Research On The Key Technology Of Infrared Imaging Detection For Low-slow-small UAV

With the development of small UAV technology,UAV has been widely used in many fields,including agriculture,express logistics,aerial photography,social security,disaster relief and military operations,and shows great advantages in effect and cost.However,the popularization of UAVs has brought huge security risks to social security.Under this situation,various anti low-slow-small UAV detection systems based on visible light,infrared and radar technology emerged.Among these detection systems,infrared imaging detection has become the standard configuration of most anti UAV systems for its good concealment,strong anti-interference and day-and-night working ability.In the infrared imaging detection system,High quality infrared image is particularly important to get high detection probability.The methods to improve the quality of infrared image include non-uniformity correction of infrared focal plane array,infrared image enhancement and background suppression of infrared dim target image.In this paper,we focus on the infrared image quality improvement and infrared dim target image background suppression algorithms in the anti UAV detection system.The main research contents and achievements are as follows.1.In order to solve the problems of slow convergence,image quality degradation and ghost artifacts in the existing non-uniformity correction(NUC)methods,two infrared image non-uniformity correction algorithms combining spatial and temporal nonlinear filtering are studied.The application of non-linear filtering in temporal domain makes the non-uniformity correction be realized by several consecutive images.It does not need hundreds of frames iteration to obtain the convergent correction parameters,and avoids the ghosting artifact,has better engineering application value.Two non-uniformity correction methods are proposed in third chapter.In the first algorithm,the non-uniformity in the infrared image is corrected by combining the trilateral filter in the spatial domain with the gradient weighted mean filtering in the temporal domain.In the second algorithm,the stripe non-uniformity is corrected by combining weighted guided image filtering with our adaptive weight and temporal non-linear diffusion equation.Due to the nonlinear filtering used in spatial and temporal domain,our method requires fewer sequential frames in a video to realize more accurate correction results.And experimental results show that thereis no ghosting artifacts in the correction results of the two non-uniformity correction algorithms,and the phenomenon of image degradation has been greatly improved.2.In order to solve the problem of low contrast and blurry details caused by high background and low contrast in infrared image,two image enhancement algorithms are proposed.The first one is based on Retinex theory and probability nonlocal-mean filtering.Before image decomposition single-scale Retinex is used to adjust the gray level of the image so that the details of the image in the dark are highlighted.And then PNLM filter is used to decompose the image into detail layer and base layer.To obtain the infrared image with more obvious contrast and better visual effect,contrast and detail enhancement are carried out for different layers.In the second algorithm,noise and detail weights are designed to extract different image information.The two weights are combined with weighted guided filtering to obtain more abundant detail and basic layers.The noise information is used to adjust the enhancement degree of detail layer adaptively,and a better enhancement result is obtained.Compared with the subjective and objective analysis of many methods,the experimental results show that the image processed by the proposed methods have more contents,details and texture,more in line with the visual effect of the human eye,thus a higher quality infrared image can be obtained.3.In order to solve the complex background suppression problem of infrared dim small target image in anti UAV infrared imaging detection system,two background suppression algorithms,based on the deep analysis of dim small target and background imaging mechanism in infrared image,are proposed in Chapter 5.These two are background suppression algorithm based on global filter and local standardized Euclidean distance and background suppression algorithm based on rolling guidance filter and double sliding window Mahalanobis distance.Global filtering and rolling guidance filtering are used to smooth the image at different scales,and the multi-dimensional information of the dim small target and background in the image are detected.Then the statistical characteristics of multi-dimensional images are used to obtain outliers(weak and small targets),so as to achieve the purpose of suppressing complex background.The experimental results show that these two algorithms solved the problem that the traditional background suppression algorithm has poor effect on strong edge suppression.A better background suppression result is obtained for the infrared dim target image.