Search Results
Optimized Deep Learning Model as a Basis for Fast UAV Mapping of Weed Species in Winter Wheat Crops
Weed maps should be available quickly, reliably, and with high detail to be useful for site-specific management in crop protection and to promote more sustainable agriculture by reducing pesticide use. Here, the optimization of a deep residual convolutional neural network (ResNet-18) for the classification of weed and crop plants in UAV imagery is proposed. The target was to reach sufficient performance on an embedded system by maintaining the same features of the ResNet-18 model as a basis for fast UAV mapping. This would enable online recognition and subsequent mapping of weeds during UAV flying operation. Optimization was achieved mainly by avoiding redundant computations that arise when a classification model is applied on overlapping tiles in a larger input image. The model was trained and tested with imagery obtained from a UAV flight campaign at low altitude over a winter wheat field, and classification was performed on species level with the weed species Matricaria chamomilla L., Papaver rhoeas L., Veronica hederifolia L., and Viola arvensis ssp. arvensis observed in that field. The ResNet-18 model with the optimized image-level prediction pipeline reached a performance of 2.2 frames per second with an NVIDIA Jetson AGX Xavier on the full resolution UAV image, which would amount to about 1.78 ha h−1 area output for continuous field mapping. The overall accuracy for determining crop, soil, and weed species was 94%. There were some limitations in the detection of species unknown to the model. When shifting from 16-bit to 32-bit model precision, no improvement in classification accuracy was observed, but a strong decline in speed performance, especially when a higher number of filters was used in the ResNet-18 model. Future work should be directed towards the integration of the mapping process on UAV platforms, guiding UAVs autonomously for mapping purpose, and ensuring the transferability of the models to other crop fields.
IoT-Based Sensor Data Fusion for Determining Optimality Degrees of Microclimate Parameters in Commercial Greenhouse Production of Tomato
Optimum microclimate parameters, including air temperature (T), relative humidity (RH) and vapor pressure deficit (VPD) that are uniformly distributed inside greenhouse crop production systems are essential to prevent yield loss and fruit quality. The objective of this research was to determine the spatial and temporal variations in the microclimate data of a commercial greenhouse with tomato plants located in the mid-west of Iran. For this purpose, wireless sensor data fusion was incorporated with a membership function model called Optimality Degree (OptDeg) for real-time monitoring and dynamic assessment of T, RH and VPD in different light conditions and growth stages of tomato. This approach allows growers to have a simultaneous projection of raw data into a normalized index between 0 and 1. Custom-built hardware and software based on the concept of the Internet-of-Things, including Low-Power Wide-Area Network (LoRaWAN) transmitter nodes, a multi-channel LoRaWAN gateway and a web-based data monitoring dashboard were used for data collection, data processing and monitoring. The experimental approach consisted of the collection of meteorological data from the external environment by means of a weather station and via a grid of 20 wireless sensor nodes distributed in two horizontal planes at two different heights inside the greenhouse. Offline data processing for sensors calibration and model validation was carried in multiple MATLAB Simulink blocks. Preliminary results revealed a significant deviation of the microclimate parameters from optimal growth conditions for tomato cultivation due to the inaccurate timer-based heating and cooling control systems used in the greenhouse. The mean OptDeg of T, RH and VPD were 0.67, 0.94, 0.94 in January, 0.45, 0.36, 0.42 in June and 0.44, 0.0, 0.12 in July, respectively. An in-depth analysis of data revealed that averaged OptDeg values, as well as their spatial variations in the horizontal profile were closer to the plants’ comfort zone in the cold season as compared with those in the warm season. This was attributed to the use of heating systems in the cold season and the lack of automated cooling devices in the warm season. This study confirmed the applicability of using IoT sensors for real-time model-based assessment of greenhouse microclimate on a commercial scale. The presented IoT sensor node and the Simulink model provide growers with a better insight into interpreting crop growth environment. The outcome of this research contributes to the improvement of closed-field cultivation of tomato by providing an integrated decision-making framework that explores microclimate variation at different growth stages in the production season.
Measuring device for air speed in macroporous media and its application inside apple storage bins
In cold storage facilities of fruit and vegetables, airflow is necessary for heat removal. The design of storage facilities influences the air speed in the surrounding of the product. Therefore, knowledge about airflow next to the product is important to plan the layout of cold stores adapted to the requirements of the products. A new sensing device (ASL, Air speed logger) is developed for omnidirectional measurement of air speed between fruit or vegetables inside storage bins or in bulk. It consists of four interconnected plastic spheres with 80 mm diameter each, adapted to the size of apple fruit. In the free space between the spheres, silicon diodes are fixed for the airflow measurement based on a calorimetric principle. Battery and data logger are mounted inside the spheres. The device is calibrated in a wind tunnel in a measuring range of 0–1.3 m/s. Air speed measurements in fruit bulks on laboratory scale and in an industrial fruit store show air speeds in gaps between fruit with high stability at different airflow levels. Several devices can be placed between stored products for determination of the air speed distribution inside bulks or bin stacks in a storage room.
