Neural network based pest identification and control in cauliflower crop using sounds of pest

Abstract

The quantity of pesticides usage in the field increases every year, which in turn affects the nutrients present in the crop. These pesticides also cause water contamination, air pollution, serious health problems in humans and finally making the soil infertile. Necessary action has to be taken to protect the environment and crops from chemicals. A well-established automatic acoustic detection of pest for early pest detection is suggested in this paper. Acoustic identification technique of signal analysis is extensively implemented in agriculture to give maximum protection of crops, ultimately resulting in better production. Successful eradication of pest lies in identification of pest without damaging the nutrients and crops. The proposed technique detects the presence of pests in the initial stage. The sounds of different pests which are dreadfully affecting the agricultural crops are collected. The signals of pests are analyzed in time domain and frequency domain specifications. The features of different pests are extracted using Mat lab programming. The various statistical features of pest are trained and given to the hidden layer of neural network where it automatically classifies the pest indicating the presence of pest. The Back-Propagation algorithm is used for training the samples. When the identification of pest has been done, the pest can be killed by the microwave shock rather than applying highly toxic chemicals. This technique also benefits the farmer by avoiding contact with the pesticides. Direct contact with the pesticides sometimes causes skin cancer to farmers. The technique suggested in this paper is harmless to the farmers and crops, hence increases the production.