Terrestrial Videogrammetry for Deriving Key Forest Inventory Data: A Case Study in Plantation

Abstract

Computer vision technology has promoted the rapid development of forest observation equipment, and video photography (videogrammetry) has provided new ideas and means for forestry investigation. According to the characteristics of videogrammetry, a spiral observation method is proposed. Meanwhile, a new point cloud data processing method is proposed, which extracts a point cloud at the diameter at breast height (DBH) section and determines the DBH of trees through cylinder fitting and circle fitting, according to the characteristics of the point cloud model and the real situation of occlusion in the sampled area, and then calculates the biomass. Through a large number of experiments, a more effective and relatively high-precision method for DBH extraction is obtained. Compared with the field survey data, the bias% of DBH extracted by videogrammetry was −3.19~2.87%, and the RMSE% was 5.52~7.76%. Compared with the TLS data, the bias% of −4.78~2.38%, and the RMSE% was 5.63~9.87%. The above-ground biomass (AGB) estimates from the videogrammetry showed strong agreement with the reference values with concordance correlation coefficient (CCC) and the RMSE values of 0.97 and 19.8 kg. Meanwhile, the AGB estimate from TLS agrees with the CCC values and the RMSE of 0.97 and 17.23 kg. Videogrammetry is not only cheap, low cost, and fast, but also can be observed in a relatively complex forest environment, with strong anti-interference ability. The experimental results prove that its accuracy is comparable to TLS and photogrammetry. Thus this work is quite valuable in a forest resources survey. We believe that the calculation accuracy of our new method can fully meet the needs of the forest survey.