Mobile Mapping System for Point Cloud Acquisition in a Forest Environment with an Action Camera

Abstract. This paper presents a methodology for generating georeferenced point clouds using a GoPro camera that incorporates a GNSS navigation receiver. The camera is mounted on a helmet worn by an operator, who simply needs to activate the camera and traverse the area of interest in overlapping strips, capturing 4K video at 60 fps. Frames are extracted from the video at an appropriate rate and processed using a structure-from-motion algorithm. This approach refines the camera's trajectory and produces a georeferenced dense point cloud. The georeferencing of the point cloud relies on the camera's GNSS-derived projection centres, which can be interpolated for each extracted frame. However, in forested environments, the reduced positional accuracy of the GNSS can introduce significant errors in the scale and orientation, limiting the accuracy of extracted dimensional parameters. To address these issues, the system incorporates a simple calibration strategy: a vertical pole of known length is placed in the surveyed area to provide a reference for scale and orientation correction. Once calibrated, the point cloud is processed to generate a canopy height model. Additionally, the point cloud can be segmented horizontally at 1.3 meters above ground level to extract individual tree rings, allowing measurements such as the diameter at breast height. The methodology is evaluated through various tests, and its accuracy is thoroughly analysed.