Solar Tracking Using Extended Mean Shift Based Color Histogram
Asepta Surya Wardhana, Astrie Kusuma Dewi

Department of Instrumentation Refinery Engineering, Politeknik Energi dan Mineral Akamigas, Cepu-Blora 58315, Indonesia

Abstract

Nowadays, there are many solar tracking applications using photodiode sensors and Solar Position Algorithm. This tracking depends on the power of light and natural conditions. Inaccurate sun tracking causes the heat concentration to become weak and miss focus on heat-receiving objects. We developed a tracking algorithm to track the sun to support the control system of the dual parabolic concentrator. This algorithm is based on Extended Mean shift to find the tracking position of an object in a video sequence. This algorithm is effective since it exploits the estimation of kernel density for searching the local maximum of a similarity measurement of the color histogram. Expectation Maximization algorithm is also employed to estimate the model parameters and to update the appearance of histogram. The updating histogram would improve the mean shift tracking accuracy and reliability. We successfully applied this algorithm for solar tracking using 148 frames of data. In this experiment, the results obtained in the form of the average value of the color similarity of an object tracking with a truth tolerance percentage of 98.39%.

Keywords: Dual parabolic concentrator, solar tracking, histogram, Extended Mean shift, Expectation Maximization

Topic: Engineering