Weed Mapping with UAS Imagery and a Bag of Visual Words Based Image Classifier
Weed detection with aerial images is a great challenge to generate field maps for site-specific plant protection application. The requirements might be met with low altitude flights of unmanned aerial vehicles (UAV), to provide adequate ground resolutions for differentiating even single weeds accurately. The following study proposed and tested an image classifier based on a Bag of Visual Words (BoVW) framework for mapping weed species, using a small unmanned aircraft system (UAS) with a commercial camera on board, at low flying altitudes. The image classifier was trained with support vector machines after building a visual dictionary of local features from many collected UAS images. A window-based processing of the models was used for mapping the weed occurrences in the UAS imagery. The UAS flight campaign was carried out over a weed infested wheat field, and images were acquired between a 1 and 6 m flight altitude. From the UAS images, 25,452 weed plants were annotated on species level, along with wheat and soil as background classes for training and validation of the models. The results showed that the BoVW model allowed the discrimination of single plants with high accuracy for Matricaria recutita L. (88.60%), Papaver rhoeas L. (89.08%), Viola arvensis M. (87.93%), and winter wheat (94.09%), within the generated maps. Regarding site specific weed control, the classified UAS images would enable the selection of the right herbicide based on the distribution of the predicted weed species. © 2018 by the authors.
Monitoring Agronomic Parameters of Winter Wheat Crops with Low-Cost UAV Imagery
Monitoring the dynamics in wheat crops requires near-term observations with high spatial resolution due to the complex factors influencing wheat growth variability. We studied the prospects for monitoring the biophysical parameters and nitrogen status in wheat crops with low-cost imagery acquired from unmanned aerial vehicles (UAV) over an 11 ha field. Flight missions were conducted at approximately 50 m in altitude with a commercial copter and camera system—three missions were performed between booting and maturing of the wheat plants and one mission after tillage. Ultra-high resolution orthoimages of 1.2 cm·px−1 and surface models were generated for each mission from the standard red, green and blue (RGB) aerial images. The image variables were extracted from image tone and surface models, e.g., RGB ratios, crop coverage and plant height. During each mission, 20 plots within the wheat canopy with 1 × 1 m2 sample support were selected in the field, and the leaf area index, plant height, fresh and dry biomass and nitrogen concentrations were measured. From the generated UAV imagery, we were able to follow the changes in early senescence at the individual plant level in the wheat crops. Changes in the pattern of the wheat canopy varied drastically from one mission to the next, which supported the need for instantaneous observations, as delivered by UAV imagery. The correlations between the biophysical parameters and image variables were highly significant during each mission, and the regression models calculated with the principal components of the image variables yielded R2 values between 0.70 and 0.97. In contrast, the models of the nitrogen concentrations yielded low R2 values with the best model obtained at flowering (R2 = 0.65). The nitrogen nutrition index was calculated with an accuracy of 0.10 to 0.11 NNI for each mission. For all models, information about the surface models and image tone was important. We conclude that low-cost RGB UAV imagery will strongly aid farmers in observing biophysical characteristics, but it is limited for observing the nitrogen status within wheat crops.
Evaluating Soil-Borne Causes of Biomass Variability in Grassland by Remote and Proximal Sensing
On a grassland field with sandy soils in Northeast Germany (Brandenburg), vegetation indices from multi-spectral UAV-based remote sensing were used to predict grassland biomass productivity. These data were combined with soil pH value and apparent electrical conductivity (ECa) from on-the-go proximal sensing serving as indicators for soil-borne causes of grassland biomass variation. The field internal magnitude of spatial variability and hidden correlations between the variables of investigation were analyzed by means of geostatistics and boundary-line analysis to elucidate the influence of soil pH and ECa on the spatial distribution of biomass. Biomass and pH showed high spatial variability, which necessitates high resolution data acquisition of soil and plant properties. Moreover, boundary-line analysis showed grassland biomass maxima at pH values between 5.3 and 7.2 and ECa values between 3.5 and 17.5 mS m−1. After calibrating ECa to soil moisture, the ECa optimum was translated to a range of optimum soil moisture from 7% to 13%. This matches well with to the plant-available water content of the predominantly sandy soil as derived from its water retention curve. These results can be used in site-specific management decisions to improve grassland biomass productivity in low-yield regions of the field due to soil acidity or texture-related water scarcity.
