Object Tracking using Superpixel Confidence Map in Centroid Shifting Method
Indian Journal of Science and Technology
Objectives: To help security system works better, many countries especially developed countries installed surveillance security cameras. They used it to help find desired person whether they are criminal or not. Methods/Statistical Analysis: In order to do object tracking task, using colour-based tracking algorithm will give more stable result. By trying to get with different approach, the method that proposed is came from two algorithms. There are Super pixel tracking and centroid shifting
... d for tracking. Because both of algorithms give promising results in order to do tracking task, it is good to take each of advantages character from both algorithms. Findings: The proposed method is used Super pixel confidence map to get region of the object and determine between object and background. By using Super pixel confidence map, the tracker will be able to discriminate by measure the value. If the value is high, it is more likely to the object. And if the value is low, it is more like to the background. Before it used super pixel confidence map value, it will do a centroid shifting based to find target location by weighted the area with mean of the centroids comparing to each color bin of the target. The experiment will compare proposed method with other previous algorithms, original tracking based on centroid shifting and super pixel tracking using a same dataset. Improvements/Applications: This algorithm can be helpful for enhanced other application such as Object Recognition, Person Re-initialization, and some other applications in deep learning especially for object recognition.