Classification of Urban Construction Land with Worldview-2 Remote Sensing Image Based on Classification and Regression Tree Algorithm

With the repaid urbanization process, the scattered existence of urban construction land and the over use of urban construction land restrict the healthy development of the city. The remote sensing monitoring of urban construction land can help to monitor the layout of urban construction and realize the meticulous management of urban construction land. Based on the Worldview-2 high resolution remote sensing image, combined with Bayesian algorithm and classification and regression tree (CART) classification algorithm, this paper objectively classified the urban construction land of the Haizhu district, Guangzhou city, China. The results showed that the segmentation scale was 40, the color parameter was 0.8, and the tightness parameter was 0.5. The segmented object can well reflect the information such as roads, water bodies, vegetation and buildings in the urban construction land. Compared with the Bayesian algorithm, the overall accuracy of the CART classification results is higher, and the expression of the ground types (green space, buildings, water bodies and roads) is better. The classification results of the CART were more uniform and the phenomena of salt and pepper was less. The consistency of the road and the water was better. In this study, high resolution remote sensing images and object-oriented machine learning algorithms are used to classify the construction sites of large cities, which can provide scientific support for urban renewal and development decision-making.