Hyperspectral Detection Model And Spatial-spectral Sensitivity Analysis Of Shadow Area

Hyperspectral target detection can accurately identify the target and background,and has unique advantages and potential in the disclosure of low detectable targets,it has become a cutting-edge research in the target detection field.The reflected light energy from the shadow area is low,and the spectral signal of the corresponding region in the hyperspectral image is weak,and the signal-to-noise ratio is significantly lower than that of the non-shadow area,because of this,the judgement of whether it is a goal in the research area has become a common and difficult problem.However,the current research on the hyperspectral target detection for shadow targets is still in the development stage.There are still some problems in the research of influencing factors about the target detection in the shadow area,and there is also no effective algorithm and feasible scheme for the detection of intra-shadow targets.Therefore,based on the study of the influence mechanism of the shadow on the target spectrum,analyse the the optimal adaptive algorithm for the detection of the shadow target,and propose a multi-target dectection algorithm for camouflage paint target,based on this,the target detection strategy is used to extract the shadow region accurately and improve the detection accuracy of the shadow.The study not only provides the technical and data support for the analysis of the hyperspectral remote sensing target detection capability,but also provides the theoretical and data basis for the evaluation of the anti-hyperspectral reconnaissance capability of the ground target,which is of great significance to solve the shadow detection application of the hyperspectral remote sensing military target.The main conclusions of this paper include:(1)The ACE algorithm in four typical supervised detection algorithms is the most stable for different detection backgrounds and different illumination conditions,followed by CEM algorithm and OSP algorithm.In contrast,the SAM algorithm is sensitive to both background and illumination conditions,with the worst performance.(2)For the shadow area target detection,the anomaly detection algorithm is difficult to achieve the desired detection result;the classical RXD algorithm can detect the target of the illumination area but can not detect the target in the shadow area directly;The LPTD algorithm is more suitable for detecting strong reflection targets,but has poor result for the target which spectral has a significant fluctuation characteristic.(3)Otsu threshold segmentation method and region growing segmentation method was used to realize the hyperspectral image shadow detection,the results showed that the regional growth method has the better Result;Secondly,the image was segmented into light,penumbra and shaded areas,to remove the shadow and recover the information in the penumbra and shaded areas respectively to make the restored image tone evenly,So when the image is subdivided into light,penumbra and shaded areas can significantly improve the quality of the restored image.(4)The idea of introducing the moment matching method to remove the shadow from the image can restore the spectral information of the shadow area to a certain extent,so that the target spectral feature which is originally suppressed by the shadow can be excavated.Doing hyperspectral target detection algorithm experiment in the reatored spectral image of the shadow region,the results show that the detection accuracy of all the algorithms is improved in different degrees,and the detection efficiency of ACE and CEM algorithm in the known target unknown background region.However,due to the detection value of the algorithm(including the target and the background of the detection value)are distributed in the detection results gray scale high gray value range,the separation of the target and the background is not very good.The improved MTCEM algorithm achieves the detection accuracy of 0.9956.(5)Aiming at the multi-intrinsic spectral characteristics of camouflage paint target,the MTCEM algorithm is used to detect the target,and the detection accuracy of the target is obviously enhanced compared with the CEM algorithm.Aiming at the inherent shortcomings of the MTCEM algorithm in the impact of the target on the background,an improved MTCEM algorithm was proposed based on Mathematical Morphology.The experimental results show that the improved MTCEM algorithm improves the detection accuracy of the MTCEM,and follows the technical route of shadow detection,shadow removal and improvement of MTCEM,Can achieve the excellent detection effect of multi-target optimal detection at the same time.(6)To analyse the spatial scale and spectral scale of target detection under shading conditions.The results show that there is a certain similarity between the target and the background in the background of the grassland,and the target of the detection effect is required to occupy more than 87% of the abundance of the pixel.The difference between the target and the background of the road is large,and the target is only need to occupy more than 37% of the abundance of the pixels.In the background of the grassland,when the spectral resolution is less than 50 nm,the reflection peak of the target and the background disappears,the detection effect is serious decreased and the target can not be detected.In the background of the road,when the spectral resolution is lower than 50 nm,the reflection characteristic of the target disappears,although the difference between the target and the background is large,the detection accuracy still has a serious downward trend.The experimental results show that when the target is close to the background and the reflection peak is absorbed,the accuracy of the target detection is highest.