Smart agricultural technology最新文献

The assessment of traits is important in determining production potential, reproductive performance, and overall health of dairy cows. The assessment of these traits typically involves visual inspection and manual measurement, which can be time-consuming, subject to bias, and potentially distressing for the animals. To address these challenges, convolutional neural networks (CNNs)-aided non-invasive computer vision system was developed in the present study. This system consists of a depth camera to acquire the RGB images and depth information of cows. The DeepLabV3+ model, having the ResNet50 model as a backbone, was utilized to segment the body parts of cows from RGB images. Image processing-based algorithms were developed to extract key pixel locations for each trait from these segmented images. The system estimated trait dimensions utilizing 3D data of respective key points. The mean-IoU (intersection-over-union) values for the developed segmentation models were 93.46%, 91.25%, and 99.27% for side-view, back-view traits, and stature, respectively. Additionally, the vision system was able to estimate the trait dimensions with mean absolute percentage error (MAPE) below 6.0%. For a few traits, MAPE, however, exceeded 10.0%, indicating higher error. Inaccurate segmentation, imprecise key point extraction, visual overlaps of specific body parts, and variations in cow postures contribute to such errors. The developed system attained a Ratio of Performance to Deviation (RPD) above 1.2 for all traits, indicating its ability to estimate the dimensions of traits efficaciously. Thus, the present study demonstrated the potential of a CNN-based computer vision-based system for automating the trait measurement process in cows.

We introduce a multimodal rapid identification and growth status discrimination method for morchella. Based on the unique biological characteristics and growth environmental requirements of morchella, the efficient and accurate identification of key growth stages of morchella is achieved through the integration of multimodal information acquisition technology. During the rapid identification process of the growth stage of Morchella, the Multi Stage Vision Enhanced Position Encoding Vision Transformer (MS-EP ViT) model is adopted. By introducing multi-stage input embedding, enhanced position encoding, and optimized Transformer Encoder layers, the performance of the model in identifying different growth stages of Morchella mushrooms is significantly improved. In the multimodal Morchella growth state discrimination method, text and image modalities are integrated, a Non downsampled Contourlet Transform Mask Region based Convolutional Neural Network (NSCT Mask R-CNN) model is designed, and a multimodal feature extraction strategy combining Non downsampled Contourlet Transform (NSCT) features with environmental features is explored. This strategy effectively achieves the goals of object detection and instance segmentation, and thus we have accurately evaluated the growth status of Morchella in the later stages of mulberry, young mushroom, and mature. The experimental results show that both models have achieved significant improvements in recognition accuracy and stability, and the rationality of hyperparameter settings has been verified through convergence and parameter sensitivity experiments. Overall, we provide a more accurate and efficient identification method for monitoring the growth of Morchella, which helps to better understand the growth of Morchella and provides scientific basis for optimizing its growth environment.

Accurate weed detection in agricultural images is a crucial challenge for improving crop management practices and reducing chemical usage. In this study, we propose an innovative segmentation model called DWUNet, inspired by popular architectures and incorporating the latest advances in the state of the art. Our model delivers remarkable accuracy, with a Jaccard index reaching 0.825, while ensuring fast inference speed of only 8 ms per image, thus providing an optimal solution for real-time applications. By comparing DWUNet to several state-of-the-art models, we demonstrate its superiority in terms of accuracy and efficiency. Furthermore, a qualitative analysis of the visual results confirms DWUNet's ability to accurately detect weeds and generalize results beyond the training data. This study represents a significant advancement in the field of precision agriculture, providing a powerful tool for sustainable crop management and reducing environmental impact.

The aim of the research is to develop a radar method for determining the physical and chemical parameters of subsurface soil horizons, providing rapid determination of moisture and specific conductivity in the area of the plant root system. The proposed method is based on a set of Fresnel equations which describe reflection of electromagnetic waves from the interface between dielectric media in vertical and horizontal polarization of the probing signal. For the practical implementation of the method, it is proposed to use two unmanned aerial vehicles that form a bistatic radar system which irradiates the Earth's surface obliquely in order to create the Brewster's effect and increase the fraction of the radio signal reflected from subsurface horizons. The percentage of moisture and the specific conductivity of soil are calculated from the measured values of the imaginary part of the complex permittivity. The required accuracy of moisture and conductivity measurements is achieved by two-step calibration of the measuring device. The values of the moisture content and specific conductivity of soil obtained by radar at a frequency of 469 MHz are in good agreement with the results of measuring these parameters using the soil moisture meter TDR 150 Spectrum Technologies, Inc.

