CAJ | 학술논문

为了提高马铃薯痂疮病无损检测识别精度，基于机器视觉和近红外光谱的多源信息融合技术，该文提出DS（dempster shafer）证据理论结合支持向量机的马铃薯痂疮病无损检测方法。试验以360个马铃薯为研究对象，在图像特征分割时，确定了差影法结合马尔可夫随机场模型法为最佳分割方法；在光谱特征提取时，确定主成分分析方法为最佳降维方法。采用支持向量机识别方法分别建立机器视觉和近红外光谱的马铃薯痂疮病识别模型，模型对测试集马铃薯识别率分别为89.17%、91.67%。采用DS证据理论与支持向量机相结合的方法对获取的图像特征和光谱特征进行融合，建立了基于机器视觉和近红外光谱技术的多源信息融合马铃薯痂疮病检测模型，该模型对测试集马铃薯识别率为95.83%。试验结果表明，该技术对马铃薯痂疮病进行检测是可行的，融合模型比单一的机器视觉模型或近红外光谱模型识别率高。
위료제고마령서가창병무손검측식별정도，기우궤기시각화근홍외광보적다원신식융합기술，해문제출DS（dempster shafer）증거이론결합지지향량궤적마령서가창병무손검측방법。시험이360개마령서위연구대상，재도상특정분할시，학정료차영법결합마이가부수궤장모형법위최가분할방법；재광보특정제취시，학정주성분분석방법위최가강유방법。채용지지향량궤식별방법분별건립궤기시각화근홍외광보적마령서가창병식별모형，모형대측시집마령서식별솔분별위89.17%、91.67%。채용DS증거이론여지지향량궤상결합적방법대획취적도상특정화광보특정진행융합，건립료기우궤기시각화근홍외광보기술적다원신식융합마령서가창병검측모형，해모형대측시집마령서식별솔위95.83%。시험결과표명，해기술대마령서가창병진행검측시가행적，융합모형비단일적궤기시각모형혹근홍외광보모형식별솔고。
The common scab is a skin disease of the potato tuber that decreases the quality of the product and significantly influences the price, so it is very necessary to find a quickly nondestructive way to detect potato scabs. In this study, machine vision technology and near infrared spectroscopy analysis technology were used to detect potato scabs. In order to improve the potato scab nondestructive recognition accuracy, multi-sensor information fusion technique was proposed to detect potato scabs based on machine vision and near infrared spectroscopy. DS evidence theory combined with support vector machine method was used for multi-sensor information fusion technique. In the research, 360 potatoes were taken as testing samples (180 qualified potatoes and 180 scab potatoes). This study concluded that the difference image method combined with the Markov random field model method was the best segmentation method in the segmentation of image characteristics through the image preprocessing. And the principal component analysis method was the best method in the spectral feature extraction through the spectroscopy preprocessing. This study compared several different spectral preprocessing methods to preprocess the near infrared spectroscopy in near infrared spectroscopy preprocessing. And from the discriminating rate of the support vector machine model with the pretreated near infrared spectroscopy, it was concluded that the dimension reduction method was the best spectroscopy preprocessing method. The support vector machine method was a good pattern recognition method, so this study used the support vector machine method to detect potato scabs based on machine vision technology and near infrared spectroscopy analysis technology. The support vector machine models to discriminate potato scab were built based on machine vision technology and near infrared spectroscopy analysis technology respectively. The discriminating rates of these two models were 89.17% and 91.67% in testing sets respectively. To improve the discriminating rates of potato scab detecting with machine vision and near infrared spectroscopy respectively, a multi-sensor information fusion technique based on near infrared spectroscopy and machine vision method was used to detect the potato scab. DS evidence theory was a good information fusion method, so DS evidence theory combined with support vector machine method model was built with image characteristics and spectral characteristics. The multi-sensor information fusion model was used to detect the testing potato samples and the discriminating rates were 95.83%in the testing set. Compared with the results from the three detecting models, it was concluded that the discriminating rate of the model built with multi-sensor information fusion was 6.66%higher than the model built with machine vision technology, and 4.16% higher than the model built with near infrared spectroscopy analysis technology. The results indicated that it was feasible to detect potato scabs by using a multi-sensor information fusion technique based on near infrared spectroscopy and machine vision. The recognizing rate of the multi-sensor information fusion model was higher than that of the model built by machine vision technology or near infrared spectroscopy analysis technology respectively. That is to say multi-sensor information fusion technology is better for potato scab nondestructive detecting than machine vision technology respectively or near infrared spectroscopy analysis technology respectively. The research can provide references for potato disease detecting with a multi-sensor information fusion technique.