Artificial Intelligence in Agriculture最新文献

Bruising reduces the edibility and marketability of fresh apples, inevitably causing economic losses for the apple industry. However, bruises lack obvious visual symptoms, which makes it challenging to detect them using imaging techniques with uniform or diffuse illumination. This study employed the structured light imaging (SLI) technique to detect apple bruises. First, the grayscale reflection images were captured under phase-shifted sinusoidal illumination at three different wavelengths (600, 650, and 700 nm) and six different spatial frequencies (0.05, 0.10, 0.15, 0.20, 0.25, and 0.30 cycles mm⁻¹). Next, the grayscale reflectance images were demodulated to produce direct component (DC) images representing uniform diffuse illumination and amplitude component (AC) images revealing bruises. Then, by quantifying the contrast between bruised regions and sound regions in all AC images, it was found that bruises exhibited the optimal contrast when subjected to sinusoidal illumination at a wavelength of 700 nm and a spatial frequency of 0.25 mm⁻¹. In the AC image with optimal contrast, the developed h-domes segmentation algorithm to accurately segment the location and range of the bruised regions. Moreover, the algorithm successfully accomplished the task of segmenting central bruised regions while addressing the challenge of segmenting edge bruised regions complicated by vignetting. The average Intersection over Union (IoU) values for the three types of bruises were 0.9422, 0.9231, and 0.9183, respectively. This result demonstrated that the combination of SLI and the h-domes segmentation algorithm was a viable approach for the effective detection of fresh apple bruises.

The Nong Han Chaloem Phrakiat Lotus Park is a tourist attraction and a source of learning regarding lotus plants. However, as a training area, it lacks appeal and learning motivation due to its conventional presentation of information regarding lotus plants. The current study introduced the concept of smart learning in this setting to increase interest and motivation for learning. Convolutional neural networks (CNNs) were used for the classification of lotus plant species, for use in the development of a mobile application to display details about each species. The scope of the study was to classify 11 species of lotus plants using the proposed CNN model based on different techniques (augmentation, dropout, and L2) and hyper parameters (dropout and epoch number). The expected outcome was to obtain a high-performance CNN model with reduced total parameters compared to using three different pre-trained CNN models (Inception V3, VGG16, and VGG19) as benchmarks. The performance of the model was presented in terms of accuracy, F1-score, precision, and recall values. The results showed that the CNN model with the augmentation, dropout, and L2 techniques at a dropout value of 0.4 and an epoch number of 30 provided the highest testing accuracy of 0.9954. The best proposed model was more accurate than the pre-trained CNN models, especially compared to Inception V3. In addition, the number of total parameters was reduced by approximately 1.80–2.19 times. These findings demonstrated that the proposed model with a small number of total parameters had a satisfactory degree of classification accuracy.

Litchi, a succulent and perishable fruit, presents a narrow annual harvest window of under two weeks. The advent of smart agriculture has driven the adoption of visually-guided, automated litchi harvesting techniques. However, conventional approaches typically rely on laboratory-based, high-performance computing equipment, which presents challenges in terms of size, energy consumption, and practical application within litchi orchards. To address these limitations, we propose a real-time litchi detection methodology for complex environments, utilizing portable, low-energy edge computing devices. Initially, the litchi orchard imagery is collected to enhance data generalization. Subsequently, a convolutional neural network (CNN)-based single-stage detector, YOLOx, is constructed to accurately pinpoint litchi fruit locations within the images. To facilitate deployment on portable, low-energy edge devices, we employed channel pruning and layer pruning algorithms to compress the trained model, reducing its size and parameters. Additionally, the knowledge distillation technique is harnessed to fine-tune the network. Experimental findings demonstrated that our proposed method achieved a 97.1% compression rate, yielding a compact litchi detection model of a mere 6.9 MB, while maintaining 94.9% average precision and 97.2% average recall. Processing 99 frames per second (FPS), the method exhibited a 1.8-fold increase in speed compared to the unprocessed model. Consequently, our approach can be readily integrated into portable, low-computational automatic harvesting equipment, ensuring real-time, precise litchi detection within orchard settings.

