking approaches. However, some of these studies suffer from numerous problems have been manipulated such as many camera motion, abrupt changes in the appearance patterns of both the object and the scene, therefore object tracking is a challenging problem. This project addressing a significant issue namely multiple objects tracking. One of significant problems that are faced the researchers is multiple objects tracking which is addressed in this project. Therefore, this project is designing and implementing an embedded vision based for multiple object (color) tracking system using FPGA-SoC. The proposed method has adopted a passive tracking vision system based on platform DE1-SoC and D5M camera. Every frame captured by the camera is processed to detect the targeted object using color based approach. The evolution of proposed method have been done based on ten experiments for two different objects (red color and blue color). As a results, the passive tracking embedded system have been tracked multi objects in different direction vision based on several angles measurements. The tracking distance of multiple objects (colors) was reached up to 30 meter for sized 15×15 cm object shape. The project is presented a smooth performance because of the DE1-SoC has the capability to carry out tasks in parallel and at a high frequency up to 1.6 GHz. The resources total of logic elements (LEs) were used in this project 9076 elements.