Coffee ripeness monitoring using a UAV-mounted low-cost multispectral camera

Coffee beverage quality is highly correlated with the degree of fruit ripeness. In this sense, monitoring fruit ripeness is of utmost importance for harvest planning and, especially for obtaining high-quality beverages. Currently, this process is carried out through manual counts of unripe fruits, which is laborious and limited to a few plants within the field. This study aimed at evaluating the potential of a low-cost multispectral camera for coffee ripeness monitoring in the Zona da Mata region of Minas Gerais State, Brazil. For that, five fields of Arabica coffee with distinct characteristics were evaluated. During the coffee ripeness period, four flights were carried using a Phantom 4 Pro quadcopter equipped with a Mapir Survey 3W camera for imagery acquisition. After that, nine vegetation indices (VIs) were obtained. For the same dates, the percentage of unripe fruits was obtained using an irregular grid in all fields. The data was split into two ripeness classes: suitable for harvest (R) with?<?30% of unripe fruits; and not suitable for harvest (U), with?>?30% of unripe fruits. Then, a principal component analysis was used to infer the importance of the VIs to discriminate plants with unripe fruits from those with ripe fruits. The first two principal components explained?>?75% of the variance in the datasets from all coffee fields. The VIs were able to discriminate the ripeness classes (U and R) in most fields; however, their performance was directly influenced by the crop yield and canopy volume.

     

Evaluation of a UAV-mounted consumer grade camera with different spectral modifications and two handheld spectral sensors for rapeseed growth monitoring: performance and influencing factors

The objective of this study was to evaluate the crop monitoring performance of a consumer-grade camera with non-modified and modified spectral ranges which are commonly used in low-altitude unmanned aerial vehicle (UAV) platforms. The camera was fixed sequentially with seven types of filters for collecting visible images and near-infrared (NIR) images with different center band locations and bandwidths. Meanwhile, field-based hyperspectral data and normalized difference vegetation index (NDVI) measured by a GreenSeeker handheld crop sensor (GS-NDVI) were collected to examine the accuracy of rapeseed growth monitoring in terms of vegetation indices (VIs) derived from UAV images. Results showed that the UAV-based RGB-VIs and optimal NIR-VIs had similar accuracy for predicting GS-NDVI. Moreover, similar results were achieved based on the hyperspectral data, indicating the importance of spectral characteristics for GS-NDVI estimation. However, the UAV-based results also indicated that the performance of VIs derived from the band combinations containing longer NIR center wavelengths and narrower bandwidths was obviously poorer than that of the RGB-VIs. The image quality of the NIR band was also found to be inferior to the visible band based on quantitative analysis, which also revealed that image quality had great impact on UAV-based results. Image quality was then related to the effects of camera exposure, spectral sensitivity, soil background and dark areas. The results from this study provide useful information for camera modifications by selecting appropriate filters that not only are sensitive to crop growth, but also ensure image quality.

     

Relationships Between Yield Monitor Data and Airborne Multidate Multispectral Digital Imagery for Grain Sorghum

Remote sensing imagery taken during a growing season not only provides spatial and temporal information about crop growth conditions, but also is indicative of crop yield. The objective of this study was to evaluate the relationships between yield monitor data and airborne multidate multispectral digital imagery and to identify optimal time periods for image acquisition. Color-infrared (CIR) digital images were acquired from three grain sorghum fields on five different dates during the 1998 growing season. Yield data were also collected from these fields using a yield monitor. The images and the yield data were georeferenced to a common coordinate system. Four vegetation indices (two band ratios and two normalized differences) were derived from the green, red, and near-infrared (NIR) band images. The image data for the three bands and the four vegetation indices were aggregated to generate reduced-resolution images with a cell size equivalent to the combine's effective cutting width. Correlation analyses showed that grain yield was significantly related to the digital image data for each of the three bands and the four vegetation indices. Multiple regression analyses were also performed to relate grain yield to the three bands and to the three bands plus the four indices for each of the five dates. Images taken around peak vegetative development produced the best relationships with yield and explained approximately 63, 82, and 85% of yield variability for fields 1, 2, and 3, respectively. Yield maps generated from the image data using the regression equations agreed well with those from the yield monitor data. These results demonstrated that airborne digital imagery can be a very useful tool for determining yield patterns before harvest for precision agriculture.     

