GeoDLS: A Deep Learning-Based Corn Disease Tracking and Location System Using RTK Geolocated UAS Imagery

Deep learning-based solutions for precision agriculture have recently achieved promising results. learning has been used to identify crop diseases at the initial stages of disease development in an effort create effective management systems. However, use deep and unmanned aerial system (UAS) imagery track spread diseases, diseased regions within cornfields, notify users with actionable information remains a research gap. Therefore, this study, high-resolution, UAS-acquired, real-time kinematic (RTK) geotagged, RGB altitude 12 m above ground level (AGL) was develop Geo Disease Location System (GeoDLS), tracking corn fields. UAS images (resolution 8192 × 5460 pixels) were acquired cornfields located Purdue University’s Agronomy Center Research Education (ACRE), using DJI Matrice 300 RTK mounted 45-megapixel Zenmuse P1 camera during V14 R4. A dataset 5076 created by splitting UAS-acquired tile simple linear iterative clustering (SLIC) segmentation. For segmentation, split into tiles sizes 250 pixels, 500 1000 resulting 1804, 1112, 570 image tiles, respectively. SLIC 865 725 superpixel obtained compactness (m) values 5 10, Five neural network architectures, VGG16, ResNet50, InceptionV3, DenseNet169, Xception, trained diseased, healthy, background DenseNet169 identified highest testing accuracy 100.00% when on size pixels. Using sliding window approach, model then calculate percentage present each image. Finally, geolocation update location 2 cm through web application, smartphone email notifications. The GeoDLS could be potential tool automated regions, provide users.