Pesticide residues exceeding the allowed figure in vegetables affects the human edible security and the trade. At present, commonly methods in analysis of pesticide residues is gas chromatography(GC), High Performance Liquid Chromatography(HPLC), gas chromatography mass spectrometry (GC-MS)etc, however these methods require expensive equipment , relatively complex procedures, specific experimental conditions and certain professional technical operating personnels, so generally do not meet the rapid detection requirements.Electronic nose as a novel bionic detection means based on principle of simulated biological sense of smell, with its fast, simple and inexpensive, just to meet the need detection of pesticide residues in vegetables. Feature extraction of measurement data is extremely important part of the electronic nose recognition process,and is the basis for pattern recognition.In this master thesis,we mainly studies the feature extraction methods of sensors in detection of pesticide residues in vegetables. The signal of pesticide residues in vegetables is extracted by different feature extraction methods. Principal Component Analysis (PCA) and Fisher discriminant analysis (FDA) are employed to identify different kinds of pesticide residues and different concentrations of pesticide residues.The main achievements in this paper are as follows:1. Describeing the five kinds of feature extraction methods commonly used in the electronic nose. PCA and FDA are employed to identify different kinds of pesticide residues in vegetables.2. Based on Wavelet Analysis method,two different feature extraction methods(reciprocal of wavelet energy and the mean value of wavelet coefficients)are proposed for discriminating different kinds of pesticide residues in vegetables and the result is very good. 3. Different feature extraction methods are researched to discriminate same kind of pesticide residues with different concentrations in vegetables; the concentration forecasting model of pesticide residues in vegetable is established based on BP neural network and good results are achieved.
|