Unmanned aerial vehicle canopy reflectance data detects potassium deficiency and green peach aphid susceptibility in canola

There is growing evidence that potassium deficiency in crop plants increases their susceptibility to herbivorous arthropods. The ability to remotely detect potassium deficiency in plants would be advantageous in targeting arthropod sampling and spatially optimizing potassium fertilizer to reduce yield loss due to the arthropod infestations. Four potassium fertilizer regimes were established in field plots of canola, with soil and plant nutrient concentrations tested on three occasions: 69 (seedling), 96 (stem elongation), and 113 (early flowering) days after sowing (DAS). On these dates, unmanned aerial vehicle (UAV) multi-spectral images of each plot were acquired at 15 and 120 m above ground achieving spatial (pixel) resolutions of 8.1 and 65 mm, respectively. At 69 and 96 DAS, field plants were transported to a laboratory with controlled lighting and imaged with a 240-band (390–890 nm) hyperspectral camera. At 113 DAS, all plots had become naturally infested with green peach aphids (Hemiptera: Aphididae), and intensive aphid counts were conducted. Potassium deficiency caused significant: (1) increase in concentrations of nitrogen in youngest mature leaves, (2) increase in green peach aphid density, (3) decrease in vegetation cover, (4) decrease in normalized difference vegetation indices (NDVI) and decrease in canola seed yield. UAV imagery with 65 mm spatial resolution showed higher classification accuracy (72–100 %) than airborne imagery with 8 mm resolution (69–94 %), and bench top hyperspectral imagery acquired from field plants in laboratory conditions (78–88 %). When non-leaf pixels were removed from the UAV data, classification accuracies increased for 8 mm and 65 mm resolution images acquired 96 and 113 DAS. The study supports findings that UAV-acquired imagery has potential to identify regions containing nutrient deficiency and likely increased arthropod performance.     

An automated stand-alone in-field remote sensing system (SIRSS) for in-season crop monitoring

A stand-alone in field remote sensing system (SIRSS) with high spatial and temporal resolution was developed in this study. System control and image processing algorithms consisted of image acquisition control, camera parameter control, crop canopy reflectance calibration, image rectification, image background segmentation and vegetation indices map generation were developed and embedded in the SIRSS. The SIRSS is able to automatically capture multispectral images over a testing field at any predefined time points during the growing season and process captured images in real-time. This paper presents the SIRSS system design, image analysis procedures and determination of vegetation indices. In a validation experiment over an 8-plot corn field with three different nutrient treatments spanning the 2006 growing season, a total of 91 images were acquired and four different vegetation indices were derived from the images of each day. The largest differences of indices values among three treatments were indentified during the V6-V8 stages which implied this period could be the best time to detect variability caused by the nitrogen stress in the cornfield. The SIRSS has shown the potential of monitoring changes in vegetation status and condition.     

Vineyard yield estimation by combining remote sensing,computer vision and artificial neural network techniques

In viticulture, it is critical to predict productivity levels of the different vineyard zones to undertake appropriate cropping practices. To overcome this challenge, the final yield was predicted by combining vegetation indices (VIs) to sense the health status of the crop and by computer vision to obtain the vegetated fraction cover (F_c) as a measure of plant vigour. Multispectral imagery obtained from an unmanned aerial vehicle (UAV) is used to obtain VIs and F_c, which are used together with artificial neural networks (ANN) to model the relationship between VIs, F_c and yield. The proposed methodology was applied in a vineyard, where different irrigation and fertilisation doses were applied. The results showed that using computer vision techniques to differentiate between canopy and soil is necessary in precision viticulture to obtain accurate results. In addition, the combination of VIs (reflectance approach) and F_c (geometric approach) to predict vineyard yield results in higher accuracy (root mean square error (RMSE)?=?0.9 kg vine^?1 and relative error (RE)?=?21.8% for the image when close to harvest) compared to the simple use of VIs (RMSE?=?1.2 kg vine^?1 and RE?=?28.7%). The implementation of machine learning techniques resulted in much more accurate results than linear models (RMSE?=?0.5 kg vine^?1 and RE?=?12.1%). More precise yield predictions were obtained when images were taken close to the harvest date, although promising results were obtained at earlier stages. Given the perennial nature of grapevines and the multiple environmental and endogenous factors determining yield, seasonal calibration for yield prediction is required.

