Research And Implementation Of Visual Mosaic System

Domestic accidents and disasters as represented by coal mine occasionally occur underthe influence of artificial, technical, natural and other factors.Search and rescue robot as oneof the most effective methods, can be the first to enter the scene of the accident, its visualsystem observation directly affect the rescue workers making rescue plan.To solve the visual problem of getting a wide viewing angle difficult for rescue robots,visual mosaic system based on Embedded Microprocessor is designed including two aspectsof research, the image stitching technology and visual mosaic platform.Image stitching is mainly composed of image registration and image fusion. Twoimproved models are proposed in image registration. One of improved models based onHarris corners, firstly, narrow its search by optimizing the extraction of corners, and then usenon maximum suppression for Harris operator to obtain coarse corner sets, according to therelationship between adjacent images and Hu moment features accurate corner matchingfinally. Another matching model based on SIFT features cut the matching image by bilateralmethod of two blocks reasonably, and extract SIFT operator in the trusted matching block,remove the wrong points using bidirectional RANSAC, obtain a series of exact matchingpoint at last.Embedded platform and PC set up to coordinate the visual mosaics system. In theembedded platform, two stepper motors and CMOS camera constitute the3D cylinder head,and relying on S3C6410processor complete the control of image acquisition, preprocessing,storage, and so on. Meanwhile, ARM associates with the host computer to send theinformation or images by Ethernet; vision system is designed to improve image mosaic modelin host computer by C++and OpenCV, finally, image having a wide viewing angle is shownrelying on MFC.Through the test, two kinds of improved algorithm increase the matching accuracy, andreduce the calculation. Its noise immunity, suitability, stability are improved obviously. Inparticular, the improved SIFT algorithm comparing with before save the running time of 22.5%. So it’s an effective algorithm for mosaic model. According to the system test, systemwith high precision, generalization ability, low cost, enhancing the rescue robot’s detectionability and work efficiency, and provide reliable basis for post drawing underground disastermap.