In recent years, the use of unmanned aerial vehicles (UAVs) to obtain high-resolution multispectral images has received increasing attention. When using reflectance data for quantitative remote sensing, radiometric calibration is needed that converts digital number to reflectance. Empirical linear radiometric calibration models with large calibration sites are feasible and convenient for remote sensing satellite wide-band sensors, but problems may exist with using this method directly in a UAV high-resolution remote sensing application. In this paper, we proposed a high precision, low cost subband empirical line (SEL) radiometric calibration method with two targets composed of black and white nonwoven fabric. To verify the results of the calibration model, analyses were performed by examining the influence of function selection, calibration materials, and the number of calibration panels. Next, the method was applied to natural field targets. The results show that the black and white panels have good spectral, Lambertian, and contrast characteristics; the power equation was more suitable for the visible bands, whereas the linear equation was more suitable for red-edge and near-infrared bands. The average absolute errors between the predicted reflectance and the measured reflectance were at around 10%. The proposed method had smaller errors and higher accuracy than the other function method, and a high correlation between the predicted and measured reflectance of the maize samples was observed. The SEL method developed in this study provides a reference for studying the radiometric calibration of other multispectral sensors.

Accurate estimation of forest volume is crucial for forest management, carbon budget balance and ecosystem monitoring. However, rapid, large-scale and high-accuracy acquisition of forest volume is still challenging. We proposed a method of coupling allometric growth model and multi-source data for forest volume estimation (CAMFVe). Firstly, the DBH estimation model is constructed by Terrestrial Laser Scanning (TLS) and Airborne Laser Scanning (ALS) to obtain more accurate measured volume. Secondly, the spectral attributes of Landsat and structural attributes of ALS are extracted and upscaled onto the 30 m plot scale, and the optimal attributes for volume estimation are selected. Thirdly, the model of CAMFVe is constructed and applied to obtain the volume of study area. Finally, the applicability of CAMFVe is evaluated under four forest growth environments (different canopy closure and slope categories), and the accuracy is compared to Multiple Linear Regression (MLR), Random Forest (RF), and Support Vector Machine (SVM). The results show that: (1) The DBH estimation model by TLS and ALS improves the DBH calculation accuracy of ALS with a 2.058 cm reduction in RMSE. (2) The mean of canopy height (Hmean) and Enhanced Vegetation Index (EVI) are identified as the optimal structural and spectral attributes, respectively. (3) The model constructed by Hmean and EVI consistently achieves higher accuracy for most forest growth environments, and the addition of spectral attribute improves volume estimation accuracy with a 10.152% reduction in RMSE compared to the Hmean-based model. (4) Compared with MLR, RF, and SVM, CAMFVe offers higher accuracy, requires fewer parameters, and is simpler and more efficient. Our proposed method, based on allometric growth model and utilizing vegetation index instead of DBH, provides a solution for large-scale and high-accuracy volume estimation by combining spaceborne LiDAR and optical satellite images.