DisView: A Semantic Visual IoT Mixed Data Feature Extractor for Enhanced Loop Closure Detection for UGVs During Rescue Operations

Disaster-altered environments present a unique challenge for unmanned ground vehicles (UGVs), navigating dynamic environments for rescue operations. Existing loop closure detection methods already face significant challenges due to environmental variations, such as changing lighting conditions and the presence of moving objects. Disaster-stricken areas present an additional layer of complexity, as previously static landmarks can also undergo significant changes, further hindering accurate loop closure detection. Inspired by humans that recognizes already visited places by prioritizing semantic information like landmarks and text rather than just visual features, this article introduces DisView, a novel approach that leverages this insight for loop closure detection in disaster-altered environments. DisView utilizes a mixed data feature extractor to analyze visual, semantic, and textual data. This combined approach goes beyond traditional methods that rely solely on visual IoT data, leading to more robust loop closure detection even in disaster-altered environments and with viewpoint changes, varying illumination conditions, and the presence of moving rescue infrastructure. DisView’s effectiveness is validated on six publicly available datasets, demonstrating superior performance in diverse lighting conditions and viewing angles. By enabling reliable loop closure detection, DisView significantly improves UGV navigation in disaster-altered environments, ultimately supporting faster and more efficient rescue operations.