Cherlet, W; Cooper, Z; Van Den Broeck, WAJ; Disney, M; Origo, N; Calders, K; (2024) Benchmarking Instance Segmentation in Terrestrial Laser Scanning Forest Point Clouds. In: Proceedings of IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium. (pp. pp. 4511-4515). IEEE: Athens, Greece.

Preview

Text Benchmarking Instance Segmentation in Terrestrial Laser Scanning Forest Point Clouds.pdf - Accepted Version Download (569kB) | Preview

Abstract

Terrestrial laser scanning (TLS) has proven to be an invaluable tool in various forest ecology applications and forestry research. A crucial step in most TLS forest point cloud processing pipelines is instance segmentation; separating individual trees from the forest. However, automation in this area proves difficult, largely due to the heterogeneity of tree features and composition as well as overlapping, dense crown areas and understory. A lack of benchmarks and standard metrics complicates intercomparison of methods and hinders development in the field. This work proposes a set of metrics and methodology for benchmarking methods, and applies this to four open source TLS instance segmentation methods on a fully segmented 1.2 hectare benchmark dataset of a deciduous forest.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item