Goss Wilt is a common and serious disease during corn production. With a goal of automatic disease monitoring, this study assessed Goss's Wilt disease severity using machine (ML) and deep learning (DL) algorithms. A dataset containing 200 corn plot images was generated from an unmanned aerial vehicle (UAV) flying at five different mission heights (15, 30, 45, 60, and 75 m) above the ground level (AGL). Three different datasets including non-augmentation, segmentation and augmentation were prepared. The augmentation dataset consisting of 6200 images was prepared using geometric augmentation techniques, such as rotation, and flip. Eight different ML algorithms (i.e., Logistic Regression, Ada Boost, Gradient Boosting, Support Vector Machine, Multilayer Perceptron, Random Forest, Naive Bayes, K-Nearest Neighbors) and two different DL algorithms (i.e., GoogLeNet and ResNet18) were implanted to classify Goss Wilt severity as a binary issue (i.e., high and low). Two different types of features, including textural (contrast, dissimilarity, homogeneity, angular second moment) and color (hue, saturation, value, lightness, chromatic components: a* and b*, red, green, blue) features were extracted from individual plot image. For ML, the Random Forest yielded 0.99 precision, 0.99 recall and 0.99 F-score in augmented dataset and outperformed all other classifiers. For DL, Resnet18 achieved slightly better results: 0.81 precision, 0.78 recall and 0.79 F-score than GoogleNet, which has 0.75 precision, 0.70 recall, and 0.73 F-score. The ML model (Random Forest) performed satisfactorily by resulting in higher precision, recall and F-score in augmented dataset. However, ML models underperformed on segmentation dataset. Therefore, Random Forest coupled with UAV imagery is a potential valuable tool for automatic assessment of Goss Wilt disease.
The agricultural sector including industry and academia is facing an increasing number of challenges regarding farm management related to an increase in agricultural production with fewer resource usages. To seek measures for optimizing resource utilization and time needed to manage farm logistics, an agricultural routing planning technology is one of the integral tasks in farm management. In this study, an infield path planner with an optimal track sequence for autonomous land preparation in the paddy field, i.e. tillage and puddling, was evaluated in the rectangular paddy field. The test platform was a 60-kW auto-guided tractor equipped with a GNSS/INS and a navigation controller running a slip estimation-based path tracking algorithm. Field tests were conducted in rectangular paddy fields of 97x37 m2 and 97x36 m2 for tillage and puddling operation, respectively to validate the autonomous land preparation tractor. The evaluation parameters focused on the reduction of non-working distance and increase of field coverage rate, comparing the previously developed system. As a result of automatic tillage work, the proposed path planner showed superior performance in guiding the tractor along a headland turn with a 16% reduction in travel distance and covering the entire field, especially in the border corner area, from 91.5% to 99.1%. In the field test of the puddling task, the autonomous tractor successfully followed the path with RMSEs of lateral deviation < 9 cm and heading error < 1.1 deg in the presence of slippery condition with 98.4% coverage rate.
Abstract. Unmanned Aerial Vehicles (UAVs) are gaining widespread agricultural usage, including in applications like biomass monitoring, crop yield estimation, disease detection, and irrigation-timing estimation. Image data acquired during flights require some integration with ground control points (GCPs) to improve data accuracy. Typical GCPs require precise geographic coordinates based on expensive surveying equipment for image location of the visible ground marker. However, GPS surveys for typical GCPs are time-consuming, labor-intensive, and costly, especially when data collection is repeated at the same location multiple times in a season, in which case multiple GPS surveys are necessary. An autonomous GCP system based in part on a wireless network was developed for information collection and integration with fixed-wing UAV image data. The GCP system improves the speed of GCP setup and provides data collection advantages that have broad application in agricultural and environmental monitoring. The current GCPs are portable and equipped with an integrated controller, two low-cost GPS modules, a solar panel, a wireless module, and a storage battery, thus simplifying the capture of accurate image data in real-time with a fixed-wing UAV. Results indicate that the current GCPs can be used to improve accuracy in geo-referencing, radiometric calibration, and calibration of crop height estimates from surface models based on structure from motion.
Unmanned aerial spraying systems (UASSs) are widely used today for the effective control of pests affecting crops, and more advanced UASS techniques are now being developed. To evaluate such systems, artificial targets are typically used to assess droplet coverage through image processing. To evaluate performance accurately, high-quality binary image processing is necessary; however, this involves labor for sample collection, transportation, and storage, as well as the risk of potential contamination during the process. Therefore, rapid assessment in the field is essential. In the present study, we evaluated droplet coverage on water-sensitive papers (WSPs) under field conditions. A dataset was constructed consisting of paired training examples, each comprising source and target data. The source data were high-quality labeled images obtained from WSP samples through image processing, while the target data were aligned RoIs within field images captured in situ. Droplet coverage estimation was performed using an encoder–decoder model, trained on the labeled images, with features adapted to field images via self-supervised learning. The results indicate that the proposed method detected droplet coverage in field images with an error of less than 5%, demonstrating a strong correlation between measured and estimated values (R2 = 0.99). The method proposed in this paper enables immediate and accurate evaluation of the performance of UASSs in situ.