The ability to automatically recognize sheep breeds holds significant value for the sheep industry. Sheep farmers often require breed identification to assess the commercial worth of their flocks. However, many farmers specifically the novice one encounter difficulties in accurately identifying sheep breeds without experts in the field. Therefore, there is a need for autonomous approaches that can effectively and precisely replicate the breed identification skills of a sheep breed expert while functioning within a farm environment, thus providing considerable benefits the industry-specific to the novice farmers in the industry. To achieve this objective, we suggest utilizing a model based on convolutional neural networks (CNNs) which can rapidly and efficiently identify the type of sheep based on their facial features. This approach offers a cost-effective solution. To conduct our experiment, we utilized a dataset consisting of 1680 facial images which represented four distinct sheep breeds. This paper proposes an ensemble method that combines Xception, VGG16, InceptionV3, InceptionResNetV2, and DenseNet121 models. During the transfer learning using this pre-trained model, we applied several optimizers and loss functions and chose the best combinations out of them. This classification model has the potential to aid sheep farmers in precisely and efficiently distinguishing between various breeds, enabling more precise assessments of sector-specific classification for different businesses.

Rice stands as a crucial staple food globally, with its enduring sustainability hinging on the prompt detection of rice leaf diseases. Hence, efficiently detecting diseases when they have already occurred holds paramount importance for solving the cost of manual visual identification and chemical testing. In the recent past, the identification of leaf pathologies in crops predominantly relies on manual methods using specialized equipment, which proves to be time-consuming and inefficient. This study offers a remedy by harnessing Deep Learning (DL) and transfer learning techniques to accurately identify and classify rice leaf diseases. A comprehensive dataset comprising 5932 self-generated images of rice leaves was assembled along with the benchmark datasets, categorized into 9 classes irrespective of the extent of disease spread across the leaves. These classes encompass diverse states including healthy leaves, mild and severe blight, mild and severe tungro, mild and severe blast, as well as mild and severe brown spot. Following meticulous manual labelling and dataset segmentation, which was validated by horticulture experts, data augmentation strategies were implemented to amplify the number of images. The datasets were subjected to evaluation using the proposed tailored Convolutional Neural Networks models. Their performance are scrutinized in conjunction with alternative transfer learning approaches like VGG16, Xception, ResNet50, DenseNet121, Inception ResnetV2, and Inception V3. The effectiveness of the proposed custom VGG16 model was gauged by its capacity to generalize to unseen images, yielding an exceptional accuracy of 99.94%, surpassing the benchmarks set by existing state-of-the-art models. Further, the layer wise feature extraction is also visualized as the interpretable AI.

In dairy farming, ensuring the health of each cow and minimizing economic losses requires individual monitoring, achieved through cow Re-Identification (Re-ID). Computer vision-based Re-ID relies on visually distinguishing features, such as the distinctive coat patterns of breeds like Holstein.

However, annotating every cow in each farm is cost-prohibitive. Our objective is to develop Re-ID methods applicable to both labeled and unlabeled farms, accommodating new individuals and diverse environments. Unsupervised Domain Adaptation (UDA) techniques bridge this gap, transferring knowledge from labeled source domains to unlabeled target domains, but have only been mainly designed for pedestrian and vehicle Re-ID applications.

Our work introduces Cumulative Unsupervised Multi-Domain Adaptation (CUMDA) to address challenges of limited identity diversity and diverse farm appearances. CUMDA accumulates knowledge from all domains, enhancing specialization in known domains and improving generalization to unseen domains. Our contributions include a CUMDA method adapting to multiple unlabeled target domains while preserving source domain performance, along with extensive cross-dataset experiments on three cattle Re-ID datasets. These experiments demonstrate significant enhancements in source preservation, target domain specialization, and generalization to unseen domains.