     

Using vegetation index and modified derivative for early detection of soybean plant injury from glyphosate

Glyphosate is a non-selective, systemic herbicide highly toxic to sensitive plant species. Its use has seen a significant increase due to the increased adoption of genetically modified glyphosate-resistant crops since the mid-1990s. Glyphosate application for weed control in glyphosate-resistant crops can drift onto an off-target area, causing unwanted injury to non-glyphosate resistant plants. Thus, early detection of crop injury from off-target drift of herbicide is critical in crop production. In non-glyphosate-resistant plants, glyphosate causes a reduction in chlorophyll content and metabolic disturbances. These subtle changes may be detectable by plant reflectance, which suggests the possibility of using optical remote sensing for early detection of drift damage to plants. In order to determine the feasibility of using optical remote sensing, a greenhouse study was initiated to measure the canopy reflectance of soybean plants using a portable hyperspectral image sensor. Non-glyphosate resistant soybean (Glycine max L. Merr.) plants were treated with glyphosate using a pneumatic track sprayer in a spray chamber. The three treatment groups were control (0 kg ae/ha), low dosage (0.086 kg ae/ha), and high dosage (0.86 kg ae/ha), each with four 2-plant pots. Hyperspectral images were taken at 4, 24, 48, and 72 h after application. The extracted canopy reflectance data was analyzed with vegetation indices. The results indicated that a number of vegetation indices could identify crop injury at 24 h after application, at which time visual inspection could not distinguish between glyphosate injured and non-treated plants. To improve the results a modified method of spectral derivative analysis was proposed and applied to find that the method produced better results than the vegetation indices. Four selected first derivatives at wavelength 519, 670, 685, and 697 nm could potentially differentiate crop injury at 4 h after treatment. The overall false positive rate was lower than the vegetation indices. Furthermore, the derivatives demonstrated the ability to separate treatment groups with different dosages. The study showed that hyperspectral imaging of plant canopy reflectance could be a useful tool for early detection of soybean crop injury from glyphosate, and that the modified spectral derivative analysis had a better performance than vegetation indices.     

Remote Sensed Spectral Imagery to Detect Late Blight in Field Tomatoes

Late blight, caused by the fungal pathogen Phytophthora infestans, is a disease that quickly spreads in tomato fields under suitable weather conditions and can threaten the sustainability of tomato farming in California, USA. This paper explores the applicability of remotely sensed images to detect disease spectral anomalies for precision disease management. We used the indices approach and generated a 5-index image that we used to identify the disease in tomato fields based on information from field-collected spectra and linear combinations of the spectral indices. Field results indicated that we were able to identify five clusters in the image space with small overlaps of a few clusters. Using the identified 5-cluster scheme to classify the tomato field images, we were able to successfully separate the diseased tomatoes from the healthy ones before economic damage was caused. Hence, the method based on a 5-index image may significantly enhance the capability of multispectral remote sensing for disease discrimination at the field level.     

RGB and multispectral UAV image fusion for Gramineae weed detection in rice fields

