查看更多>>摘要：? 2022 Elsevier B.V.High-precision positioning of agricultural robots is the key to the automation of greenhouse agricultural production. We propose a visual positioning method by leveraging the fiducial markers and factor graph to achieve reliable performance. This method consists of two parts, a front-end module and a back-end module. First, fiducial markers are involved to the front-end module to tackle the problems of unstructured and insufficient features in the greenhouse environment. A tag detection algorithm AprilTag2 is used to identify markers and extract feature points for the positioning purpose. Second, to eliminate the adverse effects of camera motion, mechanical vibration, and environmental disturbance on the positioning accuracy, the back-end module considers the characteristics of robot motion and the constraint relationship of variables, and correspondingly designs a positioning model based on the factor graph. The model combines five factors for maximum a posteriori estimation and incremental optimization of the state of the robot in real time. Finally, the positioning experiment of the wheeled robot was carried out at the motion speed of 0.26 m/s, 0.39 m/s, and 0.52 m/s. The experimental results show that the average errors of positioning are 0.056 m, 0.065 m, and 0.081 m, respectively, and the standard deviations of the errors are lower than 0.05 m. The visual positioning method based on fiducial markers and factor graph outperforms the mainstream positioning techniques in terms of positioning accuracy and robustness, and meet the high-precision positioning requirements in complex agricultural fields.
查看更多>>摘要：? 2022Firmness and weight are critical quality attributes of peaches. Non-destructive detection of peach firmness and weight is helpful to meet the market demand for high-quality peaches. In this study, the possibility of simultaneous prediction of peach firmness and weight based on fruit vibration spectra combined with a modified one-dimensional convolutional neural network (CNNm) was investigated. The vibration spectra of 216 peaches were measured by a laser Doppler vibrometer (LDV). The CNNm adopted an inception module with three parallel filters of different sizes so that it can abstract depth features at multiple scales from the full vibration spectra and has the potential to learn more useful information to achieve high prediction performance. The performance of CNNm for predicting peach firmness and weight was compared with that of a typical CNN (CNNt) and two classical chemometric methods based on both full vibration spectra and selected effective frequencies including partial least square (PLS), support vector regression (SVR), successive projections-PLS (SPA-PLS) and SPA-SVR. The results demonstrated that the CNN-based models (CNNm and CNNt) outperformed the classical chemometric models (PLS, SVR, SPA-PLS and SPA-SVR). And CNNm achieved the best performance with RP2 = 0.844, RMSEP = 0.429 N/mm and RPDP = 2.554 for peach firmness prediction and RP2 = 0.794, RMSEP = 29.954 g and RPDP = 2.223 for peach weight prediction. The preliminary results indicated that both peach firmness and weight could be accessed by the vibration spectra of fruits combined with CNNm. The proposed method provided a potential means for simultaneous online prediction of peach firmness and weight.
查看更多>>摘要：? 2022 Elsevier B.V.It is very important to monitor the seeding quantity in the precision seeding process. However, there is still room for improvement in monitoring accuracy of seeding quantity. One of the important questions is that double overlapping seeds cannot be accurately identified. They are usually counted as single seeds. The sensor signal acquisition and judgment algorithm play a key role in seed recognition and counting. This study mainly focused on the discussion and analysis of signal acquisition and judgment algorithms. Based on the ADC (Analog to Digital Converter) voltage signal acquisition, an analog signal seed sensor has been designed. The peak value judgment algorithm and the mean value judgment algorithm are compared and analyzed. At the same time, a signal output quantity (SOQ) judgment algorithm and calculation model based on ADC signal acquisition have been established. Different from the peak value judgment algorithm and the mean value judgment algorithm, the SOQ judgment algorithm did not focus on the magnitude of the acquired voltage values, but on the quantity of acquired voltage values. The test results showed that the monitoring accuracy of the judgment algorithm and calculation model for single and double overlapping model seeds had reached 100%. In order to test the monitoring accuracy of the judgment algorithm and calculation model for real seeds, maize seeds and mung bean seeds were selected for testing. The results showed that under the condition that the recognition accuracy of single maize seed was 100%, the average recognition accuracy of the double overlapping maize seeds was 91%, and the average monitoring accuracy was 95.5%. And under the condition that the recognition accuracy of single mung bean seed was 100%, the average recognition accuracy of the double overlapping mung bean seeds was 83%, and the average monitoring accuracy was 91.5%. In terms of the monitoring accuracy of double overlapping seeds, the SOQ judgment algorithm has been greatly improved compared with the peak judgment algorithm and the mean judgment algorithm. This provided a new idea for the monitoring of the seeding quantity, and it was also of great help to the improvement of the statistical accuracy of the seeding quantity.
Pierre Pott L.Jorge Carneiro Amado T.Augusto Schwalbert R.Mateus Corassa G....