To ensure sustainability in agriculture, many optimization problems need to be solved. An important one of them is harvest scheduling problem. In this study, the harvest scheduling problem for the tea is discussed. The tea harvest problem includes the creating a harvest schedule by considering the farmers' quotas under the purchase location and factory capacity. Tea harvesting is carried out in cooperation with the farmer - factory. Factory management is interested in using its resources. So, the factory capacity, purchase location capacities and number of expeditions should be considered during the harvesting process. When the farmer's side is examined, it is seen that the real professions of farmers are different. On harvest days, farmers often cannot attend to their primary professions. Considering the harvest day preferences of farmers in creating the harvest schedule are of great importance for sustainability in agriculture. Two different mathematical models are proposed to solve this problem. The first model minimizes the number of weekly expeditions of factory vehicles within the factor and purchase location capacity restrictions. The second model minimizes the number of expeditions and aims to comply with the preferences of the farmers as much as possible. A sample application was performed in a region with 12 purchase locations, 988 farmers, and 3392 decares of tea fields. The results show that the compliance rate of farmers to harvesting preferences could be increased from 52% to 97%, and this situation did not affect the number of expeditions of the factory. This result shows that considering the farmers' preferences on the harvest day will have no negative impact on the factory. On the contrary, it was concluded that this situation increases sustainability and encouragement in agriculture. Furthermore, the results show that models are effective for solving the problem.

Diseases caused by invasive pathogens are an increasing threat to forest health, and early and accurate disease detection is essential for timely and precision forest management. The recent technological advancements in spectral imaging and artificial intelligence have opened up new possibilities for plant disease detection in both crops and trees. In this study, Dutch elm disease (DED; caused by Ophiostoma novo-ulmi,) and American elm (Ulmus americana) was used as example pathosystem to evaluate the accuracy of two in-house developed high-precision portable hyper- and multi-spectral leaf imagers combined with machine learning as new tools for forest disease detection. Hyper- and multi-spectral images were collected from leaves of American elm genotypes with varied disease susceptibilities after mock-inoculation and inoculation with O. novo-ulmi under greenhouse conditions. Both traditional machine learning and state-of-art deep learning models were built upon derived spectra and directly upon spectral image cubes. Deep learning models that incorporate both spectral and spatial features of high-resolution spectral leaf images have better performance than traditional machine learning models built upon spectral features alone in detecting DED. Edges and symptomatic spots on the leaves were highlighted in the deep learning model as important spatial features to distinguish leaves from inoculated and mock-inoculated trees. In addition, spectral and spatial feature patterns identified in the machine learning-based models were found relative to the DED susceptibility of elm genotypes. Though further studies are needed to assess applications in other pathosystems, hyper- and multi-spectral leaf imagers combined with machine learning show potential as new tools for disease phenotyping in trees.

Agriculture 5.0 is an emerging concept where sensors, big data, Internet-of-Things (IoT), robots, and Artificial Intelligence (AI) are used for agricultural purposes. Different from Agriculture 4.0, robots and AI become the focus of the implementation in Agriculture 5.0. One of the applications of Agriculture 5.0 is weed management where robots are used to discriminate weeds from the crops or plants so that proper action can be performed to remove the weeds. This paper discusses an in-depth review of Machine Learning (ML) techniques used for discriminating weeds from crops or plants. We specifically present a detailed explanation of five steps required in using ML algorithms to distinguish between weeds and plants.

Crops grown on aquaponics farms are susceptible to various diseases or biotic stresses during their growth cycle, just like traditional agriculture. The early detection of diseases is crucial to witnessing the efficiency and progress of the aquaponics system. Aquaponics combines recirculating aquaculture and soilless hydroponics methods and promises to ensure food security, reduce water scarcity, and eliminate carbon footprint. For the large-scale implementation of this farming technique, a unified system is needed that can detect crop diseases and support researchers and farmers in identifying potential causes and treatments at early stages. This study proposes an automatic crop diagnostic system for detecting biotic stresses and managing diseases in four leafy green crops, lettuce, basil, spinach, and parsley, grown in an aquaponics facility. First, a dataset comprising 2640 images is constructed. Then, a disease detection system is developed that works in three phases. The first phase is a crop classification system that identifies the type of crop. The second phase is a disease identification system that determines the crop's health status. The final phase is a disease detection system that localizes and detects the diseased and healthy spots in leaves and categorizes the disease. The proposed approach has shown promising results with accuracy in each of the three phases, reaching 95.83%, 94.13%, and 82.13%, respectively. The final disease detection system is then integrated with an ontology model through a cloud-based application. This ontology model contains domain knowledge related to crop pathology, particularly causes and treatments of different diseases of the studied leafy green crops, which can be automatically extracted upon disease detection allowing agricultural practitioners to take precautionary measures. The proposed application finds its significance as a decision support system that can automate aquaponics facility health monitoring and assist agricultural practitioners in decision-making processes regarding crop and disease management.