In this paper, a new method to fuse low resolution multispectral and high resolution RGB images is introduced, in order to detect Gramineae weed in rice fields with plants at 50 days after emergence (DAE).The images are taken from a fixed-wing unmanned aerial vehicle (UAV) at 60 and 70 m altitude. The proposed method combines the texture information given by a high resolution red–green–blue (RGB) image and the reflectance information given by a low resolution multispectral (MS) image, to obtain a fused RGB-MS image with better weed discrimination features. After analyzing the normalized difference vegetation index (NDVI) and normalized green red difference index (NGRDI) for weed detection, it was found that NGRDI presents better features. The fusion method consists of decomposing the RGB image using the intensity, hue and saturation (IHS) transformation, then, a second order Haar wavelet transformation is applied to the intensity layer (I) and the NGRDI image. From this transformation, the low–low (LL) coefficients of the NGRDI image are replaced by the LL coefficients of the I layer. Finally, the fused image is obtained by transforming the new wavelet coefficients to RGB space. To test the method, a one hectare experimental plot with rice plants at 50 DAE with Gramineae weeds was selected. Additionally, to compare the performance of the method, two indices were used, specifically, the M/M_GT index which is the percentage of detected weed area, and the MP index which indicates the precision of weed detection. These indices were evaluated in four validation zones using three Neural Networks (NN) detection systems based on three types of images; namely, RGB, RGB + NGRDI, and fused RGB-NGRDI. The best weed detection performance was obtained by the NN with the fused image, with M/M_GT index between 80 and 108% and MP between 70 and 85%.     

Enhanced broadband greenness in assessing Chlorophyll a and b,Carotenoid, and Nitrogen in Robusta coffee plantations using a digital camera

Direct-leaves measurement of spectral indices using a digital camera with a portable small chamber and custom illumination is used to take images of 600 leaves from 40 coffee plants. In this research, several vegetation indices (VIs) are developed and evaluated. Through a series of experiments, Chlorophyll a and b, Carotenoids, and Nitrogen critical level of Robusta coffee plants are analyzed and evaluated using selected VIs obtained from spectra of different tools like Spectrometer, digital camera, and SPAD-502 Chlorophyll meter. The actual Nitrogen critical level was determined using Kjeldahl laboratory test. Beside Hue, the newly proposed VIs could significantly improve the correlation in estimating photosynthetic pigments (Chlorophyll a and b, Carotenoids) and Nitrogen critical level of Robusta coffee plant. Finally, consumer-grade digital camera with custom chamber is shown to be used for rapid and accurate in situ estimation of Chlorophyll a and b, Carotenoids, and Nitrogen critical level of Robusta coffee plant from direct-leaves measurement.     

Spatially discriminating Russian wheat aphid induced plant stress from other wheat stressing factors

The Russian wheat aphid (RWA) Diuraphis noxia (Mordvilko) is a major pest of winter wheat and barley in the United States. RWA induces stress to the wheat crop by damaging plant foliage, lowering the greenness of plants, and affecting productivity. The utilization of multispectral remote sensing is effective at detecting plant stress in agricultural crops. Stress to wheat plants detected in fields can be caused by several factors that can vary spatially in their presence and intensity across a field. Stress can result from factors such as nutrient deficiency, drought, diseases, and pests that can occur individually or collectively. The present study investigated the potential of using spatial pattern metrics derived from multispectral images in combination with topographic and edaphic variables to identify a set of variables to differentiate the stress induced by RWA from other stress causing factors. A discriminant function analysis was applied to 15 discriminating variables. A set of 13 variables were retained to develop a model to differentiate the three types of stress. Overall, 97 percent of patches of stress used to validate the model were correctly categorized. Stressed patches caused by RWA were 98 percent correctly classified, patches caused by drought were 94 percent correctly classified, and patches caused by agronomic conditions were 99 correctly classified. It is possible to discriminate stress induced by RWA from other stress causing factors in multispectral data when spatial attributes of the stress causing factors are incorporated in the analysis.     

Weed detection by UAV: simulation of the impact of spectral mixing in multispectral images

