RESEARCH ON INTELLIGENT RECOGNITION OF VIOLATION BASED ON BIG DATA OF URBAN CONSTRUCTION

Abstract. With the rapid development of market economy and the continuous improvement of urbanization process in China, housing construction in almost all areas of densely populated cities has shown explosive growth. The existence of illegal construction, to a certain extent, not only causes the waste of land and resources, but also leads to the increase of the cost of the development of affordable housing, but also increases the security risks. It is urgent to solve the common problem of “urban disease” in violation of construction, but the conventional means of monitoring illegal construction mainly rely on on on-site inspection by law enforcement departments and mass reporting. Due to the limited inspection power and time, there are inevitably omissions. At the same time, there are difficulties in obtaining evidence in violation of construction investigation. Therefore, a new type of monitoring method is urgently needed. There is a wide market demand in China's urban management departments and land and resources departments for automated monitoring methods to reduce the cost of urban management. In this paper, urban spatial geographic information is acquired by means of remote sensing change detection, and compared with urban construction land planning approval data, including illegal matching recognition algorithm. Based on the technology of automatic urban change detection of grid image blocks, an efficient algorithm for building change detection is proposed. Establish a threshold recognition and accuracy test algorithm of urban building construction progress model parameters, and obtain information of illegal building construction progress and area based on grid image blocks. Artificial Intelligence (AI) is used to identify and extract buildings from satellite remote sensing images in different time periods. The dynamic change information of the research area is reflected by multi-source and large data integration technology of satellite and UAV remote sensing. Several optical image sample sets and test sets are established. The convolution neural network model is designed by sample sets. The accuracy and sensitivity of illegal identification can be improved by the combination of AI and in-depth learning. Using the method of monitoring, analyzing and comparing the big data of urban construction to monitor the illegal buildings in cities is not only fast and efficient, but also provides a scientific and objective basis for the relevant departments to enforce the law.