蔬菜农药残留检测传感信息中的一种特征提取方法

A feature extraction method of sensors in detection of pesticide residues in vegetables

蔬菜中的农药残留是一个备受社会关注的问题。利用5只气敏传感器、1个温度传感器和1个湿度传感器构成测试系统,对蔬菜中2种多个浓度级别的农药残留成分进行测试。运用小波包的三尺度分解技术来分析测试结果,以小波包节点能量最大值的倒数构造特征向量,并运用主成分分析(PCA)和RBF神经网络对多浓度级别下的农药残留按类识别。结果表明:使用此方法提取的特征向量可使测试系统很好地把两种多浓度级别下的农药残留实现按类识别,且不受浓度级别的影响。

Pesticide residues in vegetables was a great problem concerned by the society. In our investigation,a feature extraction method was proposed for discriminating two kinds of pesticide residues with different concentrations in vegetables based on an array consisted of five gas sensors,one temperature and one humidity sensor. Three-scale wavelet packet decomposition was employed to analysis measure results,then feature vectors was given by calculating reciprocal of the largest energy corresponding to wavelet packet decomposition node. Principal component analysis （PCA） and radial basis function neural network （RBFNN） were employed to classify and discriminate two the kinds of pesticide residues. The results showed that feature vectors extracted by this method could make the test system correctly discriminate the two kinds of pesticide residue in vegetables,and the discrimination effect was not affected by different concentration levels of pesticide residues.