Search

The accurate large-scale measurement of peach crowns is vital in horticultural science and the optimization of orchard management. Nowadays, numerous crown parameters (e.g., crown area, height, and volume) can be obtained via the analysis of point clouds or photographs. Current laser-based sensors provide the required reliable and accurate information; however, they are costly and time-consuming. Therefore, a simpler approach for crown measurement is required. For this purpose, this study presents a pipeline for the monitoring and clustering of 259 peach tree crowns based on unmanned aerial vehicle (UAV) images of a peach orchard in Southeast China. Considering the limitation that the original aerial image dataset contains little information, a data augmentation process is adopted, and an efficient deep learning architecture based on conditional generative adversarial networks (cGANs) was designed to extract the crown area. Then, the shape of the crown area was clustered using an edge detection process and a $k$ -means algorithm. Finally, an ellipsoid volume method (EVM) was applied to estimate the crown volume. Five indicators—namely, $Q_{\mathrm {seg}}$ , $S_{\mathrm {r}}$ , Precision, Recall, and F-measure—were employed to evaluate the crown extraction effects, and the average results for testing samples were 0.832, 0.847, 0.851, 0.828, and 0.846, respectively. Compared with other approaches—namely, fully convolutional network (FCN), U-Net, SegNet21, the excess green index (ExG), and the color index of vegetation extraction (CIVE)—the proposed cGAN model performs better, achieving an accuracy improvement of 5%–25%. For the estimation of crown volume, using measurements from a light detection and ranging (LIDAR) scanner as a reference, the correlation coefficient and relative-root-mean-square error (R-RMSE) were found to be 0.836% and 14.93%, respectively. Overall, the results demonstrate that the proposed method is feasible for measuring peach tree crowns. The wide application of such technology would facilitate applied research in plant phenotyping and precision horticulture. [ABSTRACT FROM AUTHOR]

Fast and accurate counting of wheat ears in field conditions is a key element for determining wheat yield. To obtain the number of wheat ears in a field, we propose a new counting algorithm based on computer vision. This algorithm counts wheat ears in remote images through semantic segmentation regression network (SSRNet). SSRNet is a multistage convolutional neural network that we propose to achieve counting problems through regression. In SSRNet, first, the original image is cropped to increase the amount of data. This method effectively solves the small sample dataset. Next, based on the cropping results, we build a fully convolutional neural network (FCNN) to segment wheat ears in field conditions. FCNN increases the accuracy of wheat ears counting by accurately segmenting wheat ears in a complex background. Then, we build a regression convolutional neural network (RCNN) to count wheat ears based on the segmentation results of FCNN. In RCNN, we propose a new activation function positive rectification linear unit (PrLU) to process the last layer of the fully connected layer, so that RCNN can effectively count the number of wheat ears in the image. Finally, a counting strategy is proposed to count the number of wheat ears in the original image. To verify the counting performance of SSRNet, we compare the counting result of SSRNet with the real value of manual statistics. The results show that the average accuracy (Acc), $R^{2}$ , and root mean squared error (RMSE) of the SSRNet count results on the test set in this article are 0.980, 0.996, and 9.437, respectively. It can be seen from the results that our proposed method can accurately count wheat ears in field conditions. At the same time, the counting time (0.11 s) shows that SSRNet can quickly estimate the number of wheat ears in field conditions. We concluded that this study can provide important technical support for the high-throughput field wheat ears counting task in large-scale phenotyping work. [ABSTRACT FROM AUTHOR]

