半透射高光谱多指标同时检测马铃薯内外部缺陷

目的应用半透射高光谱成像技术结合支持向量机(support vector machine,SVM)模型实现马铃薯内外部缺陷多指标同时检测。方法采集310个马铃薯样本半透射高光谱图像,并分别采用标准正态变量变换(standard normalized variate,SNV)、归一化(normalize)和平滑处理(smoothing)对光谱信息进行预处理。进一步采用竞争性自适应重加权算法结合无信息变量消除法(competitive adaptive reweighed sampling algorithm,uninformative variable elimination,CARS-UVE)进行特征波长选择,提高模型识别率。结果原始光谱信息经归一化预处理和竞争性自适应重加权算法结合无信息变量消除法(CARS-UVE)降维后所建的支持向量机(SVM)模型识别结果最优,该方法对合格、绿皮和黑心马铃薯样本预测结果分别为90.7%、88.9%、95.7%,混合识别率为91.3%。结论采用半透射高光谱成像技术结合CARS-UVE方法所建SVM模型能够实现马铃薯内外部缺陷多指标同时检测。

Simultaneous detection of multiple index for internal and external defects of potato based on semi-transmission hyperspectral

Objective To realize multi-index simultaneous detection of internal and external defects of potato by semi-transmissive hyperspectral imaging technology combined with support vector machine (SVM) model. Methods A total of 310 semitransparent hyperspectral images of potato samples were collected, spectral information of that were pretreated by using standard normal variate transformation (SNV), normalize and smoothing, respectively. Further, characteristics of wavelength were selected using competitive adaptive re-weighed sampling algorithm combined with uninformative variable elimination (CARS-UVE) to improve pattern recognition rate. Results The model of SVM which was built by original spectral information that got normalize pretreatment and CARS-UVE dimension reduction was the optimal identification results. Using this method to forecast eligibility, green skin and black heart potato sample rate was 90.7%, 88.9% and 95.7%, respectively, and hybrid recognition rate could reach to 91.3%. Conclusion The established SVM model of potato by semi-transmissive hyperspectral imaging technologies combined with CARS-UVE method can realize potato internal and external defects simultaneous detection in multiple index.