Enhanced detection of cement plants in remote sensing images using modified faster R-CNN

Accurate extraction of cement plants is important for the regulation of polluting enterprises and environmental protection. Conventional cement plant identification methods are characterized by low efficiency, high cost, and limitations incomprehensive and precise monitoring. Considering the successful application of deep learning in visual object detection, this research presents a modified Faster R-CNN network tailored specifically for detecting cement plants in remote sensing images. Our approach utilizes a multi-level fusion structure, integrating deep semantic features with more superficial detail features. This combination yields multi-scale feature maps that provide precise positional data along with deep semantic context. The experimental results demonstrate that our proposed method effectively detects multi- scale cement factory targets in remote sensing images, reducing the omission rate and improving target localization accuracy.