. 2023 Feb 24;9(3):57. doi: 10.3390/jimaging9030057

Table 5.

Comparison of LiDAR data in weed, crop and soil detection and crop growth estimation.

LiDAR Type	Analytical Method	Model	Features	Results
Terrestrial	SLR; binary logistic regression; CDA		Plant height; reflection value	R² between LiDAR measured height and actual plant heights was 0.75. The predicted values from binary logistic regression shows an accuracy of 95.3% for vegetation and 82.2% for non-vegetation/soil, with an overall accuracy of 92.7%. Using canonical discriminant analysis (CDA), the overall success to discriminate was 72.2%. The soil and dicots were classified with 92.4% and 64.5% accuracy, respectively [40].
Terrestrial		CropPointNet; PointNet; DGCNN	Crop height	CropPointNet model had an overall accuracy of 81.5%. PointNet and DGCNN had overall accuracies of 55% and 66.5%, respectively. CropPointNet, DGCNN and PointNet models discriminated cabbage with 91%, 82% and 72% accuracy, respectively, eggplant with 88%, 83% and 69% accuracy, respectively, and tomato crop with 65%, 61% and 60% accuracy, respectively [41].
Terrestrial	SLR		Canopy height	R² between LiDAR measured height and manual measurement, between UAS measured height and manual measurement and between ultrasonic-sensor-measured height and manual measurement were 0.97, 0.91 and 0.05, respectively [42].
Airborne	Power regression; SLR	DTM	Sugarcane height	The ratio of ground to non-ground returns with LiDAR had R² = 0.971. The ratio of ground to non-ground returns with photogrammetry had R² = 0.993. R² between maximum crop height obtained from LiDAR and those obtained from photogrammetry was 0.885. R² between the mean crop height obtained from LiDAR and those obtained from photogrammetry was 0.929 [43].
Airborne	MLR; SMR; GLM; GBM; KRLS; RFR	DEM; DSM	AFW	R² between observed AFW and fitted AFW via RFR was 0.96, the highest value for R² among the six models [44].