Research Of Moving Object Detection Key Technologys Of Vehicle-mounted Photoelectric System Under Raining Situation

With the progress of science and technology,photoelectric detection equipment as an effective method of reconnaissance in the modern battle has attracted an increasing number of attention.Photoelectric detection equipment takes high precision camera as imaging unit,and converts the optical radiation signals into corresponding electrical signals(images and videos).The resulted electrical signals can be collected,analyzed as well as processed,and finally finish tasks,such as target identification,tracking,and some complicate operations.Subgrade vehicle photoelectric detection equipment as one of the most common method which to collect informations,it is an indispensable significant component in trinity reconnaissance warning system for the future advanced technology,high informatization joint warfares.The study on photoelectric detection technology for armored vehicles has important military strategy significance.Moving object detection is the main approach to detect the valid target in battlefield for photoelectric detection system.The working condition of armored vehicle is abominable and changeable,which makes the moving object detection more difficult.Based on the working condition of photoelectric system,this paper discusses the related technical problems of moving object detection.Furthermore,the aims of this paper are listed following,First,to explore a kind of digital image stabilization(DIS)method that adaptive to inter-frame motion and solve the poor precision under various coupling dither interference factors;Moreover,on the basis of DIS,researching a kind of moving object detection method that robust to the changing view content;Last but not least,to explore a rain removal method to enhance video sequence and improve the accuracy of moving object detection under rainy condition.For these purposes,the main research contents of the article are as follows: First,In order to overcome the weakness of adaptability of traditional DIS method under various interference factors,this paper proposes a motion filtering method based on complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and relative entropy.The paper outlines the interframe motion parameters models.Combined with the practical application environment,the problems existing in the motion estimation method and the traditional filter method are analyzed.This method decomposes the global motion vector into multidimensional intrinsic mode functions(IMF)through the CEEMDAN,which realizes the adaptive separation between intentional motion component and random jitter component.The IMFs are classified by using the relative entropy,and the intentional motion vector and random jitter vector are summed to reconstruct.According to the reconstructed random jitter vector,the video sequence is compensated and the video stabilization realized.Simulation results show that,compared with the best result of other DIS method,the intentional motion vector obtained by proposed approach has smaller root mean square error(RMSE)mean and variance,which the mean and variance of RMSE can be reduced by 10.0% and 17.1% at least.The resulted video sequence by using the proposed DIS method has higher peak signal-to-noise ratio(PSNR),which can be increased 13.1% and 16.7% respectively.Second,in order to solve the low precision prblem of traditional moving object detection,an improved background codebook algorithm based moving object detection method has been proposed.This paper summarizes the background codebook model,and demonstrates the reasons of high detection error by experiments.By redefining computing method of color distortion and brightness threshold interval,the lack of codeword problem and inaccurate brightness interval problem have been solved.Based on the improved background codebook model,the moving object detection method under dynamic background has been established.New method can adaptively adjust parameter according to the brightness,which has strong robustness to brightness changes.Simulation experiments show that,compared with the best results of traditional moving object detection methods,the T/JE and NAE of the results obtained by presented method can be reduced by 30.7%,and 33.8% at least.The simulations on jitter video sequence show that the detection results can accurately reflect the moving object in the field.Third,in order to solve the rain effection on moving object detection under rainy weather,this paper puts forward a method based on the 2D variational mode decomposition(VMD)and selective median compensate strategy.According to the rain imaging frequency characteristic,the rainy image is decomposed via 2D VMD,and the sub-image contains the rain is selected.According to the rain imaging brightness and spatial characteristics,the raindrop positions are determined by threshold detection method,and the rain pixels are compensated by averaging the non-rain pixel values around the rain pixels.This method only compensates the rain pixels on the narrow band of sub image compensation,which effectively avoid the useful information losing caused by the de-rain operating.By comparison with the experimental results of different rain removal methods,proved that the method can not only get better rain removal effect,but also get clearer reconstructed image.Finally,the actual rain scene video moving object detection experiments are carried out.Simulation experiment results show that the moving object detection is more accurate after the derain process,the T/JE and NAE can be reduced by 41.16% and 68.90%.