An autonomic building detection method based on texture analysis, color segmentation, and neural classification

Building detection is a beneficial task that can be utilized to various kinds of application, such as disaster management, military application, and traffic planning. With this reason, an automatic building detection based on texture analysis and color segmentation is proposed in this paper. The proposed technique can be divided into three main steps. Firstly, in a preprocessing step, all color images were improved by filtering, color segmenting, and adjusting contrast with histogram. Then, in the feature extraction step, Gray-Level Co-Occurrence Matrix (GLCM) and Single Value Decomposition (SVD) were applied into the enhanced image to extract features. Lastly, all derived features were fed to neural classifier based on multilayer perceptron (MLP) to detect building objects. Results showed that the proposed method outperformed the compared method. It can efficiently detect building objects containing several shades of colors, can differentiate the closed building objects, and can distinguish the difference between building objects and other similar areas, while the method based on morphological operation produces many incorrect detected buildings. Additionally, the proposed method can achieve high performance values in accuracy, recall, precision, and F-measure. In addition, the proposed method has shown that it can be used as an efficient way for autonomic building detection.