A Dual Dynamic Feature-based Deep Learning and Computer Vision–Based Model for Multi-Object Classification Using Geospatial Satellite Imagery

This paper introduces a Dual Dynamic Feature-based Deep Learning and Computer Vision–Based Model for Multi-Object Classification Using Geospatial Satellite imagery using dynamic feature extraction neural network and selecting the feature selection multi object classification images. The proposed framework is developed multi neural networks for features and advanced YOLOv8 architecture to enhance detection accuracy and efficiency for real-world objects and counting objects such as people, vehicles, and trees across various imaging modalities. The integration of high-resolution data from satellite systems offers a comprehensive view of both ground-level and aerial environments, supporting large-scale object detection and geospatial analysis. To address the limitations of conventional single-source models, the proposed dynamic model utilizes adaptive feature extraction and multi-scale learning strategies, which improve generalization across different resolutions and environmental conditions. Experimental results show that the model applies to two verified datasets that achieves superior performance in terms of detection precision and computational efficiency, confirming its potential for practical applications in urban monitoring, and environmental surveillance. This paper contributes to the advancement of Dual Dynamic Feature-based Deep Learning and Computer Vision–Based AI Model to geospatial analytics by providing multi neural networks to identify multiple object types. The experimental accuracy results for multi object classification via two satellite imagery datasets achieve 99.9%.