Computer vision workflows for physical carrying capacity analysis and monitoring in linear recreation corridors

Linear recreation corridors are increasingly managed through carrying-capacity frameworks, yet implementation is often constrained by a lack of scalable, transparent measurements of the physical conditions that shape where and when recreation use concentrates. This study develops and demonstrates two computer vision workflows for quantifying physical infrastructure and activity indicators in a heavily used river recreation corridor in Idaho, USA. The first combines high-resolution aerial imagery, model-assisted human labeling, and semantic segmentation to map corridor-scale parking opportunity. The second applies object detection to motion-triggered photographs from fixed cameras to derive temporal indices of parking-area vehicle activity and relative on-river activity. Across the corridor, the first workflow mapped 13.8 acres of parking-related surface (2.8 paved and 11.0 dirt/gravel) and the second processed 406,869 photographs, yielding 233,139 vehicle and 2036 watercraft detections. The application demonstrates how machine learning can transform imagery at a scale impractical for manual review into repeatable infrastructure and activity indicators that complement social and ecological standards in recreation carrying-capacity planning.