Automatic Detection and Extraction Building Complex Geometric of the Pixel-based and Object-based from Aerial image and Lidar Data

Automatic Building Detection and Extraction in multi-source aerial data are important in many applications, including map updating, city modeling, and urban growth analysis and monitoring of informal settlements. The main purpose of this paper presents a new automated method for building extraction from fusion of Lidar data and aerial images. For this purpose, at first part, we reviewed and evaluated building detection methods (ANN, SVM, MD) in Pixel-base and Object-base levels. The result for study area show a better performance of SVM with an overall accuracy of 95.9% and Omission error of “Building” class equal to 6.2% and Commission error of “Building” class equal to 3.2% which Errors of commission represent pixels that belong to another class that are labeled as belonging to the class of interest. Errors of omission represent pixels that belong to the ground truth class but the classification Technique has failed to classify them into the proper class. The second part of this paper is to extract and reconstruct the building edges based on the proposed method of building detection. For this purpose, building the edges are segmented based on spectral and geometric segmentation. In continue, based on least squares equation, the edge segments of each building are reconstructed. The results extraction show has proved an overall accuracy of 96.¬8% and an Omission error of “Building” class equal to 5.9% and commission error of “Building” class equal to 2.5%.