Research Of Curve Matching Algorithm Based On Multi Support Domain

Computer vision, as an important branch of artificial intelligence, has been widely used in a variety of industrial production and consumer applications, such as overlapping photos will be seamlessly stitched together a single panorama; the objects or characters One or more snapshots are transformed into their 3D models; they can also be authenticated. Image matching as a key technical point in the fields of vehicle identification, 3D modeling, image stitching, and enhancement of reality. The main task is to find the feature primitives from two or more images taken at different times or different perspectives of the same scene Between the correspondence (points, curves and regions).This paper starts from the perspective of texture features, combines the understanding of the current texture, explains the definition of the texture, how to express it and the development process?and introduces the technology applied to the current curve matching algorithm and analyzes it. Of the technology to complete the curve matching. And the descriptor proposed in this paper is applied to this matching process, and a curve automatic matching algorithm with good robustness is designed.The most important is based on the existing descriptors, the introduction of multi-support region, proposed a novel multi-support region based on the brightness curve descriptor (MRIOCD). This description enhances the matching error caused by the similar curve and improves the accuracy of the curve matching by increasing the support area radius according to certain law.Has good robustness.In the automatic matching experiment, this paper compares the influence of the radius size and the number of support regions on the total number and the correctness of the image curve through the multi-group matching experiment, and selects the appropriate original radius and the number of support regions. Then, by comparing the total number and correctness of the curve matching of the IOCD and MRIOCD descriptor pairs under the conditions of rotation, illumination change, compression,angle transformation and so on, it is verified that the descriptor has good invariance under different conditions, And the curve matching of individual images under IOCD and MRIOCD is presented respectively. The matching of the images under the two algorithms is displayed visually. Finally, the descriptor is used for the image splicing experiment to test the feasibility of the descriptor. Experiments show that the descriptor of this paper has good invariance, and the designed curve matching algorithm has good robustness.