Greenhouse Mapping and Crop Type Classification for Small-Scale Farms Using Airborne Laser Scanning

The increasing frequency of extreme climatic events has resulted in significant crop losses, prompting many farmers to adopt greenhouses as a climate adaptation strategy. Greenhouses, constructed with transparent materials to allow sunlight penetration, are widely used for high-quality vegetable and fruit cultivation. These structures have both upper and lower layers; airborne laser scanning can penetrate the upper plastic layer and detect crop in the lower layer. This study analyzed 222 small-scale greenhouses in Taiwan, covering a total area of 1,443 ha. ALS data were used to derive four indices: normalized digital surface model, first echo intensity, laser penetration index, and surface roughness. These indices were used to classify greenhouse areas using a support vector machine, achieving an overall accuracy of 87.32% and an F1-score of 0.93. A cloth simulation filter was then applied to separate point data into upper and lower layers, enabling the removal of upper-layer points. Greenhouse area crops were further classified into bare ground, tall crops, low-lying crops, and mixed crops, with 199 greenhouse areas correctly identified and an overall accuracy of 92.56%. The F1-scores for each crop class ranged from 0.87 to 0.97. This method accurately reflected actual cultivation conditions within the GAs.