Object retrieval system based on feature matching technology

As the underlying visual technology, image matching technology plays an extremely important role in many fields. In these many fields, in order to get better results, most of them currently use deep learning as the mainstream algorithm, but deep learning algorithms have high requirements for the performance of the device, so it is not enough for mobile terminal to support its carrying deep learning tasks. This thesis attempts to build a suitable application scenario for them using image matching technology which is more suitable than deep learning. The main content of this thesis is to focus on the most widely used SIFT algorithm in the classic feature-based matching technology, and based on the SIFT algorithm, combined with the knowledge of homography to construct an object retrieval system for small mobile terminal. The system can retrieve the object in a picture containing the object in the system. The system uses the classic SIFT algorithm as the mainstream algorithm to meet the needs of these small mobile terminal. At the same time, the system also has a very high retrieval success rate.