LiDAR no setor florestal: cubagem de pilhas de madeira e ferramenta para diagnóstico de árvores urbanas

Abstract

The measurement and monitoring of forest resources depend largely on costly field methods that are sample-based and subject to the evaluator's subjectivity, whether in quantifying timber stocks or in diagnosing urban trees. In this context, computer vision—and, in particular, LiDAR (Light Detection and Ranging) technology, now available in portable, affordable devices—emerges as an alternative capable of replacing visual estimates with instrumental measurement. Accordingly, the objective of this study was to evaluate the application of computer vision in the forestry sector, using LiDAR in three contexts: mapping the state of the art, developing a method for scaling log piles, and characterizing urban trees for fall-risk analysis. The systematic review, conducted in the Web of Science and Scopus databases following the PRISMA protocol, identified 4,418 documents, of which 155 articles met the final eligibility criteria. A marked increase in publications was observed from 2019 onward, distributed across five themes: tree species classification (50 articles), forest monitoring (41), tree parameter measurement (26), tree segmentation and counting (20), and defect detection (18). Authorship was concentrated in China and the United States. A method was then developed for the volumetric estimation of Eucalyptus sp. log piles from point clouds obtained with an iPhone 14 Pro and a Hand-held Mobile Laser Scanner (HMLS). The method comprised noise removal, normalization, estimation of the face area by concave hull, and adaptive voxelization to calculate the stacking factor, with automation in a web application developed in the R language. Errors relative to the reference measurement ranged from −6.8% to +8.3%, all within the acceptable range. Finally, the HMLS was applied to the diagnosis of urban trees in a public square in the municipality of Piedade, São Paulo, through an adaptation of the TRAQ (Tree Risk Assessment Qualification) protocol. Of the 19 individuals evaluated, DBH ranged from 14.7 to 91.4 cm and the mean stem inclination was 5.24°, with only two individuals in the attention range. It is concluded that the use of portable LiDAR constitutes a viable, replicable, and low-cost alternative to conventional procedures, as it enables the assessment of timber stocks without exposing the operator to risks and objectifies the geometric parameters of arboreal risk analysis.