Species distribution modeling (SDM) can be a valuable tool to improve perilla production by identifying optimal locations for its cultivation in the North Eastern Hill (NEH) region of India. Numerous habitat suitability modeling techniques are available; however, requirement of sophisticated hardware and software for their execution limits their in-situ utility to agriculturalists in real-time. Integrating SDM on edge devices for habitat suitability predictions is challenging due to the computational demands and complexity of current modeling techniques. Hence, in the present study, we developed an artificial intelligence (AI)-based mobile application to predict perilla habitat suitability solely from geographical location. Maximum Entropy (MaxEnt) software with perilla occurrence data from the NEH region was utilized to generate an accurate suitability map (Area Under Curve (AUC) for test data = 0.88). Probabilities and corresponding locations extracted from the suitability map were used as training data for AI models including Random Forest Regression (RFR), Support Vector Regression (SVR), and Artificial Neural Network (ANN). The ANN model, architecture being optimized using a genetic algorithm, achieved the best performance (R² = 0.81). All models show good predictive ability (R² > 0.75) in predicting actual habitat suitability (Residual Prediction Deviation (RPD) > 2.10), and a high degree of relationship between predicted and actual probabilities (AUC > 94.0 %) was also observed. The mobile application integrated with the ANN model achieved high AUC (>97.0 %) and R² (0.82) values for testing locations and predicted known perilla locations with an accuracy of 76.0 %. This shows the practical utility of AI-based mobile application for species distribution modeling and emphasizes its potential for perilla cultivators. The developed user-friendly mobile application may help farmers of NEH region to predict optimal locations for perilla cultivation in real-time with a single click, thereby enhancing sustainable production efficiency and biodiversity conservation efforts in their locales.

Flue-cured tobacco is a main economic crop, and leaves are the direct product of tobacco plant. Monitoring leaf area index (LAI) of tobacco plant is important to field management, yield prediction, and industry regulation. Hyperspectral data have hundreds of narrow spectral bands in continuous spectral ranges, significantly advancing monitoring of LAI. However, considering spectral responses of LAI vary with wavelength, the use of the full spectral range in estimation of LAI is redundant. To reduce spectral redundancy and improve estimation of LAI, spectral subsets were selected based on importance of spectral bands. Variable importance in the projection (VIP) score obtained by partial least squares regression (PLSR) was adopted to measure the importance. The study was conducted in Yunnan Province, China. Canopy reflectance spectra of tobacco plant were measured in two consecutive growth seasons. Genetic algorithm (GA) and PLSR were used for model calibration. The identified important spectral regions for estimation of LAI were red edge, near-infrared (NIR), and green regions. In estimation of LAI of tobacco plant, compared with the estimation using the full spectral range of VNIR reflectance spectra, normalized root mean square error (NRMSE) and coefficient of determination (R²) values were improved from 10.30% and 0.83 to 7.68% and 0.90 and from 18.78% and 0.43 to 7.65% and 0.90 by using the reflectance spectra in identified spectral regions in the growth seasons in 2021 and 2022 separately. In addition to the identified continuous spectral regions, central bands of the identified spectral regions also achieved estimation of LAI, with the highest R² value reaching 0.72. The selected spectral subsets improved estimation accuracy and reduced model complexity. The results indicate that selected spectral subsets are effective and promising in estimation of LAI, providing an alternative for estimation of LAI using hyperspectral data.

