Accurate detection and mapping of manholes are essential for urban infrastructure management, facilitating efficient maintenance and safety. This paper introduces a novel methodology that integrates the open-world object detection model, Grounding DINO, with geographic information systems (GIS) to detect and geolocate manholes in urban environments. Unlike traditional object detection approaches that rely on extensive labelled datasets and predefined object categories, Grounding DINO, a transformer-based model, leverages natural language processing for adaptable, scalable detection. Grounding DINO processes natural language descriptions to detect the manholes in an open-world context, overcoming the limitations of predefined object categories. Detected manholes are localized using multi-view triangulation, which refines their 3D positions by leveraging redundant camera viewpoints and intrinsic calibration parameters, which ensures accurate geometric mapping of manhole centers. The resulting geospatial coordinates are transformed into the WGS84 system using a global navigation satellite system/inertial navigation system (GNSS/INS) for compatibility with GIS platforms. The proposed approach achieved sub-meter precision, with mean localization errors of 0.36 meters in easting and 0.34 meters in northing, evaluated on KITTI dataset sequences under various urban conditions. The seamless integration of object detection and g