Sensör füzyonu ve parçacık filtrelemesi kullanarak hareketli nesnelerin tespiti ve takip edilmesi

Object detection and tracking problem has long been an important topic in the literature. The importance of this subject continues to increase due to the advantages that object detection and tracking provides especially for every day practical life and hence the need for its pervasive use. There are numerous different types of object tracking and various different tracking methods for each type of tracking. This thesis presents a moving object detection and tracking system with a Particle Filter algorithm. The goal is to build an infrastructure that will allow following an unknown moving object in a region with numerous other dynamic objects. Several components are used to determine objects; to estimate self-localization; and match the determined objects in the next iteration with the previously determined objects in order to tag each object with a particular identification. Specifically, LIDAR is used to determine the objects, IMU(Inertial Measurement Unit) to estimate relative translation and rotation, Odometer and GPS to help increase the accuracy of the self-position that is calculated by the IMU. The Particle Filter algorithm predicts self-position, utilizing the data received from both the IMU, the Odometer and the GPS. Computational cost is also taken into account during the clustering and matching stages. Performance and detection accuracy tests are carried out using various sized objects, as well as different environmental settings in order to conduct a comparison analysis for the gathered data.