Key Technologies Of Image Stitching For Single Strip Of UAV In Power Line Patrol

In recent years,one of the trends in remote sensing is the UAV remote sensing.UAV remote sensing is a new type of remote sensing,compared with traditional remote sensing methods such as satellite remote sensing,UAV remote sensing with mobile,flexible,low cost,high resolution and other characteristics,and in smaller areas or traditional aerospace is difficult to reach Of the area quickly obtain high-resolution remote sensing images with significant advantages.Despite the significant advantages of UAV remote sensing,but in the follow-up processing of UAV images than the satellite and large aircraft images are complex,image processing need to consider the following points: First,the poor weather conditions or flight conditions are poor,The image of the unmanned aerial vehicle may be inclined,the angle of rotation,the triangular element of the pitch angle is too large and irregular.The overlapping degree of the heading is too small or irregular,the brightness of the image is large,the color difference is large,the color distribution is not uniform,The image is difficult to successfully match,stitching,access to the stitching image deformation,low precision,can not meet the needs of practical applications;Second,the image is small,large amount of data,a single image Can not cover the entire study area,resulting in late panoramic image acquisition difficulties;Third,the flight routes twists and turns,different routes will be cross,etc.,the image of the overlap is too small or irregular,adjacent airports can not choose a sufficient number Of the connection point,the accuracy of the aerial triangulation is low,difficult to stitch between different airports.These problems related to the UAV image stitching process of the difficulties,the quality of the pros and cons,especially for some single-band UAV images stitching,such as high-voltage transmission line visits and disaster emergency rescue,in order to save the cost of inspection line or Improve the efficiency of emergency disaster prevention,in which case the UAV is generally used in a single flight,but the actual flight belt relative to the conventional case of single-band in terms of more tortuous,complex,the corresponding image distortion,image rotation angle,Heading overlap,the degree of overlap,etc.become more irregular,follow-up image stitching work is very difficult.Therefore,UAV image stitching technology has become the most important factor in the development of UAV remote sensing technology.It is also very important to study the image mating algorithm with good robustness and fast and efficient UAV.This paper is based on the scale-invariant feature transform(SFIT)image matching algorithm to achieve image mosaic,in a large number of access to UAV image stitching principle and image mosaic algorithm based on the system of learning unmanned The key technology of fast image splicing.With VS2010 as the platform,the image mosaic algorithm is implemented in C ++ language,and the OpenCV library function is configured to improve the efficiency of image processing code compilation.In the experimental stage,40 sets of UAV image data were selected to analyze the SIFT algorithm,and four representative images were selected to show the results.It mainly includes the following steps: image preprocessing,local feature extraction and matching,image transformation parameter calculation,image merging,image fusion,output full image.The purpose of image preprocessing is to eliminate the geometric distortion and radiation distortion of UAV images in the acquisition phase,and to perform uniform light,uniform color,denoising and so on,and prepare for the later image fusion splicing.The task of scale invariant feature extraction and matching is to extract the image feature points and the character description,adopt the appropriate search strategy,find the same name in the overlapping area of the two images,and then purify the matching points to eliminate the unsuccessful points Correct.After establishing the correct pair of the same name,select a certain number of pairs to solve the image transformation matrix parameters,construct the geometric transformation model between images,and unify the images into the same coordinate system to complete the image simple overlay.And then the fusion image fusion processing,remove the color,stitching traces,and finally the image sequence splicing processing output full image.