AETC: an automated pest detection and classification model using optimal integration of Yolo plus SSD and adaptive ensemble transfer CNN with IoT-assisted pest images

Insects harm or destroy the crops and plants in agriculture fields by causing infection to the plants or destroying the valuables, which is called as a pest. When a plant is invaded by pests, the quality of the food it produces decreases drastically. So, it is highly essential to detect the pests before they attack the plants. However, the existing pest detection and categorizing techniques need suggestions and decisions from entomologists, and also this process consumes more time. If pests are identified at an early stage, then it could help the farmer to eliminate the necessity for pesticides and also increase food production. Because of its almost similar look, detecting and classifying the pests associated with a crop is complex work for the farmer, especially during the initial stage of plant growth. The sudden and productive growth in the Internet-of-Things (IoT) technology also finds its application in agriculture, resulting in a transition from statistical to quantitative methods. To alleviate the issues in the agricultural sector, a new framework for an IoT-assisted Automatic Pest Prediction and Classification (APDC) model using ensemble transfer learning of the convolutional neural network (CNN) method is developed. At first, IoT sensors are used to capture pest images from the agricultural field. These images are stored in the standard database, from which these images are taken for conducting experiments. The gathered images are then subjected to image pre-processing for contrast enhancement by median filter (MF). After that, the pests are detected from the pre-processed image by means of a Hybrid You Only Look Once (Yolo) v3 and Single Shot multi-box Detector (HYSSD) model. In this model, two algorithms, namely the Beetle Swarm Optimization (BSO) and the Salp Swarm Algorithm (SSA), are combined to optimize the parameters. An adaptive ensemble transfer CNN (AETC) is used to identify the pests after it has been detected. DenseNet, MobileNet, and ResNet are the three models that constitute this ensemble model. Finally, various metrics are used to verify the effectiveness of the proposed classification model. The findings from the results show that the recommended method has better classification accuracy.