Research On Real - Time Matching Of Landmark Image Supported By Random Fern Bunling Algorithm

With the rapid development of computer science and technology, and extensive application of artificial intelligence technology, image matching as one of the key technology of artificial intelligence, also rapidly developed. Considering positioning and navigation functions of landmark, a new real-time landmark image matching algorithm was proposed by combined image matching with landmarks.Real-time matching of landmark images can be real-time positioning research, and navigation guidance and the like.The key technology of landmark image matching is image matching, the existing image matching methods can be broadly divided into three categories:based on the gray image,based on image feature and based on transform domain. But each method has a little less more or less, such as low matching accuracy, long matching time and poor stability.In this paper, a new image matching algorithm was presents, which starting from the feature point matching and based on random ferns algorithm, hoping to make up for lack of existing methods, main contents include:(1)Focused on the image features, introduces several image matching algorithm based on feature points:SIFT algorithm, SURF algorithm and random forest algorithms, introducing those algorithms’principle and implementation process, and analyzing advantages and disadvantages of each algorithm by experimental results.(2)This paper focuses on image matching method supported by random ferns algorithms and achieve recognition of the landmark, the matching process is divided into online and offline stage, greatly reducing the image matching time; using the improved FAST feature points and affine enhancing strategies to training random ferns, to ensure the matching algorithm performance while making the matching time can meet real-time requirements.Using a large number of landmark experimental data to evaluating the performance of the proposed algorithm from matching time and matching accuracy. Comparing the experimental results with the three feature points matching algorithm were mentioned above. Experimental results show that this algorithm takes place in the standard image scaling, rotation, occlusion and complex background conditions can efficiently, quickly matched, with good robustness. In order to more in-depth discussion on the application of this algorithm, human face and book covers also used to do matching study, it can produce better matching results and strong stability.