Abstract. Thermal remote sensing for the measurement of soil and crop surface temperatures has potential for various applications including the monitoring of crop stresses (like diseases and lack of soil moisture) and the planning of irrigation and harvesting. The use of unmanned aerial vehicles (UAVs) to acquire highly accurate thermal image data requires integration with thermal references on the ground. The primary objective of this paper was to demonstrate that it is possible to combine thermal remote sensing with UAV and temperature controlled ground references for calibrated crop temperature measurements. The references used for calibration of thermal images were two 61 cm square aluminum panels – one equipped with an integrated heating controller, thermal sensors, and thermoelectric modules to serve as a high-temperature reference, and the other equipped with an integrated cooling controller, thermal sensors, and coolers to serve as a low-temperature reference. To demonstrate the feasibility of using the calibration references in thermal images, three groups with three 61 cm square color tiles (light gray, medium gray, and dark gray) were distributed on the ground at different tilted angles (0 degrees, 25 degrees, and 50 degrees) to consider the effect that variations in crop temperature have on stem or leaf bending. Correlations between UAV-based thermal image estimates and ground truth were strong (R=0.98) on both un-calibration and calibration procedures. It is clear that a thermal calibration method based on ground temperature controlled references to improve UAV-based wheat temperature estimates was applied to a lower error by about 4%.
Soil moisture is an important factor determining yield. With the increasing demand for agricultural irrigation water resources, evaluating soil moisture in advance to create a reasonable irrigation schedule would help improve water resource utilization. This paper established a continuous system for collecting meteorological information and soil moisture data from a litchi orchard. With the acquired data, a time series model called Deep Long Short-Term Memory (Deep-LSTM) is proposed in this paper. The Deep-LSTM model has five layers with the fused time series data to predict the soil moisture of a litchi orchard in four different growth seasons. To optimize the data quality of the soil moisture sensor, the Symlet wavelet denoising algorithm was applied in the data preprocessing section. The threshold of the wavelets was determined based on the unbiased risk estimation method to obtain better sensor data that would help with the model learning. The results showed that the root mean square error (RMSE) values of the Deep-LSTM model were 0.36, 0.52, 0.32, and 0.48%, and the mean absolute percentage error (MAPE) values were 2.12, 2.35, 1.35, and 3.13%, respectively, in flowering, fruiting, autumn shoots, and flower bud differentiation stages. The determination coefficients (R2) were 0.94, 0.95, 0.93, and 0.94, respectively, in the four different stages. The results indicate that the proposed model was effective at predicting time series soil moisture data from a litchi orchard. This research was meaningful with regards to acquiring the soil moisture characteristics in advance and thereby providing a valuable reference for the litchi orchard’s irrigation schedule.
Unmanned aerial vehicle (UAV)-based aerial images have enabled a prediction of various factors that affect crop growth. However, the single UAV system leaves much to be desired; the time lag between images affects the accuracy of crop information, lowers the image registration quality and a maximum flight time of 20–25 min, and limits the mission coverage. A multiple UAV system developed from our previous study was used to resolve the problems centered on image registration, battery duration and to improve the accuracy of crop phenotyping. The system can generate flight routes, perform synchronous flying, and ensure capturing and safety protocol. Artificial paddy plants were used to evaluate the multiple UAV system based on leaf area index (LAI) and crop height measurements. The multiple UAV system exhibited lower error rates on average than the single UAV system, with 13.535% (without wind effects) and 17.729–19.693% (with wind effects) for LAI measurements and 5.714% (without wind effect) and 4.418% (with wind effects) for crop’s height measurements. Moreover, the multiple UAV system reduced the flight time by 66%, demonstrating its ability to overcome battery-related barriers. The developed multiple UAV collaborative system has enormous potential to improve crop growth monitoring by addressing long flight time and low-quality phenotyping issues.
'%3e%3cg id='b'%3e%3cpath id='a' stroke='%23000' stroke-width='2' d='M-6-25H6m-12 3H6m-12 3H6'/%3e%3cuse xlink:href='%23a' y='44'/%3e%3c/g%3e%3cpath stroke='%23fff' d='M0 17v10'/%3e%3ccircle r='12' fill='%23cd2e3a'/%3e%3cpath fill='%230047a0' d='M0-12A6 6 0 0 0 0 0a6 6 0 0 1 0 12 12 12 0 0 1 0-24z'/%3e%3c/g%3e%3cg transform='rotate(-123.69)'%3e%3cuse xlink:href='%23b'/%3e%3cpath stroke='%23fff' d='M0-23.5v3M0 17v3.5m0 3v3'/%3e%3c/g%3e%3c/svg%3e)
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)