International Journal of Engineering Trends and Technology (IJETT) – Volume 21 Number 5 – March 2015 A Complete Processing Chain for Shadow Detection and Reconstruction in VHR Images Vinaya V. Kulkarni #1, Priyanka S. Borkar #2, Rani S. Kapare #3 Prof. Neelam S. Labhade *4 # Student , *Prof & E&TC & Savitribai Phule Pune University,India. 1 Vinaya V.Kulkarni @ BE ECE,At Jspm’s ICOER Wagholi,Pune Priyanka S.Borkar @ BE ECE,ICOER Wagholi, Pune 3 Rani S. Kapare@ BE ECE,ICOER Wagholi,Pune 4 Prof. Neelam S. Labhade @,BE E&Tc, ME in Signal Processig. 2 Abstract— Building information of automated update in maps from highresolution aerial imagery is one of the most important and challenging researches in the field of remote sensing and photogrammetric. Up-to-date building information is needed for many practical applications such as GIS application analysis, city planning, fixed assets inventory, etc. Urban land used to focused on building extraction and height estimation from space borne optical imagery and study. The merits of such methods is a GIS databases for decision makers, 3D visualization of urban areas, and digital urban mapping. In particular, for efficient building extraction from optical multi-angular imagery, first essential to eliminate the presence of shadows in very high resolution (VHR) images, because shadow in VHR can represent a serious obstacle for their full exploitation. This paper proposes to face this problem as a whole through the proposal of a complete processing chain for Shadow Detection and Reconstruction in VHR images. After, a template matching algorithm is formulated for automatic relative height and estimation of the relative building height are utilized in conjunction with a support vector machine (SVM)-based classifier for extraction of non-buildings from buildings. The final results are presented as a building mapand an approximate 3Dimentional model of buildings extraction. The building detection accuracy of the proposed method is improved to 88%, compared to 83% without using multi-angular information. They represent an important problem for both sellers and users of remote sensing images. As a consequence, shadows can impact negatively in the exploitation of VHR images, influencing detailed mapping, leading to erroneous classification (e.g., biophysical parameters such as water, vegetation, or soil indexes), due to the partial or total loss of information in the image. To remove these demerits and, thus, to increase image exploitability, two stages are needed: 1) shadow detection and 2) shadow compensation. An example of the importance of getting shadow-free images is the massive tsunami in 2004 where it was crucial to obtain such images in a very short time in order to take rapid and crucial decisions in rescue missions [1]. Index Terms— Image enhancement, image restoration, missing data, shadow detection, shadow reconstruction, support vector machines (SVMs), very high resolution (VHR) images. Fig.1 Illustration of cast and self shadows. I. Introduction: such as the fact that shadow areas have lower Now a days, very high resolution (VHR) satellite images opened a new era in the remote sensing field. Because of the increase of spatial resolution, classification, new analysis and change detection techniques are needed. Although, VHR images exhibit resolutions which allow differentiating very well detailed features from tiny objects, like little building structures, vehicles, and roofs, trees. Although it is feasible to explain shadow characteristics to detect building position and to estimate their height and other useful components usually, shadows are viewed as undesired information that strongly affects images. ISSN: 2231-5381 The former requires primary information about the sensor and scenario. since, such knowledge is not available, most of the detection algorithms are based on shadow properties, brightness higher saturation, , and greater hue values. Other algorithms rely on the idea of adding characteristics capable to better discriminate shadow areas (e.g., normalized difference vegetation index [2], normalized saturation-value difference index (NSVDI) [3], and maximally stable external regions [4]). Another technique applies the principal component analysis to isolate the luminance component in an RGB image, where the detection of shadows appears more accurate [5]. Finally, physical properties (e.g., temperature) of a blackbody radiator have been exploited in a earlier method to detect shadows. In order to compensate shadow areas, there exist essentially three different categories : 1) linear correction; 2) http://www.ijettjournal.org Page 257 International Journal of Engineering Trends and Technology (IJETT) – Volume 21 Number 5 – March 2015 histogram matching; and 3) gamma correlation. The results obtained with two methods, namely, gamma correction and linear correlation, are compared. In [6], it is assumed that the restoration of shadows almost depends on the spectral signature of the spectral bands. Accordingly, first, the bands are threshold in an independent way, calculating the optimal threshold values by visual inspection. Then, a linear regression in each spectral band is carried out to correct the shadow effects. In this paper, an another method is proposed to solve both problems of detection and reconstruction of shadow areas. Shadow detection is performed through a hierarchical supervised classification scheme, while the proposed reconstruction relies on a linear correlation function, which exploits the information returned by the classification. The whole processing chain includes also two important capabilities: 1) a rejection mechanism to limit as much as possible reconstruction errors 2) explicit handling of the shadow borders. II. LITRATURE SURVEY The detection and removal of shadows have been simulated using different techniques. A survey of these techniques and their demerits are discussed in this section. In some applications, especially inspection system and, traffic analysis the subsistence of shadows may cause stern nuisance while segmenting and tracking objects. Shadows can cause object unification as a consequence of this, shadow detection is applied to situate the shadow part and discriminate shadows from foreground objects. Algorithms dealing with shadows were