15页
查看更多>>摘要：? 2022 Elsevier B.V.Crop type mapping is essential for agricultural monitoring, but lack of terrestrial labels (herein termed as ground truth data) limit these models around the globe. In this study, we tested different methods for creating crop type maps for soybean (Glycine max L.) and corn (Zea mays L.), with aims of i) compare ground truth data related to crop modeling; ii) evaluate agricultural field masks generated by Environmental Rural Register, MapBiomas product, and random forest model; iii) evaluate models of a) unsupervised classification, b) supervised classification with ground truth data of 1 year, c) with ground truth data of 2 years, d) supervised classification with crop modeling only, e) with the combination of crop modeling and ground truth data of 1 year, and f) with ground truth data of 2 years; iv) test sample size for training the model utilizing ground truth data and crop modeling in transfer learning approach; and v) examine spatial remote sensing features to guide crop data collection. The APSIM-NG crop model was utilized for generate crop modeling simulations to compare with satellite images harmonic regression of ground truth data. We found similarity of harmonic regression coefficients derived from crop modeling and satellite imagery of Sentinel-2 of the labeled ground truth data for both soybean and corn crops. Agricultural masks showed efficiency for crop area estimation for soybean, with highest accuracy with random forest model. Crop modeling aggregated with growing season data as input for supervised learning presented the greater model performance with overall accuracy of 0.94. Crop area prediction was most accurate for soybean [R2 = 0.93, and mean absolute error (MAE) = 2,052 ha] for the model with the combination of crop modeling and ground truth data of 2 years, and least for corn, [R2 = 0.18, and MAE = 1,146 ha] when only using crop modeling. Model performance was influenced by sample size, with greater accuracy (0.93) for aggregating crop modeling (150 samples) and year data (100 samples). In addition, considering spatial field data variability as model input increased overall accuracy from 0.84 to 0.93, with higher impact when only ground truth data was utilized for input in the model. Our results suggest that these methods offer options for crop type classification when less adequate ground truth data is available and with unsupervised learning. On the other hand, supervised learning utilizing crop modeling with presence of field data did improve model performance overall.
查看更多>>摘要：? 2022 Elsevier B.V.Measuring the ripeness of fruits is one of the critical factors in achieving real-time quality control and sorting of fruit by growers and postharvest managers. However, recent tactile sensing approaches for fruit ripeness detection have suffered setbacks due to: (1) the nonlinear relationship between the sensor output and the true stiffness of fruits; and (2) the angle of contact, referred to as the inclination angle, between the sensor and the outer surface of the fruit. In this paper, we propose a non-destructive tactile sensing approach for estimating the stiffness of fruits, using kiwifruit as a case study. Our sensor configuration is based on a three-probe piezoresistive cantilever beam, allowing us to obtain relatively stable sensor outputs that are independent of the inclination angle of the fruit surface. Our stiffness estimation approach is based on the combination of instantaneous sensor outputs with 63 regression-based machine learning models comprising of neural networks, Gaussian process, support vector machines, and decision trees. For experiments, we used several kiwifruit samples at diverse ripeness levels. The extracted sensor data was used to train the learning models over a 10-fold cross-validation technique, allowing us to find the nonlinear relationships between the instantaneous sensor outputs and the ground truth stiffness of the fruit. Our pairwise statistical comparison by the Wilcoxon test at 5% significance revealed the competitive performance frontiers of our approach for stiffness prediction; the Gaussian process kernel functions and the binary trees outperformed other models at a mean squared error (MSE) of 1.0 and 2×10?23, respectively. Most neural network models achieved competitive learning performance at MSE less than 10?5 and the utmost performance being a pyramidal class of feed-forward neural architectures. The results portray the potential of achieving accurate ripeness estimation of fruit using intelligent tactile sensors with fast machine learning schemes across the supply chain.
查看更多>>摘要：? 2022 Elsevier B.V.Computational Fluid Dynamics (CFD) approach is a versatile technique being applied to multiple areas of knowledge. CFD approach has contributed to comprehensively understand the complexity of biological systems and it has gained relevance since the end of the twentieth century. CFD became the key approach to replicate spatiotemporal phenomena combining fundamentals of physics, chemistry and biology. A considerable development derived from numerous CFD studies has been accomplished in a large range of applications in agriculture such as biological systems, and in agricultural buildings in particular. This paper provides an examination of recent progress (last 2 decades) in CFD studies, mainly applied to greenhouses and livestock buildings with the aim to depict current status and trends and potential research directions. Current status comprises CFD applied to regional and local climate, housing design and operation and animal or plant interaction with its surroundings. Integrating different scales from animal or plant physiologic responses to an agricultural building and to a region, CFD capabilities have made it possible to considerably improve the realism of simulations. Applications also cover a focus on specific equipments with the aim to assess benefits on energy and water fluxes. Additionally, the issues of validation and reliability of CFD simulations are addressed as well as the perspective of coupling CFD to operational devices or decision-making systems, based on ICT, IoT, or Artificial Intelligence.