This study aimed to assess the spectral information potential of images captured with an unmanned aerial vehicle, in the context of crop–weed discrimination. A model is proposed in which the entire image acquisition chain is simulated in order to compute the digital values of image pixels according to several parameters (light, plant characteristics, optical filters, sensors…) to reproduce in-field acquisition conditions. The spectral mixings in the pixels are modeled, based on an image with a 60 mm spatial resolution, to estimate the impact of the resolution on the ability to discriminate small plants. The classification potential (i.e. the ability to separate two classes) in soil and vegetation and in monocotyledon and dicotyledon classes is studied using simulations for different vegetation rates (defined as the proportion of vegetation covering the surface projected in the considered pixel). The classification is unsupervised and based on the Mahalanobis distance computation. The results of soil-vegetation discrimination show that pixels with low vegetation rates can be classified as vegetation: pixels with vegetation rate greater than 0.5 had a probability to be correctly classified between 80 and 100%. Classification between monocotyledonous and dicotyledonous plants requires pixels with a high vegetation rate: to obtain a probability to be correctly classified better than 80%, vegetation rates in the pixels have to be over 0.9. To compare the results with data from real images, the same classification was tested on multispectral images of a weed infested field. The comparison confirmed the ability of the model to assess vegetation–soil and crop–weed discrimination potential for specific sensors (such as the multiSPEC 4C sensor, AIRINOV, Paris, France), where the acquisition chain parameters can be tested.     

Determination of differences in crop injury from aerial application of glyphosate using vegetation indices

Crop injury caused by off-target drift of herbicide can seriously reduce growth and yield and is of great concern to farmers and aerial applicators. Farmers can benefit from identifying an indirect method for assessing the level of crop injury. This study evaluates the combined use of statistical methods and vegetation indices (VIs) derived from multispectral images to assess the level of crop injury. An experiment was conducted in 2009 to determine glyphosate injury differences among the cotton, corn, and soybean crops. The crops were planted in eight rows spaced 102 cm apart and 80 m long with four replications. Seven VIs were calculated from multispectral images collected at 7 and 21 days after the glyphosate application (DAA). At each image collection date, visual injury estimates were assessed and data were collected for plant height, chlorophyll content, and shoot dry weight. From the seven VIs evaluated as surrogate for glyphosate injury identification using a canonical correlation analysis (CCA), the Chlorophyll Vegetation Index (CVI) showed the highest correlation with field-measured plant injury data. CVI image values were subtracted from the CVI average values of the non-injured area to generate CVI residual images (CVI_res). Frequency distribution histograms of CVI_res image values were calculated to assess the level of injury between crops. These data suggested that injury increased from 7DAA to 21DAA with corn exhibiting higher severity of injury than cotton or soybean, while only moderate injury was observed for cotton. The techniques evaluated in this study are promising for estimating the level of glyphosate herbicide drift, which can be used to make appropriate management decisions considering crop proximity.     

Detection of target spot and bacterial spot diseases in tomato using UAV-based and benchtop-based hyperspectral imaging techniques

Early and accurate diagnosis is a critical first step in mitigating losses caused by plant diseases. An incorrect diagnosis can lead to improper management decisions, such as selection of the wrong chemical application that could potentially result in further reduced crop health and yield. In tomato, initial disease symptoms may be similar even if caused by different pathogens, for example early lesions of target spot (TS) caused by the fungus Corynespora cassicola and bacterial spot (BS) caused by Xanthomonas perforans. In this study, hyperspectral imaging (380–1020 nm) was utilized in laboratory and field (collected by an unmanned aerial vehicle; UAV) settings to detect both diseases. Tomato leaves were classified into four categories: healthy, asymptomatic, early and late disease development stages. Thirty-five spectral vegetation indices (VIs) were calculated to select an optimum set of indices for disease detection and identification. Two classification methods were utilized: (i) multilayer perceptron neural network (MLP), and (ii) stepwise discriminant analysis (STDA). Best wavebands selection was considered in blue (408–420 nm), red (630–650 nm) and red edge (730–750 nm). The most significant VIs that could distinguish between healthy leaves and diseased leaves were the photochemical reflectance index (PRI) for both diseases, the normalized difference vegetation index (NDVI850) for BS in all stages, and the triangular vegetation index (TVI), NDVI850 and chlorophyll index green (Chl green) for TS asymptomatic, TS early and TS late disease stage respectively. The MLP classification method had an accuracy of 99%, for both BS and TS, under field (UAV-based) and laboratory conditions.

