Shadow On-line Processing Of High Resolution Images In The Process Of Taking Aerial Images

High resolution images contain rich spatial information which have significant applications in environmental resource protection,earthquake relief,emergency response and so on.Aerial remote sensing is the main platform to obtain high-resolution remote sensing images.In the process of aerial photography,some areas are blocked by surface features and cannot accept sunlight,which form shaded areas with lower brightness.These shadows lead to quality degradation of high-resolution image which cannot meet the requirements of interpretation and other follow-up applications.If we choose to fill the fly film,it not only leads to time-consuming and high cost,but also cannot provide images which meet the quality requirements timely and cannot be applied to earthquake relief and emergency response.The significance of on-line shadow processing for aerial imagery is to evaluate the quality of the image in real time and to determine whether it is necessary to make a fly back to improve the quality of aerial photography.How to realize shadow detection and compensation quickly,check the image quality timely,reduce or remove the shadow effects effectively,reduce flight costs and increase the utilization of high-resolution remote sensing images is the main issues discussed in this paper.In this paper,the spectral characteristics of shadows in different feature components is summarized.We can observe that the shadow has the high values on the C3 component,which is beneficial to the shadow region extraction.The key of shadow detection is image segmentation.Considering that regional growth is an object-oriented segmentation method that the limitation of segmentation is reduced and the segmentation result is more accurate.In this paper,regional growth is chosen as the main method of extracting shadows.In order to improve the automation of shadow detection,this paper designs the automatic selection strategy of the regional growth parameters-seed points and thresholds.After processed by Gaussian filter on the C3 component,the initial results of shadows are obtained by region growing.In order to solve the problem of "empty" and small shaded areas in the initial test results,it is necessary to deal with morphological phenomena to reduce the false detection rate of shadow pixels and improve the accuracy of shadow detection.In this paper,the commonly used shadow compensation models are compared and summarized.Local compensation model which is simple and has less parameters and comprehensive compensation mode which is adaptive strong and has good compensation effect are selected as an experimental models.In order to obtain the compensation parameters automatically,it is necessary to obtain a large number of similar pairs of points in the shadow area and non-shadow area.It is very important to select the appropriate strategy of homogeneous points matching.To solve the problem of correlation coefficient matching strategy,the paper proposes an improved ISODATA classification similarity matching strategy which can select the appropriate number of points with higher precision to acquire compensation parameters automatically.Through a large number of experiments,it is found that the similarity point matching strategy based on ISODATA classification is closer to the ideal value of compensating parameter than the correlation coefficient method,and the compensation result is more stable.In this paper,the shadow automatic detection method based on C3 component and the algorithm of shadow automatic compensation based on ISODATA classification are evaluated.According to the result,it is proved to the proposed method can be the technical support for the on-line processing of shadowing during aerial photography.