Methods to obtain accurate phenotypic data of the seedling stage of maize are receiving ever-increasing research attention because such data are very important for crop growth and for estimating crop yield. To obtain such data, we propose herein an algorithm that uses computer vision to accurately recognize maize seedlings from a digital image. First, the red–green–blue (RGB) images acquired by a manned ground vehicle (MGV) and an unmanned aerial vehicle (UAV) are transformed into grayscale image to clarify the details of the images, and the Otsu threshold-segmentation method, which is based on threshold optimization, is used to separate the maize seedlings from the soil background. Next, the external pressure method is used to segment the results of skeleton extraction. After removing the skeleton burr by using the skeleton-deburring method based on saliency theory, the principal component analysis is used to determine the direction of the principal axis of the maize-seedling skeleton. The principal axis is then introduced as a reference to identify the direction angle of the principle axis in binary images. This process allows us to obtain from RGB images the multiple plant parameters that characterize the maize seedling stage. To verify these statistical computer-generated results, we compare them with field measurements of the maize plots. The result of applying the proposed algorithm to the MGV and UAV data sets correlate strongly ($R = 0.77$ –0.86) with the manually collected data. An average of 0.014 s is required to calculate the number of maize seedlings in a plot from two images showing different vegetation coverage, which shows that the proposed algorithm is computationally efficient. These results indicate that the proposed method provides accurate phenotypic data on maize seedlings. [ABSTRACT FROM AUTHOR]

Vertical heterogeneity of the canopy is being increasingly recognized in remote estimates of vegetative properties. Given the current limited knowledge of this issue, this paper investigated the effects of different vertical distributions of crop structure [e.g., leaf angle (LA) and leaf area index (LAI)] and biochemical parameters [e.g., chlorophyll a and b content (\textChl{a+{b}}) and water content (Wc)] on canopy reflectance and vegetation indices (VIs). A recently developed multiple-layer canopy reflectance model (MRTM) was tested for winter wheat and used to run a simulation analysis of different canopy scenarios. The results showed that the MRTM performed well to model winter wheat canopy reflectance with regard to spikes and vertical distributions of leaf properties. The vertical profiles of LA and LAI influenced canopy reflectance at almost all wavelengths, whereas the vertical profile of \textChl{a+{b}} mainly affected reflectance in the visible region, and that of Wc only affected reflectance in the near-infrared region. Changes in vertical distribution of the LA resulted in clear variations in VIs related to the LA, LAI, and \textChl{a+{b}} estimates. The vertical LAI and \textChl{a+{b}} profiles mainly influenced the VIs related to the LAI and \textChl{a+{b}} estimates. The Wc vertical profile primarily affected the VIs used to estimate crop water properties. The sensitivities of the VIs were mainly associated with the spectral responses and penetration characteristics of the bands they used. These findings suggest that the sensitivity of VIs to the vertical distributions of crop parameters should be considered when establishing models for remote crop monitoring. [ABSTRACT FROM PUBLISHER]

The total leaf chlorophyll content (TLCC) provides valuable information about the physiological status of crops. The objectives of this study were 1) to analyze the leaf area index (LAI) and soil factors that influences the estimation of TLCC using the PROSAIL model, which is a combination of the PROSPECT leaf model and the SAIL canopy model; 2) to propose newly combined spectral indices that reduce the influence of LAI and soil factors in order to improve the TLCC estimation; and 3) to test and validate the relationship between TLCC and the newly combined spectral indices. Ground-based hyperspectral data and concurrent TLCC parameters of samples were acquired at the Shihezi University Experiment Site, Xinjiang Province, China, during the 2009 and 2010 cotton growing seasons. The results showed that the newly combined spectral indices [double-peak canopy nitrogen index I (DCNI I), the ratio of the structure insensitive pigment index to the ratio vegetation index III (SIPI/RVI III), the ratio of the plant pigment ratio to the normalized difference vegetation index (PPR/NDVI), and the modified MERIS terrestrial chlorophyll index (MMTCI)] were more sensitive to chlorophyll and more resistant to LAI than the PPR, SIPI, and MERIS terrestrial chlorophyll index alone. In this study, DCNI I proved to be the best spectral index for estimating chlorophyll content, with determination coefficients (R^2) and root mean square error (RMSE) values of 0.80 and 8.31\;\mu\g\cdot cm^- 2, respectively. PPR/NDVI was also strongly correlated with chlorophyll content, with corresponding R^2 and RMSE values of 0.79 and 9.45\;\mu\g \cdot cm^-2, respectively. This study concluded that DCNI I and PPR/NDVI, in association with indices related to nitrogen, have good potential for assessing nitrogen content. [ABSTRACT FROM PUBLISHER]

