Research On Reference Map Preparation And Embedded Software Optimization Technology In Image-aided Target Recognition

The preparation of target recognition reference map is one of the key technologies of aircraft target recognition.This technology provides basic data guarantee for scene matching.High-precision and time-sensitive reference image preparation technology is very important to improve the accuracy of recognition.The pixel positioning error in the reference image directly affects the accuracy of recognition.In recent years,with the improvement of the accuracy of inertial navigation-related sensors and the maturity of UAV control technology,the use of aerial image data captured by drones has become the trend of preparing reference maps for identifying ground targets.In this paper,the research on the preparation method of reference map is carried out by taking images based on the UAV with imaging platform.After the preparation of the reference map,the requirements of the target recognition algorithm are analyzed in depth,the quantitative evaluation indicators are discussed according to the characteristics of the reference map,and the software is designed and implemented in conjunction with specific engineering applications.The optimization method of scene matching algorithm performance improvement was carried out,and the experimental results showed that good results were achieved.The main work of the thesis includes the following three aspects:1.Based on the drone flying image with platform,the position and posture information of the drone,the installation posture of the turntable camera and the internal and external parameters of the camera are analyzed,and the method of correcting the original image taken obliquely to the front-view plane image is discussed in detail.And based on the geometrically corrected UAV images,the stitching algorithm is studied.Due to the diversity of the GPS/INS system of the UAV platform and the complicated definition of the coordinate system,this paper selects one of the aircraft coordinate definitions and implements the correction algorithm under the definition of the coordinate system.2.Based on the visual characteristics of the human visual system that are sensitive to edges,a scene matching algorithm based on the Sobel gradient direction is designed,which can adapt well to the heterogeneous grayscale distribution differences between heterogeneous images.The transplantation and optimization of the algorithm on the DSP platform are completed.3.The quantitative index of the matchability analysis of the reference image was studied,and the quantitative index analysis was carried out from the perspective of information amount,stability and regional uniqueness of the reference image,which provided an auxiliary method for the analysis of the matchability during the preparation of the reference image.