Research On Key Technologies Of Space-based Aerial Target Detection And Recognition

Realizing the detection of moving targets based on space-based platform is the frontier research content in the current detection field.The current research in this field mainly focuses on ground-based radar or infrared,and there are certain detection limitations.Therefore,this dissertation focuses on the mechanism and method of aerial target detection on space-based platform,the characteristics of the infrared weak target and its background were studied,the detection system and method were deeply studied and analyzed,and researched on the recognition method of small infrared targets at low resolution in visible light.The research work and main research results of this subject were described below:Aiming at the problem that detection needs of aerial target on the complex background,the motion characteristics of targets at different detection angles and spatial positions were studied.Combined with the actual situation,the characteristics of the pitch angle,take-off angle and the rotation angle during the detection process were analyzed to realize the simulation of target radiation characteristics.And discussed the effects of complex cloud and sea background during the detection process.Finally,an integrated space-based infrared feature description and evaluation system for target and background combined with the detection system was established.Aiming at the problem that how to realize the detection of aerial target on the space-based platform,the specific detection system and method used for moving target were studied,and the corresponding technical indicators were established:resolution,detection spectrum range,detection distance and mode.When the relevant signal reached the image plane,analyzed the relationship between signal to background,signal to noise,signal to clutter,and determined that theis best within 50m at the band of 3-5μm,and theis best within 60m at the band of 8-12μm.The detection spectrum needed to be considered together with the detection distance.When the detection distance is within 500km,the reasonable detection spectrum in the mid-wave band is 4.25μm-4.52μm and 4.19μm-4.57μm,and the reasonable band length is0.27μm-0.38μm.The reasonable detection spectrum in long-wave band is 8.02μm-8.58μm and 9.24μm-9.96μm,and the corresponding shortest wavelength intervals are0.56μm and 0.74μm.In order to realize the search and discovery targets,a multi-line push-broom search imaging and area array gazing focus detection and multi-spectral band joint working mode are jointly used.Aiming at the high-performance detection of infrared weak and small targets on complex background,a method based on horizontal and vertical standard deviation multi-scale gray-scale value different weighted bilateral filtering was proposed.The discontinuity between the target area and the surrounding area was used to enhance the difference between target and background.Adaptive local threshold segmentation and global threshold segmentation were used to extract candidate targets.Experimental results show that this method was significantly better than other methods with weak signal-to-noise ratio,and it can increase the target by 6-30 times while suppressing the background.The(?)also confirm that our method was 5-12 times more effective than other methods.When the average single-to-noise ratio(?),the average detection probability of our algorithms is 95.71%,which is 50%higher than other methods.Aiming at the problem of aircraft target recognition in low-resolution visible light images,combined with the actual moving target detection mode,a method of target recognition in the infrared and visible band combined with PCA and nearest neighbor algorithm was proposed.It used the target detection result in infrared as the suspected position,and the proposed a slice of a certain size around the suspected position,and the used PCA and k NN to identify the slice.This method can achieve a recall rate of80.3%for targets with a resolution of less than 3m,and an average accuracy rate of97.52%.Finally,a simulation verification system and an actual ground multi-spectral band detection platform were designed in this subject.The validity of the detection system and processing algorithm proposed in this paper and the rationality of the processing framework were verified using simulation and measured image data.In summary,this subject has introduced the relevant theories of aerial target detection on space-based platform in detail,and put forward corresponding solutions to the difficulties in the current research,and conducted detailed comparison and experimental verification.This topic provides a strong theoretical support for the subsequent actual construction of aerial targets on space-based platform.