CAJ | 학술논문

探索一种能对长枣表面农药种类进行准确鉴别的方法。利用900 nm~1700 nm的高光谱成像系统对喷洒不同农药的长枣进行原始光谱数据采集，利用Savitzky-Golay卷积平滑的方法对900 nm~1700 nm波段范围内的原始光谱进行预处理，运用偏最小二乘回归系数选择7个特征波长（Seven Wavelengths， SW），采用对农药种类进行赋值的方法建立偏最小二乘（PLS）模型进行鉴别分析，结果表明，不同种类的农药在一定程度上能被有效的鉴别，鉴别正确率达到88.75%，但是在鉴别吡虫啉与哒螨灵两种农药时容易出现鉴别错误。进一步采用SW-LDA对不同农药种类进行鉴别，建立7个特征波长下的线性判别分析（LDA）模型，研究结果显示，不同农药种类的鉴别准确率达到90%以上。说明近红外高光谱联合SW-LDA能够准确地鉴别长枣表面不同的农药种类，为长枣表面农药种类的鉴别探索出了一种新方法。
탐색일충능대장조표면농약충류진행준학감별적방법。이용900 nm~1700 nm적고광보성상계통대분쇄불동농약적장조진행원시광보수거채집，이용Savitzky-Golay권적평활적방법대900 nm~1700 nm파단범위내적원시광보진행예처리，운용편최소이승회귀계수선택7개특정파장（Seven Wavelengths， SW），채용대농약충류진행부치적방법건립편최소이승（PLS）모형진행감별분석，결과표명，불동충류적농약재일정정도상능피유효적감별，감별정학솔체도88.75%，단시재감별필충람여달만령량충농약시용역출현감별착오。진일보채용SW-LDA대불동농약충류진행감별，건립7개특정파장하적선성판별분석（LDA）모형，연구결과현시，불동농약충류적감별준학솔체도90%이상。설명근홍외고광보연합SW-LDA능구준학지감별장조표면불동적농약충류，위장조표면농약충류적감별탐색출료일충신방법。
In this paper, a new non-destructive method for discriminating varieties of Pesticides on Lingwu long jujubes'surface by near-infrared hyperspectral technique was developed. Raw spectral of long jujubes which have been sprinkled by different pesticides was gathered by hyperspectral imaging system between 900 nm and 1 700 nm. Savitzky-Golay smoothing was used to deal with the raw spectral between 900 nm and 1 700 nm and 7 optimal wavelengths (SW) were selected by PLS. Different numbers was assigned to different Pesticides that PLS was established to identify. The results indicate that different Pesticides can be discriminated in a sense and discrimination accuracy is 88.75 %. However, the identification error can occur between Bichonglin and Damanling. Different Pesticides were discriminated by SW-LDA model. The correct answer rate of LDA model based on 7 wavelengths is more than 90%. The results indicate that variety discrimination of Pesticides can be achieved accurately by SW-LDA model based on near-infrared hyperspectral technique. SW-LDA model based on near-infrared hyperspectral technique is a new method for discriminating different Pesticides on long jujubes'surface.