The Use of a Pressure-Indicating Film to Determine the Effect of Liner Type on the Measured Teat Load Caused by a Collapsing Liner
During milking the teat cup liner is the interface between the teat of a dairy cow and the milking system, so it should be very well adapted to the teat. Therefore, the aim of the present study was to determine the effect of liner type on the directly measuring teat load caused by a collapsing liner with a pressure-indicating film. The Extreme Low pressure-indicating film was used to detect the effect of six different liners on teat load. For each liner, six positions in the teat cup were specified, and six repetitions were performed for each position with a new piece of film each time. Analysis of variance was performed to detect differences between the six liners, the positions within a liner, and the measuring areas. The pressure applied to the teat by a liner depends on the technical characteristics of the liner, especially the shape of the barrel, and for all tested liners, a higher teat load was found at the teat end. In conclusion, with the help of pressure-indicating film, it is possible to determine the different effects of liner type by directly measuring teat load due to liner collapse. © 2017 by the authors. Licensee MDPI, Basel, Switzerland.
Comparison of Calibration Approaches in Laser-Induced Breakdown Spectroscopy for Proximal Soil Sensing in Precision Agriculture
The lack of soil data, which are relevant, reliable, affordable, immediately available, and sufficiently detailed, is still a significant challenge in precision agriculture. A promising technology for the spatial assessment of the distribution of chemical elements within fields, without sample preparation is laser-induced breakdown spectroscopy (LIBS). Its advantages are contrasted by a strong matrix dependence of the LIBS signal which necessitates careful data evaluation. In this work, different calibration approaches for soil LIBS data are presented. The data were obtained from 139 soil samples collected on two neighboring agricultural fields in a quaternary landscape of northeast Germany with very variable soils. Reference analysis was carried out by inductively coupled plasma optical emission spectroscopy after wet digestion. The major nutrients Ca and Mg and the minor nutrient Fe were investigated. Three calibration strategies were compared. The first method was based on univariate calibration by standard addition using just one soil sample and applying the derived calibration model to the LIBS data of both fields. The second univariate model derived the calibration from the reference analytics of all samples from one field. The prediction is validated by LIBS data of the second field. The third method is a multivariate calibration approach based on partial least squares regression (PLSR). The LIBS spectra of the first field are used for training. Validation was carried out by 20-fold cross-validation using the LIBS data of the first field and independently on the second field data. The second univariate method yielded better calibration and prediction results compared to the first method, since matrix effects were better accounted for. PLSR did not strongly improve the prediction in comparison to the second univariate method.
Evaluation of different sensing approaches concerning to nondestructive estimation of leaf area index (LAI) for winter wheat
Different approaches of non-destructive estimation of the LAI in winter wheat were compared. Plant height had weak relation with the LAI, while estimated biomass showed high logarithmic relationship (R2=0.839). NDRE and REIP were logarithmically well related to the LAI (R2=0.726 and 0.779 respectively). Saturation effect of NDRE and REIP was less than NDVI. Some RGB-based indices also showed good potential to estimate the LAI. Among the indices, Gm, GMB, RMB, and NRMB were better related to the LAI. The results indicated that digital cameras can be used as an affordable and simple approach for assessment of the LAI of crops.
Growth Height Determination of Tree Walls for Precise Monitoring in Apple Fruit Production Using UAV Photogrammetry
In apple cultivation, spatial information about phenotypic characteristics of tree walls would be beneficial for precise orchard management. Unmanned aerial vehicles (UAVs) can collect 3D structural information of ground surface objects at high resolution in a cost-effective and versatile way by using photogrammetry. The aim of this study is to delineate tree wall height information in an apple orchard applying a low-altitude flight pattern specifically designed for UAVs. This flight pattern implies small distances between the camera sensor and the tree walls when the camera is positioned in an oblique view toward the trees. In this way, it is assured that the depicted tree crown wall area will be largely covered with a larger ground sampling distance than that recorded from a nadir perspective, especially regarding the lower crown sections. Overlapping oblique view images were used to estimate 3D point cloud models by applying structure-from-motion (SfM) methods to calculate tree wall heights from them. The resulting height models were compared with ground-based light detection and ranging (LiDAR) data as reference. It was shown that the tree wall profiles from the UAV point clouds were strongly correlated with the LiDAR point clouds of two years (2018: R2 = 0.83; 2019: R2 = 0.88). However, underestimation of tree wall heights was detected with mean deviations of −0.11 m and −0.18 m for 2018 and 2019, respectively. This is attributed to the weaknesses of the UAV point clouds in resolving the very fine shoots of apple trees. Therefore, the shown approach is suitable for precise orchard management, but it underestimated vertical tree wall expanses, and widened tree gaps need to be accounted for.