Accurate Registration Of Airborne Laser Point Cloud And Optical Image Based On Placement Parameter Optimization

Airborne laser scanning technology is a kind of all-weather,high-efficiency and popular remote sensing technology.The laser pulse emitted by it has strong penetrating power and is less affected by the environment.It can obtain high-precision and high-density data,which is an important technical means to obtain surface information.Due to the lack of ground object surface texture information in point cloud data,it is difficult to process the data.Therefore,the semantic information of optical images and point cloud data are usually used to achieve complementary advantages.The accurate registration between UAV laser point cloud and UAV image is the premise and key to realize the fusion of UAV laser point cloud and UAV image.Due to the small error of placement parameters and exposure delay of the high resolution camera in the airborne laser scanning system,the registration parameters calibrated in advance cannot meet the registration effect of UAV laser point cloud and optical image required by the experiment.Existing laser point cloud and image registration methods are mainly based on affine transformation to establish the registration relationship between point cloud and image,which cannot describe the complex transformation relationship between laser point cloud and image.Placed on parameter error,this paper adopts a unmanned aerial vehicle(uav)laser point cloud based on settlement parameter optimization and precision optical image registration method,set the camera parameters as optimization parameters,normalized mutual information as the laser point cloud and optical images,the similarity measure between the modified Powell algorithm as the optimization strategy,and get the optimal registration parameters,realizes the unmanned aerial vehicle(uav)laser point cloud and accurate registration between the unmanned aerial vehicle(uav)images.Experimental results show that the registration method based on camera placement parameters can achieve automatic registration between UAV laser point cloud and optical image,and the registration accuracy can reach pixel level.In this paper,the registration of airborne laser-point cloud data and optical images is mainly studied as follows:(1)Firstly,based on the theoretical basis of registration of airborne laser point cloud data and optical image,according to the principle of image generation of airborne laser point cloud intensity characteristics,the camera placement parameters are adopted as registration parameters.The experimental technical route was designed based on the placement parameters.Considering the experimental efficiency and accuracy,the experimental discussion was conducted on the similarity measure and optimization strategy.(2)Considering the experimental efficiency.Based on the mutual information registration primitive,the basic formula of mutual information is simplified.In the experimental environment of this paper,the information entropy of the reference image is identified as a constant,and the simplified experimental efficiency is verified.The verification results show that the experimental speed has been improved to a certain extent.(3)Through experiments,the method of UAV laser point cloud and optical image registration is verified.The experimental results are analyzed qualitatively and quantitatively,and a Powell algorithm(algorithm 2)is used for analysis and comparison.Qualitative analysis of the results will be 5 × 5 uniform block stacking,the block stacking map of the improved Powell algorithm(algorithm 1)contains obvious features such as road traffic lines,street lights,buildings,vegetation,etc.these features have high convergence degree in the image block and stacking area of(algorithm 1),and the interface is basically aligned with high coincidence degree.The result of algorithm 1 is visually accurate,and the registration effect is consistent with the high-precision evaluation of qualitative analysis.The quantitative analysis is based on the superimposed image,the uniform statistics of 12 image features superimposed on the image of the convergence pixel error.In quantitative analysis,the feature features at the image overlay are selected evenly,and the error value of the convergence of the feature images is counted,and then the average value is calculated.In algorithm 1,the maximum pixel error is 3.2 pixels,the minimum pixel error is 0,and the average pixel error is 1.6 pixels.The performance of the qualitative analysis of Powell algorithm(algorithm 2)is similar to that of algorithm 1.The number of iterations of quantitative analysis is more than 9 rounds,and the average pixel error corresponding to the same feature is 1.57 pixels.The results show that the accuracy of the two algorithms is close,and the average pixel error difference fluctuates in individual pixels,which shows more iterations when the accuracy is relatively high.Experiments show that the registration method based on the optimization of camera placement parameters can achieve automatic registration between UAV laser point cloud and optical image,and achieve pixel level registration accuracy.