Development of AI-Driven Mobile GIS for Real-Time Urban Land Use Mapping

Thesis Abstract

The rapid growth of urban populations and expanding land use demands necessitate innovative approaches to urban planning and land management, particularly in real-time monitoring and mapping. This study addresses the critical challenge of timely, accurate, and cost-effective urban land use data acquisition by developing an AI-driven mobile Geographic Information System (GIS) framework capable of real-time land use classification and mapping. The primary aim is to design and evaluate a mobile GIS platform integrated with artificial intelligence algorithms to facilitate instantaneous urban land use updates, thereby supporting sustainable urban development initiatives. The specific objectives include (1) designing a mobile GIS architecture optimized for real-time data collection and processing; (2) integrating machine learning models—specifically convolutional neural networks (CNNs)—for land use classification; (3) evaluating the system’s accuracy, usability, and efficiency in a dynamic urban environment; and (4) assessing the impact of the AI-enhanced mobile GIS on urban land management practices. The research adopts a mixed-methods approach, combining quantitative data analysis with qualitative usability assessments. The population comprises urban land parcels within a medium-sized city with diverse land use types, totaling approximately 15,000 land plots. A stratified random sampling technique was employed to select 600 land parcels representing residential, commercial, industrial, green spaces, and transportation zones. Data collection involved deploying a custom-built mobile GIS application on 50 field surveyors' smartphones, equipped with high-resolution cameras, GPS, and sensors for environmental data. The AI component utilized trained convolutional neural networks (CNNs) based on a dataset of 10,000 labeled satellite and drone-based aerial images, processed through transfer learning techniques to classify land use types. Data analysis incorporates accuracy assessment metrics, such as precision, recall, and F1-score, to evaluate AI classification performance. Spatial data collected in the field was integrated into a geodatabase, with post-processing involving Geographic Information System (GIS) environments for spatial analysis. The system's usability was examined through heuristic evaluation and user satisfaction surveys, analyzed via thematic analysis. Statistical validation involved regression analysis to identify relationships between AI classification accuracy and land use variability, complemented by descriptive statistics to portray the system’s operational efficiency. Expected findings suggest that the AI-driven mobile GIS will achieve classification accuracies exceeding 85%, significantly outperforming traditional manual mapping methods. The system is anticipated to reduce land use mapping time by over 60%, offer real-time updates, and demonstrate high user satisfaction. The integration of CNNs is expected to enhance land use discrimination in complex urban settings, supporting more responsive urban planning and management. Additionally, the study aims to establish a comprehensive framework for the deployment of AI-powered mobile GIS tools across different urban contexts, emphasizing scalability and adaptability. This research contributes to knowledge by bridging the gap between emerging AI techniques and mobile GIS applications in urban land management, providing a replicable model for real-time spatial decision support systems. It extends existing theories of spatial intelligence and mobile GIS usability, incorporating the Theory of Technological Acceptance Model (TAM) to explain user adoption and efficiency gains. The novel integration of transfer learning CNN models with mobile GIS represents a significant advancement in the field, offering scalable solutions for urban planners, policymakers, and stakeholders. In conclusion, the study affirms that AI-driven mobile GIS can substantially improve the accuracy, speed, and usability of urban land use mapping processes. It recommends further research to incorporate additional AI models, such as object detection and semantic segmentation, and to expand the system’s application to larger urban regions. Policymakers are encouraged to adopt AI-enhanced GIS tools as integral components of smart city initiatives, aligning urban planning processes with technological innovations to foster sustainable and resilient urban environments.