This study evaluated the potential of using machine vision in combination with deep learning (DL) to identify the early blight disease in real-time for potato production systems. Four fields were selected to collect images (n = 5199) of healthy and diseased potato plants under variable lights and shadow effects. A database was constructed using DL to identify the disease infestation at different stages throughout the growing season. Three convolutional neural networks (CNNs), namely GoogleNet, VGGNet, and EfficientNet, were trained using the PyTorch framework. The disease images were classified into three classes (2-class, 4-class, and 6-class) for accurate disease identification at different growth stages. Results of 2-class CNNs for disease identification revealed the significantly better performance of EfficientNet and VGGNet when compared with the GoogleNet (FScore range: 0.84–0.98). Results of 4-Class CNNs indicated better performance of EfficientNet when compared with other CNNs (FScore range: 0.79–0.94). Results of 6-class CNNs showed similar results as 4-class, with EfficientNet performing the best. GoogleNet, VGGNet, and EfficientNet inference time values ranged from 6.8–8.3, 2.1–2.5, 5.95–6.53 frames per second, respectively, on a Dell Latitude 5580 using graphical processing unit (GPU) mode. Overall, the CNNs and DL frameworks used in this study accurately classified the early blight disease at different stages. Site-specific application of fungicides by accurately identifying the early blight infected plants has a strong potential to reduce agrochemicals use, improve the profitability of potato growers, and lower environmental risks (runoff of fungicides to water bodies).

The uniform application (UA) of agrochemicals results in the over-application of harmful chemicals, increases crop input costs, and deteriorates the environment when compared with variable rate application (VA). A smart variable rate sprayer (SVRS) was designed, developed, and tested using deep learning (DL) for VA application of agrochemicals. Real-time testing of the SVRS took place for detecting and spraying and/or skipping lambsquarters weed and early blight infected and healthy potato plants. About 24,000 images were collected from potato fields in Prince Edward Island and New Brunswick under varying sunny, cloudy, and partly cloudy conditions and processed/trained using YOLOv3 and tiny-YOLOv3 models. Due to faster performance, the tiny-YOLOv3 was chosen to deploy in SVRS. A laboratory experiment was designed under factorial arrangements, where the two spraying techniques (UA and VA) and the three weather conditions (cloudy, partly cloudy, and sunny) were the two independent variables with spray volume consumption as a response variable. The experimental treatments had six repetitions in a 2 × 3 factorial design. Results of the two-way ANOVA showed a significant effect of spraying application techniques on volume consumption of spraying liquid (p-value < 0.05). There was no significant effect of weather conditions and interactions between the two independent variables on volume consumption during weeds and simulated diseased plant detection experiments (p-value > 0.05). The SVRS was able to save 42 and 43% spraying liquid during weeds and simulated diseased plant detection experiments, respectively. Water sensitive papers’ analysis showed the applicability of SVRS for VA with >40% savings of spraying liquid by SVRS when compared with UA. Field applications of this technique would reduce the crop input costs and the environmental risks in conditions (weed and disease) like experimental testing.

Fertilizer spreaders capable of variable rate application are increasingly important for enhancing nutrient management in horticultural crops because they improve placement and increase nutrient uptake efficiency. Matching applied fertilizer to fertilizer requirements represents a significant input cost saving for the grower and a reduction in potential pollutant loading to ground and surface water. Variable rate fertilization (VRF) is a precision agriculture technology made possible by embedded high-speed computers, accurate Global Positioning System (GPS) receivers, Geographic Information Systems (GIS), remote sensing, yield or soil maps, actuators, and electronic sensors capable of measuring and even forecasting crop properties in real time. For tree crops like Florida citrus ( Citrus spp.), the most important function of the VRF spreader is to detect and avoid fertilizing spaces of the orchard not occupied by trees. Treeless spaces are becoming more common in Florida as diseases such as citrus greening ( Candidatus Liberibacter asiaticus) and canker ( Xanthomonas axonopodis ) cause the removal of thousands of trees every year. VRF works best under those conditions. Because VRF exploits crop and soil variability, it has no value in a perfectly uniform field. VRF enables smaller trees including resets to be fertilized at lower, most appropriate rates, thus minimizing any excess application. This article examines the existing knowledge on using precision agriculture and variable rate technology to keep water and nutrients in the root zone of horticultural crops, thus facilitating maximum uptake efficiency.

Deep learning convolutional neural networks (CNNs) are an emerging technology that provide an opportunity to increase agricultural efficiency through remote sensing and automatic inferencing of field conditions. This paper examined the novel use of CNNs to identify two weeds, hair fescue and sheep sorrel, in images of wild blueberry fields. Commercial herbicide sprayers provide a uniform application of agrochemicals to manage patches of these weeds. Three object-detection and three image-classification CNNs were trained to identify hair fescue and sheep sorrel using images from 58 wild blueberry fields. The CNNs were trained using 1280x720 images and were tested at four different internal resolutions. The CNNs were retrained with progressively smaller training datasets ranging from 3780 to 472 images to determine the effect of dataset size on accuracy. YOLOv3-Tiny was the best object-detection CNN, detecting at least one target weed per image with F₁-scores of 0.97 for hair fescue and 0.90 for sheep sorrel at 1280 × 736 resolution. Darknet Reference was the most accurate image-classification CNN, classifying images containing hair fescue and sheep sorrel with F₁-scores of 0.96 and 0.95, respectively at 1280 × 736. MobileNetV2 achieved comparable results at the lowest resolution, 864 × 480, with F₁-scores of 0.95 for both weeds. Training dataset size had minimal effect on accuracy for all CNNs except Darknet Reference. This technology can be used in a smart sprayer to control target specific spray applications, reducing herbicide use. Future work will involve testing the CNNs for use on a smart sprayer and the development of an application to provide growers with field-specific information. Using CNNs to improve agricultural efficiency will create major cost-savings for wild blueberry producers.

This study was carried out with the purpose of testing the ability of deep learning machine vision to identify microscopic objects and geometries found in chemical crystal structures. A database of 6994 images taken with a light microscope showing microscopic crystal details of selected chemical compounds along with 180 images of an unknown chemical was created to train and test, respectively the deep learning models. The models used were GoogLeNet (22 layers deep network) and VGG-16 (16 layers deep network), based on the Caffe framework (University of California, Berkeley, CA) of the DIGITS platform (NVIDIA Corporation, Santa Clara, CA). The two models were successfully trained with the images, having validation accuracy values of 97.38% and 99.65% respectively. Finally, both models were able to correctly identify the unknown chemical sample with a high probability score of 93.34% (GoogLeNet) and 99.41% (VGG-16). The positive results found in this study can be further applied to other unknown sample identification tasks using light microscopy coupled with deep learning machine vision.