Cassava crop is in the forefront among the arable crops mostly cultivated in Nigeria. It is a crop that International Institute of Tropical Agriculture (IITA) have invested more efforts to develop mobile digital solutions for farmers to address some of the agronomic issues threatening its productivity. The ultimate aim of the digital solution is to bring innovations closer to farmers as well as to bring about the transformation needed to achieve food security in Africa. However, farmers’ (users) quality rating and usability of agricultural mobile apps remains uncertain thereby, creating knowledge gap. Feedbacks of users are critical to inform developers of the apps about necessary modifications. This study therefore examined the usability attributes of Agricultural Mobile Phone Applications (AMPAs) among cassava farmers in South-west Nigeria using users-centered approach. Because of the selected crop, apps evaluated in this study were Akilimo App, Airtel 4-21 call App and IITA herbicide calculator App. Akilimo App is an advisory tool that provides site-specific recommendations to cassava farmers in order to increase their cassava-based cropping systems; Airtel 4-2-1 call App provides a voice-based tutorial on cassava production, weather, education, rice amongst others by simply dialing 4-2-1 on any mobile phone with an Airtel sim card; while IITA herbicide is a mobile application developed to prevent herbicide abuse, such as over-dosing and/or under-dosing. Four hundred and ten (410) cassava farmers were recruited as respondents for the study. Data were analyzed using both descriptive and inferential statistics, including frequency counts, percentages, means and standard deviation. As the major contribution, it was shown for the first time, how users (farmers) rated the three Apps in terms of engagement, functionality, aesthetics, information quality, and subjective quality while usability of the Apps were rated for efficiency, effectiveness, satisfaction, learnability, memorability, cognitive load, accuracy and error. Similarly, pattern of usage, benefits of usage and constraints to usage of the Apps were further investigated.

In recent decades the agricultural intensification led to a landscape simplification, where once heterogeneous landscapes turned into more monocultured cropping regions. The main driver behind it was to be able to use larger and more powerful agricultural machinery in order to increase productivity and specially to decrease labor costs. With the advent of robots in combination with new methods of AI it is obvious that this is not only a new step of automated mechanization. New robotic systems open up new opportunities of how to cultivate crop plants. Different robotic concepts with different properties are discussed and how these might contribute to achieve the indisputably needed improvements in sustainable and environmental-friendly crop production systems. In a performed case study, it is demonstrated that even if larger robotic solutions when operated in isolation still can achieve higher area outputs, smaller and often specialized robotic systems can achieve a substantially improved work quality in the use case of seeding.

Vision Transformer (ViT) has recently attracted significant attention for its performance in image classification. However, studies have yet to explore its potential in detecting and classifying plant leaf disease. Most existing research on diseased plant leaf detection has focused on non-transformer convolutional neural networks (CNN). Moreover, the studies that applied ViT narrowly experimented using hyperparameters such as image size, patch size, learning rate, attention head, epoch, and batch size. However, these hyperparameters significantly contribute to the model performance. Recognising the gap, this study applied ViT to Java Plum disease detection using optimised hyperparameters. To harness the performance of ViT, this study presents an experiment on Java Plum leaf disease detection. Java Plum leaf diseases significantly threaten agricultural productivity by negatively impacting yield and quality. Timely detection and diagnosis are essential for successful crop management. The primary dataset collected in Bangladesh includes six classes, ‘Bacterial Spot’, ‘Brown Blight’, ‘Powdery Mildew’, and ‘Sooty Mold’, ‘healthy’, and ‘dry’. This experiment contributes to a thorough understanding of Java Plum leaf diseases. Following rigorous testing and refinement, our model demonstrated a significant accuracy rate of 97.51%. This achievement demonstrates the possibilities of using deep-learning tools in agriculture and inspires further research and application in this field. Our research offers a foundational model to ensure crop quality by precise detection, instilling confidence in the global Java Plum market.

Many horticultural tasks require a definition of and means to measure ‘plant state’, a quantity that is used to represent and compare plants. The plant state's evolution through time (plant growth) can then be tracked. This is often performed informally by humans based on heuristic features for particular plant types. It can be used for example to recommend interventions to catch up on expected growth, or to measure the effects of experimental interventions on growth. This work provides a purely data-driven definition, that is easy to train, easy to non-destructively acquire training data, does not require expert annotations, and is easy to compute for any new plant type and growing conditions. The presented method is applied to a dataset of lettuce plants where it exceeds the performance of a hard baseline. This work also demonstrates that the presented method retains the intuitive properties expected for a plant growth model. Open source code implementation and data is provided.