ORB-based Template Matching Through Convolutional Features Map
Dong Xiang Zhou, Yingxin Tian, Xiang Li, Jiefei Wu · 2019
Template matching is an important part of computer vision, but most of common methods do not work well in some cases, such as complicated background clutter, deformation and partial occlusion. Therefore, we present a novel method for image matching, which is useful, robust and fast. Its essence is the ORB-based Convolutional Features Map (CFM), which is used to measure the similarity between template and target image. We study its properties and apply it to a real-world dataset in complex environment. The result of experiments demonstrates that our algorithm outperforms other commonly used algorithm.