查看更多>>摘要：? 2022Good humidity control is helpful to prevent the occurrence of waterfowl diseases, so it is necessary to predict and control the humidity in waterfowl houses. Traditional sequence methods such as RNN, LSTM, and GRU face challenges in prediction accuracy and parallel performance for long sequence prediction. In this study, a novel neural network model called SRU–SRU-dense is proposed to predict the waterfowl indoor humidity for the next 6 h. Simple recurrent unit(SRU) has better parallel performance compared with LSTM and GRU, which can effectively reduce the inference time. The proposed SRU–SRU-dense model is a seq2seq model based on SRU, and the experimental results show that this model has faster prediction speed and more accurate shorter prediction accuracy than seq2seq models based on RNN, LSTM, and GRU. In addition, we also compared the performance of two different seq2seq structures, seq2seq-dense and seq2seq-sequence, and the experimental results show that the seq2seq-dense structure has faster prediction speed and better prediction accuracy in the prediction of waterfowl indoor humidity in the next 6 h.
查看更多>>摘要：? 2022This paper presents the development of an autonomous harvesting robot system for tomato, a representative crop cultivated in the facility horticulture smart farm. The automated harvesting work using a robotic system is very challenging because of the appearance, environmental features, such as the atypical directions of the peduncles or their growing form in a bunch. Also, the robot system should enable the harvesting of the target fruit only without damaging other fruits, stems, and branches. Hence, this paper presents a deep learning network pipeline, Deep-ToMaToS, capable of three-level maturity classification and 6D pose (3D translation + 3D rotation) estimation of the target fruit simultaneously. Due to the difficulties encountered in building a large-scale dataset to train and test the deep learning model for the 6D pose estimation in the real world, we presented an automatic data collection scheme based on a photo-realistic 3D simulator environment. The robotic harvesting system includes a harvesting motion control algorithm based on the result of the 6D pose estimation. The overall process of the motion control phase is described along with the decision way of the appropriate final posture of the harvesting module mounted at the end-effector of the robot manipulator via removal of invalid motions getting out of the valid workspace or redundant motions. We conducted experiments on the 6D pose estimation based on the Deep-ToMaToS and the harvesting motion control in virtual and real smart farm environments. The experimental results showed a 6D pose estimation accuracy of 96 % based on the ADD_S metric, and the proposed harvesting motion control algorithm achieves the harvesting success rate of 84.5 % on average. The experimental results reveal that the harvesting robot system has significant potential to extend to harvesting works for other fruits and crops.
查看更多>>摘要：? 2022 Elsevier B.V.With increasing population, the demand for agricultural productivity is rising to meet the goal of “Zero Hunger”. Consequently, farmers have optimized the agricultural activities in a sustainable way with the modern technologies. This integration has boosted the agriculture production due to high potentiality in assisting the farmers. The impulse towards the technological advancement has revived the traditional agriculture methods and resulted in eco-friendly, sustainable, and efficient farming. This has revolutionized the era of smart farming which primarily alliance with modern technologies like, big data, machine learning, deep learning, swarm intelligence, internet-of-things, block chain, robotics and autonomous system, cloud-fog-edge computing, cyber physical systems, and generative adversarial networks (GAN). To cater the same, a detailed survey on ten hot-spots of smart farming is presented in this paper. The survey covers the technology-wise state-of-the-art methods along with their application domains. Moreover, the publicly available data sets with existing research challenges are investigated. Lastly, the paper concludes with suggestions to the identified problems and possible future research directions.
查看更多>>摘要：? 2022 Elsevier B.V.Evapotranspiration (ET) is the crucial parameter of agricultural irrigation and the hydrological cycle. To obtain the optimal estimation model of ET with film-mulching for spring maize, the extreme learning machine model (ELM) optimized by sparrow search algorithm (SSA) was built. The ET results were compared with four machine learning models, including artificial bee colony algorithm optimized ELM model (ABC-ELM), particle swarm algorithm optimized ELM model (PSO-ELM), genetic algorithm optimized ELM model (GA-ELM), ELM model, and two empirical models, including the modified Shuttleworth-Wallace model (SW) and Priestley-Taylor model (PT). We evaluated the accuracy of different models using the root mean square error (RMSE), coefficient of determination (R2), mean absolute error (MAE), coefficient of efficiency (Ens) and GPI. The results showed that the SSA-ELM models show high accuracy under different input combinations in different growth periods. Throughout the growing season of spring maize, the slope of the fitting equation of the SSA-ELM9 model was 0.895. The RMSE, R2, Ens, MAE and GPI were 0.433 mm/d, 0.895, 0.895, 0.342 mm/d and 1.382, respectively. The SSA-ELM models showed the highest accuracy for ET estimation of spring maize in different growth periods, followed by PSO-ELM, ABC-ELM and GA-ELM models. The accuracy of the SSA-ELM models was better than that of the SW PT models. Therefore, the SSA-ELM model can estimate spring maize ET with film mulching.
NSTL
Elsevier