Profenofos Insecticide Residue Detection on Red Chili (Capsicum annum L.) through Image Augmentation and CNN (Convolutional Neural Network)

Abstract

The use of i nsecticide in agriculture is rapidly increasing but also faces challenges related to their negative impacts on human health and the environment. One commonly used insecticide is profenofos. Farmers frequently use Profenofo s to control pests on red chili plants in Indonesia. Therefore, detecting profenofos insecticide residue on chili is crucial to ensure the safety of chili consumption. This study aims to develop a new method for detecting profenofos insecticide residues on red chilies (Capsicum Ann um L.) through image augmentation and CNN (Convolutional Neural Network). In this study, insecticide solution preparation, chili image acquisition, testing with GC (Gas Chromatography) Agilent 6890 , image preprocessing, and CNN model implementation were co nducted. Insecticide solution spraying was conducted on 15 chili es with concentrations of 0 and 10 mg/l, followed by smartphone image acquisition, resulting in 30 images . Subsequently, each image was augmented 50 times, resulting in 1530 images. This proce ss involves rotation, shifting, zoom , a nd horizontal and vertical flipping. The CNN model, consisting of convolution layers and fully connected layers, was trained with the Adam optimizer, categorical cross - entropy loss function, the learning rate of 0,000 1, and trained for 20 epochs. The analysis results indicate that the model achieved an accuracy of 85% with precision, recall, and F1 - score values of 0,86 , 0,85, and 0,84, respectively. Based on these results, it can be concluded that the image augmentatio n and CNN method successfully detected profenofos insecticide residues on red chilies at concentrations of 0 and 10 mg/l.