Research On Two-dimensional And Three-dimensional Image Matching Algorithms Based On Compound Eye Model

The SR-3000camera is an active camera, which can also obtain the depth distance information of the object when shooting the scene to get its grayscale information, but the images obtained by the camera are with low resolution, poor image quality, increasing the difficulty for the follow-up research. While Images obtained by the two-dimensional color camera are with high resolution, good image quality, but without the depth of information of the environmental objectives. To complement each other’s advantages of the image information obtained from the SR-3000camera and the two-dimensional color camera image, and to improve the accuracy and reliability of the3D object extraction and recognition, we match the image information from SR-3000camera and the two-dimensional color camera of the same scene at the same time. Therefore, researching the matching problem of different resolution images has important theoretical and practical significance.In order to conduct the matching of the image information obtained from the two cameras, and get a better composite image information as well as make the effective perception of the surrounding environment, we set up the hardware and software platform for the compound eye camera system, fixing the SR-3000camera and two-dimensional color camera on the same device to carry out the work on image acquisition.For the problem that the resolutions of the images to be matched are quite different, we learn and study the traditional image matching method, and meanwhile research on different scale image matching algorithms. We work in realistic scenarios to match different resolution images by our program, improve visualization effect of the match results, and analyze the experimental results.For the problem that the ICP point cloud integration technology is very sensitive to noise, we utilize the SIFT feature match and the second match of the matching points, building the positional relationship between the images from the three dimensional camera at different time, as well as achieving the three dimensional scene splicing at different time.