The prevalence of powdery mildew (PM) in winter wheat field has a severe impact on crop production. An effective and timely forecast of the disease at a regional scale is necessary to control and prevent it. In this study, both meteorological and remotely sensed observations associated with crop characteristics and habitat traits were integrated for modeling the PM occurrence probability. With an effective feature selection procedure, four meteorological factors, including precipitation, temperature, sun radiation, humidity, and two remotely sensed features including reflectance of red band (R_R) demonstrate that the disease risk maps were able to depict the approximately spatial distribution of PM and its temporal dynamic in the study area. Compared with the model constructed with meteorological data only, the integrated model constructed with both remote sensing and meteorological data has produced a higher accuracy (increasing overall accuracy from 69% to 78%) of forecasting the PM occurrence. This suggests that there would be a great potential for predicting the PM occurrence probability by integrating both meteorological and remote sensing data at a regional scale. In the future, multiple forms of information (e.g., Web sensors networks data) are expected to be incorporated in the disease-forecasting model to further improve its performance for forecasting the disease occurrence (e.g., PM) at a regional scale. [ABSTRACT FROM PUBLISHER]

Data assimilation as an approach for crop Leaf Area Index (LAI) estimation has been rapidly developed in the field of agricultural remote sensing. Many studies have attempted to integrate sequential remotely sensed observations in the dynamical operation of physical models, aiming to improve model performance of LAI estimation by using various data assimilation schemes. In this study, a new data assimilation algorithm is proposed. For this algorithm, the background field of free parameters is constructed according to the ensemble construction scheme of EnKF algorithm, and a cost function is constructed based on the cost function construction strategy in 4DVAR algorithm in order to analyse and update the free parameters. The last updated free parameters are input into the model for LAI estimation until all the observations are assimilated. Additionally, the cost function in data assimilation procedure is optimised using VFSA algorithm. Winter wheat in Beijing in 2002 is selected as the experimental object. The crop growth model CERES-Wheat and the radiative transfer model PROSAIL are coupled in the assimilation process. Sequentially observed NDVI of winter wheat are assimilated into the coupled model using different assimilation algorithms to quantitatively compare and analyse the LAI estimation results. The performance of the new algorithm is better than CERES-Wheat model, EnKF algorithm, and 4DVAR algorithm, with a lesser RMSE by 43.68%, 41.67%, and 28.99%, respectively, and with increased R^2 by 110.53%, 90.48%, and 33.33%, respectively. Moreover, for the subset LAI \geq 3.00, higher estimation precision and greater efficiency of the new algorithm are obtained. [ABSTRACT FROM PUBLISHER]

This paper presents a new forest leaf area index (LAI) inversion method from multisource and multiangle data combined with a radiative transfer model and the strategy of k -means clustering and artificial neural network (ANN). Four scenes of Landsat-5 Thematic Mapper (L5TM) and Beijing-1 small satellite multispectral sensors (BJ1) images, acquired at different times, were selected to construct multisource and multiangle image data in this study. Considering a vertical distribution of forest LAI from both overstory and understory, a hybrid model of the invertible forest reflectance model (INFORM) was used to support the retrieval of forest LAI to eliminate the dependence of understory vegetation. The simulated data from INFORM outputs, added with a random noise, were first clustered by k -means method, and were then trained by ANN to obtain the inversion model for each group (cluster). Next, the inversion model was applied to the different combinations of multiangle data to retrieve the forest LAI. Finally, a validation of inverted results with Moderate Resolution Imaging Spectroradiometer LAI product and field measurements was conducted. The experimental results indicate that the accuracy of the inverted forest LAI can be improved through the addition of observation angle data, if the quality of the image data is ensured. The inversion accuracy of LAI with the multiangle image data is improved by 30% compared to the average accuracy of the inverted LAI with the single angle data after considering the addition of random noise to the ANN training data. [ABSTRACT FROM AUTHOR]