Research Of Multi-Source Data Registration And Fusion Algorithm For Helicopter Platform Under Complex And Low Altitude

Typically,the helicopter is featured by low-altitude and low-speed flight,strong mobility,high flexibility,and vertical taking-off and landing,which has been widely used in military and civilian sectors.However,since China expands low-altitude airspace,helicopter industry has been exposed to an increasingly complex flying environment,where low-altitude helicopters are threatened by considerable obstacles such as power lines,drones,trees and houses,resulting in greater dangers in flight safety.At present,the existing helicopter platform obstacle detection and identification systems mostly use single source sensors to work independently,such as visible light,infrared and so on.However,a single sensor will gradually accumulate due to its own errors,complex climatic conditions and environmental factors;resulting in poor reliability of obstacle detection and low recognition accuracy.Generally,using a multi-source cooperative sensor is one of the most important solution to facilitate detecting the complex low-altitude environment threats and improving the recognition ability.Multi sensors enable obtaining more spatial information of threats.Moreover,this approach makes use of the complementarity among different sensors,which could effectively overcome the limitation of single sensor in a complex environment,so as to improve the robustness of the system detection and threat identification.The accurate registration and efficient fusion of multi-sensor images is the premise and basis of obstacle detection.However,multi-sensor data has spatial and gray information differences in the time domain and spatial domain,which brings challenges to registration and fusion.To address the problem above,this thesis aims to carry out the research on infrared and visible image registration and fusion and other related studies.The detailed contents and innovativeness of the research are as follows:(1)To solve problems under the complex low-altitude environment such as poor texture details,difficult feature extraction,and the inefficacy in the algorithm of the existing infrared and visible image registration,this thesis proposes a fast image registration algorithm based on Singular Value Decomposition(SVD)and Scale Invariant Feature Transform(SIFT).Firstly,extracting the edge information of source image with Canny detection operator and finding the key points by constructed scale space detection.After that,the main direction of the key points is determined on the basis of the gradient direction of the key points,The feature descriptor matrix is generated,then reduce the matrix dimension by SVD;Finally,according to Euclidean distance matching of the feature descriptor,a transformation model is established to complete image resampling and realize image registration.The experimental data verify the effectiveness of the proposed method.Canny operator is used to extract image edge information,so that the algorithm has certain anti-noise ability.Meanwhile,the invariance of SIFT and the dimensionality reduction superiority of SVD reduce the computational complexity and improve the real-time performance of the registration algorithm.(2)To address the issue of strong scattering particles under the complex low-altitude environment,which causes color deviation and distortion of the infrared and visible light image resulting in the low contrast of the subsequent multi-source fusion results,this thesis proposes a multi-scale image fusion algorithm based on multi-index adaptive weighted map.Firstly,the source image is preprocessed with smoothing and denoising to extract hue H,saturation S and brightness V of visible image;After that,the weight graph is generated by combining multiple indicators such as information entropy,contrast and mean square deviation.A new brightness component is obtained by adaptive weight graph.The new HSV space is made up of the initial hue,initial saturation,and new brightness.Moreover convert HSV space to the RGB color space.Finally,to achieve enhanced visual effect on the fusion image,color correction and sharpening are processed.The experimental results verify the effectiveness of the algorithm.The fusion weighted graph of this algorithm is adaptive,and the proposed multi-scale space is sensitive to image size information.The fusion result has high contrast,clear details and bright colors.