ALGORITHMIC AND SOFTWARE IMAGE RECOGNITION METHOD FOR TRACKING OBJECT GEOLOCATION

Abstract

This article introduces a software method for image recognition designed to track the geolocations of objects captured in digital imagery. The proposed approach combines advanced computer vision techniques with geospatial data processing to enable automatic identification, classification, and spatial mapping of objects of interest within photographic content. At the core of the method lies a convolutional neural network architecture, which ensures high recognition accuracy by extracting and analyzing visual features from images. A distinctive feature of the system is the integration of an environmental context classification module, which segments the surrounding scene into semantic subclasses such as urban, natural, or industrial environments. This segmentation provides the model with additional spatial and contextual cues, significantly improving the reliability and precision of geolocation predictions, particularly in complex or ambiguous visual settings. The method is tailored for applications in environmental monitoring, reconnaissance, digital cartography, archival image analysis, and autonomous navigation systems. Experimental results demonstrate the effectiveness of the proposed approach compared to traditional algorithms, highlighting its potential to automate geolocation processes for diverse visual datasets. The implementation utilizes convolutional neural networks ResNet50 and EfficientNetB0, both pre-trained on large image datasets, which ensures high generalization capability. Experiments conducted on the Im2GPS3k dataset demonstrated a country-level classification accuracy of 57.3% and also 74.6% when the correct country was one of the five suggested by the network. The findings support the method’s scalability and future development, including the integration of multimodal data sources and more sophisticated deep learning models for enhanced accuracy and generalization across varying geographic and visual conditions.