classified in two-layer taxonomy. A. Statistical non-parametric (SNP) approach This approach considers the color constancy ability of human eyes and exploits the Lambertian hypothesis (objects with perfectly matte surfaces) to consider color as a product of irradiance and repentance. The algorithm is based on pixel modeling and background subtraction. A statistical learning procedure is used to automatically determine the appropriate thresholds. i. ii. iii. More assumption Difficult process Complex C. Deterministic non-model based (DNM1) approach It performs object detection by means of a background suppression and one of its major novelties is the way the background model is computed . The model of the background is deļ¬ned ad a combination of statistical and knowledge-based assumptions. A point of the background assumes the value (in the RGB color space) that has the higher probability in a given observation. Drawback of this approach i. ii. Problem of ghost shadow Complex D. Deterministic non-model based (DNM2) approach This is also a deterministic non-model based approach, but we have included it because its capability of detecting penumbra in moving cast shadow. This approach is one of the most complete proposed in literature. Drawback of this approach i. ii. III. Very complex Less flexible BLOCK DIAGRAM: Drawback of this approach i. Less accurate ii. Distortion B. Statistical parametric (SP) approach Traffic scene shadow detection is an example of statistical parametric (SP) approach This algorithm uses two sources of information: local (based on the appearance of the pixel) and spatial (based on the assumption that the objects and the shadows are compact regions). Fig 2: Block Diagram Drawback of this approach ISSN: 2231-5381 http://www.ijettjournal.org Page 258 International Journal of Engineering Trends and Technology (IJETT) – Volume 21 Number 5 – March 2015 Support Vector Classification- IV. The classification problem can be restricted to consideration of the two-class problem without loss of generality. In this problem the goal is to separate the two classes by a function which is induced from available examples. The goal is to produce a classifier that will work well on unseen examples, i.e. it generalises well. Consider the example in Figure 3.Here there are many possible linear classifiers that can separate the data, but there is only one that maximises the margin. This linear classifier is called as the optimal separating hyperplane. Intuitively, we would expect this boundary to generalise well as opposed to the other possible boundaries. The key concept of SVMs, which were originally first developed for binary classification problems, is the use of hyperplanes to define decision boundaries separating between data points of different classes. SVMs are able to handle both linear, simple, classification tasks, as well as more complex, i.e. nonlinear, classification problems. 1. Start 2. Satellite image acquisition 3. Binary classification between Shadow and non-shadow Regions using SVM 4. Morphological Filtering on classified regions 5. Border Creation. 6. Multiclass Classification between following classes i. Vegetation ii. Soil iii. Roads and Buildings iv. Ash fault 7. Quality Improvement. 8. Shadow Reconstruction. 9. Pixel Interpolation for smoothing edges. 10. Reconstructed Image. 11. Stop. V. Fig.3: Optimal Separating Hyper plane Both non separable and separable problems are handled by SVMs in the nonlinear and linear case. The idea behind SVMs is to map the original data points from the input space to a high dimensional feature space such that the classification problem becomes simpler in the feature space. The basic principle of SVMs is a maximum margin classifier. Using the kernel methods, the data can be first implicitly mapped to a high dimensional kernel space. The maximum margin classifier is determined in the kernel space and the corresponding decision function in the original space can be non-linear [2]. The non-linear data in the feature space is classified into linear data in kernel space by the SVMs. The aim of SVM classification method is to find an optimal hyper plane separating irrelevant vectors and relevant by maximizing the size of the margin. ISSN: 2231-5381 ALGORITHM CONCLUSION: This project has dealt with the important and challenging problem of reconstruction of VHR images obscured by the presence of shadows. The proposed methodology is supervised. The shadow areas are not only detected but also classified so as to allow their customized compensation. The classification tasks are implemented by means of the state-ofthe-art SVM approach. A quality check mechanism is integrated in order to limit misreconstruction problems. Moreover, borders are explicitly handled by adaptive morphological filters and linear interpolation for the prevention of possible border artifacts in the reconstructed image. In general, from the obtained results, different considerations may be deduced. VI. http://www.ijettjournal.org RESULT Page 259 International Journal of Engineering Trends and Technology (IJETT) – Volume 21 Number 5 – March 2015 sensing image based on HIS space transformation and NDVI index,” in Proc. 18th Int. Conf. Geoinf., Jun. 2010, pp. 1–4. [3] H. Ma, Q. Qin, and X. Shen, “Shadow segmentation and compensation in high resolution satellite images,” in Proc. IEEE IGARSS, Jul. 2008, vol. 2, pp. 1036–1039. [4] H. Y. Yu, J. G. Sun, L. N. Liu, Y. H. Wang, and Y. D. Wang, “MSER based shadow detection in high resolution remote sensing image,” in Proc. ICMLC, 2010, pp. 780–783. Fig: Reconstructed Image REFERENCES [1]K. Kouchi and F. Yamazaki, “Characteristics of tsunamiaffected areas in moderate-resolution satellite images,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 6, pp. 1650–1657, Jun. 2007. [2] D. Cai, M. Li, Z. Bao, Z. Chen,W.Wei, and H. Zhang, “Study on shadow detection method on high resolution remote ISSN: 2231-5381 [5] S. Wang and Y. Wang, “Shadow detection and compensation in high resolution satellite images based on retinex,” in Proc. 5th Int. Conf. Image Graph., 2009, pp. 209– 212. [6] F. Yamazaki, W. Liu, and M. Takasaki, “Characteristic of shadow and removal of its effects for remote sensing imagery,” in Proc. IEEE IGARSS, Jul. 2009, vol. 4, pp. 426– 429. http://www.ijettjournal.org Page 260