Design and Development of a Leaf Blight Detection System Using Image Processing with Intrusion Notification via SMS

Abstract

The color of plant leaves is used as an indicator to detect leaf blight disease in plants using digital image processing. Early detection of leaf blight disease in plants allows proper protection to avoid the widespread of the disease. The general objective of this study was to design and develop a leaf blight detection system using image processing and a motion sensor with a laser tripwire alert system to detect intrusion and send SMS alerts. The data were randomly divided into three portions, with 75% being allocated as the training dataset, 22% as validation, and 3% as testing. The YOLOv5 object detection model was trained to detect leaf blight disease, resulting in a mean average precision of 38.80% (mAP 0.5) and 13.3% (mAP 0.5:0.95), with a precision of 56% and a recall of 37.90%. The low performance can be attributed to the low number of datasets involved, also, there were no essential steps carried out in image analysis that helped enhance the data in images namely, noise reduction, contrast enhancement, image resizing, color correction, and segmentation. To improve the performance of the system, an additional data enhancement in images may be implemented to simplify the image and easier to analyze namely, preprocessing, feature extraction, feature selection, and classification. Also, the use of additional prevailing algorithms namely, SVM + Deep Features, ANN, and Fast R-CNN can help improve the accuracy of results. In addition, the motion sensor and laser tripwire for detecting intrusions, have high accuracy rates of 90% and 95%, respectively. Furthermore, the motion sensor and GSM module responsiveness has an average of 6.54 seconds out of 20 trials while the tripwire sensor and GSM module responsiveness has an average of 6.53 seconds out of 20 trials as well.

Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗

The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.