Research On Image Mosaic Of Multiple Wide Angle Cameras Based On IMX6Q

With the development of sensor technology and communication technology,wireless multimedia sensor network is more and more widely used,especially in disaster relief,environmental monitoring and other fields.As an important part of WMSN nodes,camera can play an increasingly important role in acquiring unknown environmental information.However,the current node mainly adopts a single camera,which has limited perception and a certain blind zone,so it is difficult to obtain comprehensive panoramic information.In response to this problem,this thesis studies the multiple wide-angle camera image mosaic,through the acquisition of six video to generate panoramic images to solve the problem of a single camera blind spot.On this basis,this article transplants the stitching algorithm to the IMX6Q embedded platform,realizes the real-time transmission of the panoramic stitching video,and allows the user to observe remotely.The research has positive and practical significance.The thesis analyzes the imaging model of the camera by analyzing the world coordinate system,the camera coordinate system,the image plane coordinate system and the image coordinate system.The polynomial model and the inverse division model are respectively used to correct the distortion of the original image.On this basis,the principles of cylindrical projection and homography matrices are studied,and preparation is made for the subsequent splicing.Since the panoramic image comes from multi-channel camera,this thesis studies the precise alignment algorithm of multi-image.Currently,there are some problems in stitching images,such as "ghosting" and local detail distortions,whether based on the global matrix alignment algorithm or the APAP-based alignment algorithm.In order to solve this problem,this thesis proposes an improved APAP algorithm.Two thresholds are set on the boundary of overlap region,and the weight of different feature points is calculated to solve the block matrix.Experimental results show that this method eliminates the problems of "ghosting" and distortion of local detail,and has better alignment effect.In order to enhance the image mosaic effect,this thesis studies the shape preserving algorithm of the mosaic image.When there is obvious parallax in the adjacent images,the transformation matrix of the image is subject to scaling and rotation,and the shape distortion will gradually accumulate with the subsequent splicing,which eventually results in the failure of splicing of multiple images.In order to solve this problem,a shape preserving algorithm based on gradually changing matrix is proposed in this thesis.The stitched image is divided into three regions.The left region uses the global matrix,the right region adopts the translation matrix,and the middle part adopts the gradually changing matrix.The experimental results show that the proposed algorithm can solve the problems of scaling,rotation and other problems generated in the image transformation process well and can generate a smooth panoramic image.In addition,this method can be combined with APAP to achieve precise alignment of overlapping regions.Based on the IMX6Q embedded platform,this thesis verifies the feasibility of multiple wide-angle camera images mosaic proposed in this thesis.The table generated by splicing PC is loaded into the IMX6Q file system,realizing six channels video splicing in real time.The nginx-rtmp server is built in the cloud to enable pushing flow and pulling flow,enabling multiple clients to remotely view panoramic information.In order to improve the real-time performance of the entire system,the thesis gives full play to the advantages of the hardware platform.This thesis uses multi-threaded approach to the task processing,and use the processor comes with IPU transcoding,VPU compression,and by way of table lookup splicing.Experimental results show that the panoramic video frame rate of 1504*288 resolution can reach 8?10 frames.