Estimating the Real Positions of Objects in Images by Using Evolutionary Algorithm
Zhixing Huang, Weili Liu, Jinghui Zhong · 2017
Estimating the real positions of objects in images is a fundamental operation in many data-driven modeling approaches. However, due to the perspective principle, the positions extracted directly from images usually are not accurate enough. Traditional data-driven modeling approaches using these data may not be able to build satisfying models. To solve this problem, this paper proposes an evolutionary algorithm based method to estimate the real positions of objects in images. Specifically, we formulate the problem of estimating the real positions of objects in distorted images as a parameter optimization problem by using some reference information of the images. Then, the differential evolution (DE) is utilized to search for the optimal solution. Our method is tested on two real-world datasets and the experimental results demonstrate that the proposed method is effective to estimate the real positions of objects in images.