•Our paper proposed a method to analyze the urban development utilizing the build-up area maps generated from multiple high-resolution remote sensing (RS) images. The previous researches mainly operate on medium-resolution RS images which lack of the details of sub-pixel features, having impacts on the reliability of mapping and analysis. Obviously, high-resolution RS images are more desirable for urban development observation. However, these images with higher resolution but smaller width, bring some challenges to mapping the whole urban with large coverage. Therefore, our paper develops a transferable build-up area extraction (TBUAE) algorithm using multiple satellite images to mapping the build-up area. This algorithm integrates the advantages of deep learning and transfer learning, and efficiently alleviates the pressure of deep learning on the demand for new satellite image samples which require time-and labor-consuming labeling manually. The main contributions are as follows:•Our method of built-up area mapping only requires automatic operation of the acquired remote sensing images, instead of consuming a lot of manpower which convenient to urban development analysis.•The TBUAE algorithm relives the pressure of obtaining high-resolution data in a large area of close time when carrying out urban analysis.•This study extracted the built- up areas of Zhengzhou in 2016, 2018, and 2020 by built-up area mapping to analyze its urban development in the past five years. Accordingly, driven factors are elaborately analyzed from specific aspects of economy and policy, respectively. Analysis of built-up areas—the most significant artificial urban areas—reveals physical development processes. Unlike previous research involving medium-resolution remote sensing (RS) images, this study used built-up area maps generated from multiple high-resolution RS images with abundant built-up area edge information to analyze development in Zhengzhou. A transferable built-up area extraction (TBUAE) algorithm was developed to map the built-up area maps. The algorithm allows the developed deep learning model used on a certain satellite image to be eligible for other types of satellite images by altering the data distribution with adaptive Wallis filtering (DT-AWF). The proposed method alleviates the pressure of deep learning on the demand for new satellite image samples that are time-consuming and laborious. Additionally, the accuracy of built-up area mapping using this method exceeds 90%. Quantitative and qualitative analyses were conducted on the map results to observe the urban development of Zhengzhou from 2016 to 2020. We found that Zhengzhou has expanded rapidly since it was defined as a central city in central China in 2016. Additionally, the suburban built-up area has expanded rapidly and developed together with the central city. Further, affected by the policy, the built-up areas in different regions of Zhengzhou has changed differently, the urban edge is more simplified, and the urban internal structure is more compact.