霉变玉米电子鼻识别及其传感器阵列优化

RECOGNITION OF MOLDY CORN USING ELECTRONIC NOSE AND OPTIMIZATION OF SENSOR ARRAY

收集了玉米样品40份,利用电子鼻技术对样品进行模式识别,并对电子鼻传感器阵列进行优化.结果表明,电子鼻能够对正常与霉变样品进行区分.在优化传感器阵列后,主成分分数较优化前的84.36%提高至97.54%.对测试集的判别采用4种算法(Euclid、Malahanobis、Kohonen和DFA)进行判别,电子鼻判别率较优化前均有不同程度的提高,其中Kohonen法判别率可达90.63%.

In the paper, we collected 40 corn samples, carried out the pattern recognition of the samples using electronic nose technique, and optimized the electronic nose sensor array. The results showed that the electronic nose technique could discriminate the normal sample from the moldy ones. After the sensor array was optimized, the fraction of the principal components was improved from the 84.36% （before optimization） to 97.54%. The test set was discriminated by four algorithms （Euclid, Malahanobis, Kohonen and DFA）, and the discrimination rate of electronic nose technique was improved to different degrees in comparison to that before optimization, wherein the discrimination rate of Kohonen algorithm was up to 90.63%.