Human Tracking Objects in Dark Situations Based on Thermal Using Support Vector Machines, L1 Tracker Using Accelerated Proximal Gradient Approach, and Kernelized Correlation Filter Methods

Abstract


 
 
 
 Pedestrian safety on pedestrian lanes on the side of highways or roads in housing with heavy or quiet traffic conditions needs to be a public concern. Security that must be considered, object tracking is needed to carry out surveillance in improving pedestrian security, and it is also necessary to install thermal camera devices to find out the position of objects such as humans, in various positions of viewpoints that can be applied and applied to monitor the environment. To classify objects such as humans who are in dark or low-light conditions, namely by using the Kernelized Correlation Filter (KCF) tracking method, Support Vector Machines (SVM), and L1 Tracker Using Accelerated Proximal Gradient Approach (L1APG) based on a distance of 10 meters, 15 meters, 20 meters and the size of the object in the dataset. The results of the study with 1684 image inputs. Good performance for each success plot distance on the SVM method is 99.25%, 99.75%, 98.74% because it can track successfully based on the object being traced. Good performance for each precision plot distance on the KCF method of 51.88%, 46.8%, 63.81% has precise accuracy results against the object being tracked.