     

Detection of bacterial wilt infection caused by Ralstonia solanacearum in potato (Solanum tuberosum L.) through multifractal analysis applied to remotely sensed data

Potato bacterial wilt, caused by the bacterium Ralstonia solanacearum race 3 biovar 2 (R3bv2), affects potato production in several regions in the world. The disease becomes visually detectable when extensive damage to the crop has already occurred. Two greenhouse experiments were conducted to test the capability of a remote sensing diagnostic method supported by multispectral and multifractal analyses of the light reflectance signal, to detect physiological and morphological changes in plants caused by the infection. The analysis was carried out using the Wavelet Transform Modulus Maxima (WTMM) combined with the Multifractal (MF) analysis to assess the variability of high-resolution temporal and spatial signals and the conservative properties of the processes across temporal and spatial scales. The multispectral signal, enhanced by multifractal analysis, detected both symptomatic and latently infected plants, matching the results of ELISA laboratory assessment in 100 and 82%, respectively. Although the multispectral method provided no earlier detection than the visual assessment on symptomatic plants, the former was able to detect asymptomatic latent infection, showing a great potential as a monitoring tool for the control of bacterial wilt in potato crops. Applied to precision agriculture, this capability of the remote sensing diagnostic methodology would provide a more efficient control of the disease through an early and full spatial assessment of the health status of the crop and the prevention of spreading the disease.     

Evaluating high resolution SPOT 5 satellite imagery to estimate crop yield

High resolution satellite imagery has the potential to map within-field variation in crop growth and yield. This study examined SPOT 5 satellite multispectral imagery for estimating grain sorghum yield. A 60 km × 60 km SPOT 5 scene and yield monitor data from three grain sorghum fields were recorded in south Texas. The satellite scene contained four spectral bands (green, red, near-infrared and mid-infrared) with a 10-m spatial resolution. Subsets were extracted from the scene that covered the three fields. Images with pixel sizes of 20 and 30 m were also generated from the individual field images to simulate coarser resolution satellite imagery. Vegetation indices and principal components were derived from the images at the three spatial resolutions. Grain yield was related to the vegetation indices, the four bands and the principal components for each field, and for all the fields combined. The effect of the mid-infrared band on estimates of yield was examined by comparing the regression results from all four bands with those from the other three bands. Statistical analysis showed that the 10-m, four-band image and the aggregated 20-m and 30-m images explained 68, 76 and 83%, respectively, of the variation in yield for all the fields combined. The coefficient of determination between yield and the imagery increased with pixel size because of the smoothing effect. The inclusion of the mid-infrared band slightly improved the R ² values. These results indicate that high resolution SPOT 5 multispectral imagery can be a useful data source for determining within-field yield variation for crop management.     

Phenotyping a diversity panel of quinoa using UAV-retrieved leaf area index,SPAD-based chlorophyll and a random forest approach

Given its high nutritional value and capacity to grow in harsh environments, quinoa has significant potential to address a range of food security concerns. Monitoring the development of phenotypic traits during field trials can provide insights into the varieties best suited to specific environmental conditions and management strategies. Unmanned aerial vehicles (UAVs) provide a promising means for phenotyping and offer the potential for new insights into relative plant performance. During a field trial exploring 141 quinoa accessions, a UAV-based multispectral camera was deployed to retrieve leaf area index (LAI) and SPAD-based chlorophyll across 378 control and 378 saline-irrigated plots using a random forest regression approach based on both individual spectral bands and 25 different vegetation indices (VIs) derived from the multispectral imagery. Results show that most VIs had stronger correlation with the LAI and SPAD-based chlorophyll measurements than individual bands. VIs including the red-edge band had high importance in SPAD-based chlorophyll predictions, while VIs including the near infrared band (but not the red-edge band) improved LAI prediction models. When applied to individual treatments (i.e. control or saline), the models trained using all data (i.e. both control and saline data) achieved high mapping accuracies for LAI (R²?=?0.977–0.980, RMSE?=?0.119–0.167) and SPAD-based chlorophyll (R²?=?0.983–0.986, RMSE?=?2.535–2.861). Overall, the study demonstrated that UAV-based remote sensing is not only useful for retrieving important phenotypic traits of quinoa, but that machine learning models trained on all available measurements can provide robust predictions for abiotic stress experiments.

