A Practical Approach for Crop Insect Classification and Detection Using Machine Learning

Insect identification is one of the most pressing difficulties for Indian farmers, as numerous insect species harm a vast number of crops and hence diminish the quality of harvests, resulting in financial losses for both farmers and the country. However, in agriculture, the combination of IoT and machine learning (ML) allows for ease and innovation, allowing farmers all over the world to better their farming operations. On the other hand, in India, a very little amount of farmers is aware of smart farming and its benefits. Various study on research paper shows that the proper use of IoT devices embedded with the machine learning algorithm can reduce the task of farmer at very early stage of the plant life and thus saving the crops from being degraded, also included the survey of various research done across the globe and identified the potential methods which must be included for the current era farmers in order to minimize the insect effect on the crops. The aim of our experiment is to involve ML and IoT technology to sense the crop conditions in terms of quality and whether it is affected by insect or not for this a experimental study with the help of image processing has been performed thus calculation of results done accordingly. There are various sensors, which are equipped with ML technology like computer vision algorithm, which make the sensor powerful, and images being captured by these sensor can be analysed automatically and thus trigger the automated pesticide treatment systems using a ML-based decision support model. In this paper, study of Convolution Neural Network (CNN), Long Short Term Memory (LSTM), Support Vector Machine (SVM), Grid search based SVM (Grid-SVM), and K-nearest Neighbour classifier has been done Among them based on the required performance, nd the CNN-based model is much accurate for predicting the required treatment. The CNN has achieved up to 88% of accurate classification. Further, the model has been extending by incorporating the regression analysis, which enables the system to recommend the quantity of the required treatment. In this context, study on the KNN regression and Support Vector Regression (SVR) model has been made, among them the KNN regression provides up to 99.8% accurate prediction for treatment quantity prediction.