Research On Image Clearness Of Grain Depot Monitoring System Under Fog And Dust

With the upgrading of industrial technology,equipment such as intelligent security monitoring,image acquisition,target recognition and tracking have been widely promoted in the security of grain depot.The security video monitoring system can monitor the working conditions of some important places such as the main import and export grain storage passages,storage areas,operating points,equipment depots,and drug depots in real time,and warn the abnormal behavior such as personnel gathering,crossing the border,regional invasion and operation personnel violating regulations.It not only reduces the on-site inspections,reduces the cost of manual management of the grain depot,but also facilitates the management and data collection in the process of grain storage.However,the emergence of dust and haze weather in the grain depot causes the quality of the grain depot image obtained by the monitoring system to be seriously degraded,and the details are blurred,which limits the further recognition,analysis and use of the image.Therefore,in this paper,based on the method of machine learning,a comprehensive study is conducted on how to improve the image quality of grain depots in fog and dust environments.The specific research content is as follows:1.For the background of the images collected in the fog and dust environment of grain depots,most of them contain white areas.However,the existing defogging methods can affect the accuracy of the atmospheric light value estimation due to the presence of white areas.So the method based on quadtree decomposition is proposed to estimate atmospheric light accurately in the sky area.At the same time,based on the existing defogging methods that are easily limited by manual feature extraction and assumptions,this paper proposes an improved method based on neural network.First,using the multi-scale convolutional neural network to obtain a rough transmission map.Then using image fusion method to refine it.Finally,bringing the estimated parameters into the atmospheric scattering model to reverse the clear image.The quantitative and qualitative experimental results of synthetic and real-world grain depot fog and dust images show that the algorithm has a good effect on image texture details and sky area processing,and has high robustness and universality.2.Aiming at the problem of poor universality of each defogging algorithm when dealing with the degraded images with different fog concentrations,an adaptive foggy image defogging method based on support vector machine is proposed.First,combining the characteristics of the degradedimages of the haze weather,extracting the image features,and using the support vector machine algorithm to realize the two classification of the haze images,that is,foggy and fog free.Then,through the defined defogging algorithm,evaluates the image quality of the defogging results obtained by different defogging algorithms for the same image,and outputs the highest evaluated defogged image.Finally,the output results are classified again until the output image is satisfied as a clear index.The experimental results show that: from the perspective of visual effects,the proposed algorithm shows better contrast,brightness,and color saturation,and its scene adaptability and robustness are improved.