IoT-Enabled Plant Leaf Disease Detection Using HPJSO_SqueezeNet

Abstract

The Internet of Things (IoT) has become a highly effective tool over the past few decades, finding relevance in real-time applications. In agriculture, the use of automated technologies for detecting plant diseases offers immense benefits but also poses significant challenges. To address existing challenges, a hybrid framework named Hunter Prey Jellyfish Search Optimization (HPJSO) enabled SqueezeNet (HPJSO_SqueezeNet) has been developed for multi-classification of plant leaf disease detection in IoT. Here, HPJSO combines Hunter Prey Optimization (HPO) and Jellyfish Search Optimization (JSO). The IoT nodes are simulated. Then, the Cluster Head (CH) is executed employing the Low-energy adaptive clustering hierarchy (LEACH) protocol. After that, the routing is executed using the selected CH and it is given to the Base Station (BS) utilising HPJSO. At BS, the pre-processing phase is performed using a Gaussian filter. Thereafter, plant leaf segmentation is carried out by a Psi-Net trained with HPJSO. Moreover, the classification process is done by Deep Convolutional Neural Network (Deep CNN), which is trained by HPJSO. Finally, the multi-classification of plant leaf disease detection is achieved using SqueezeNet trained with the proposed HPJSO. In addition, the overall performance of the proposed HPJSO_SqueezeNet method for multi-classification is compared with other existing methods such as sine cosine algorithm-based rider neural network (SCA based RideNN), IoT-based Fuzzy Based Function Network (IoT_FBFN), Taylor-Water Wave Optimization-based Generative Adversarial Network (Taylor-WWO-based GAN), Smart Farm Monitoring System (SFMS), Deep Learning, Improved Quantum Whale Optimization with Principal Component Analysis (IQWO-PCA), HPO-based SqueezeNet and JSO-based SqueezeNet. Additionally, the simulation outcomes of HPJSO are examined with Energy efficient routing, Secure and Scalable Routing protocol (SARP), Trust aware routing and Competitive Versatile Flower Pollination (CVFP) based routing. The HPJSO has achieved the highest energy of 73.80%, throughput of 77.60%, delay of 24.50% and distance of 10028.40%. As well as, the HPJSO_SqueezeNet attained the accuracy of 0.898 and sensitivity of 0.937. The proposed HPJSO model achieves higher energy compared to several other methods, with improvements of 52.30% over Energy efficient routing, 50.81% over SARP, 18.97% over Trust aware routing, and 20.87% over CVFP-based routing based on routing. Likewise, the proposed HPJSO_SqueezeNet model achieves higher accuracy compared to several other methods with improvements of 8.91% over SCA-based RideNN, 5.68% over IoT_FBFN, 3.67% over Taylor-WWO-based GAN, 3.23% over SFMS, 2.34% over Deep Learning, 1.00% over IQWO-PCA, 1.67% over HPO-based SqueezeNet, and 1.00% over JSO-based SqueezeNet. The code for the proposed approach is found at ‘https://github.com/SridharM87/HPJSO_SqueezeNet.git’.