     

Separability of coffee leaf rust infection levels with machine learning methods at Sentinel-2 MSI spectral resolutions

Coffee leaf rust (CLR) caused by the fungus Hemileia vastarix is a devastating disease in almost all coffee producing countries and remote sensing approaches have the potential to monitor the disease. This study evaluated the potential of Sentinel-2 band settings for discriminating CLR infection levels at leaf levels. Field spectra were resampled to the band settings of the Sentinel-2, and evaluated using the random forest (RF) and partial least squares discriminant analysis (PLS-DA) algorithms with and without variable optimization. Using all variables, Sentinel-2 Multispectral Imager (MSI)-derived vegetation indices achieved higher overall accuracy of 76.2% when compared to 69.8% obtained using raw spectral bands. Using the RF out-of-bag (OOB) scores, 4 spectral bands and 7 vegetation indices were identified as important variables in CLR discrimination. Using the PLS-DA Variable Importance in Projection (VIP) score, 3 Sentinel-2 spectral bands (B4, B6 and B5) and 5 vegetation indices were found to be important variables. Use of the identified variables improved the CLR discrimination accuracies to 79.4 and 82.5% for spectral bands and indices respectively when discriminated with the RF. Discrimination accuracy slightly increased through variable optimization for PLS-DA using spectral bands (63.5%) and vegetation indices (71.4%). Overall, this study showed the potential of the Sentinel 2 MSI band settings for CLR discrimination as part of crop condition assessment. Nevertheless further studies are required under field conditions.     

A new color index for vegetation segmentation and classification

Color vegetation indices enable various precision agriculture applications by transforming a 3D-color image into its 1D-grayscale counterpart, such that the color of vegetation pixels can be accentuated, while those of nonvegetation pixels are attenuated. The quality of the transformation is essential to the outcomes of computational analyses to follow. The objective of this article is to propose a new vegetation index, the Elliptical Color Index (ECI), which leverages the quadratic discriminant analysis of 3D-color images along a normalized red (r)—green (g) plane. The proposed index is defined as an ellipse function of r and g variables with a shape parameter. For comparison, the ECI’s performance was evaluated along with six other indices, by using 240 color images as a test sample captured from four vegetation species under different illumination and background conditions, together with the corresponding ground-truth patterns. For comparative analysis, the receiver operating characteristic (ROC) and the precision–recall (PR) curves helped quantify the overall performance of vegetation segmentation across all of the vegetation indices evaluated. For a practical appraisal of vegetation segmentation outcomes, this paper applied Gaussian filtering, and then the thresholding method of Otsu, to the grayscale images transformed by each of the indices. Overall, the test results confirmed that ECI outperforms the other indices, in terms of the area under the curves of ROC and PR, as well as other performance metrics, including total error, precision, and F-score.

     

水稻多时相植被指数特征及覆盖度提取研究

为准确快速获取水稻的植被指数特征和植被覆盖度信息,利用无人机采集水稻分蘖期、抽穗期和结实期的多光谱影像数据,选择不同类型的植被指数,利用样本统计法和植被指数交点法,提取并探究水稻3个生长期在地块和像元尺度下的植被指数特征,并运用阈值分割法提取水稻植被信息及覆盖度信息。结果表明,水稻3个生长期内,在像元和地块尺度下均表现出明显的物候特征,且与杂草和树木存在明显区别;多光谱植被指数的植被覆盖度提取精度整体高于可见光植被指数;归一化植被指数（normalized difference vegetation index,NDVI）对水稻3个时期植被覆盖度提取精度最高,提取误差分别为0.40%、0.43%和0.81%,R²为0.77、0.92和0.98,均方根误差（root mean square error,RMSE）为9.09%、2.97%和0.38%;可见光波段差异植被指数（visible-band difference vegetation index,VDVI）提取精度高于超绿红蓝差分指数（excess green-red-blue difference index,EGEBDI）和过绿减过红指数（excess green-excess red index,ExG-ExR）,提取误差分别为4.30%、1.36%和1.60%,R²分别为0.53、0.77和0.80,RMSE分别为14.62%、3.70%和5.50%。该研究成果可为作物长势监测及其植被覆盖度提取提供技术支撑。