基于高光谱成像技术的苹果表面缺陷无损检测

以"红富士"苹果为研究对象,提出基于高光谱成像技术结合图像分割技术的苹果表面缺陷的无损检测方法。采用高光谱图像采集系统(400 nm～1 000 nm)采集完好无损和表面有缺陷苹果的高光谱图像;对采集到的高光谱图像进行最小噪声分离变换,提取感兴趣区域的平均光谱反射率;采用图像分割技术提出苹果表面缺陷的无损检测方法。结果表明:采用最小噪声分离变换可有效地消除苹果高光谱图像中的噪声;在700 nm～800 nm以及900 nm～1 000 nm波段范围内完好无损和表面有缺陷的苹果的光谱反射率值具有明显的差异,同时选取特征波长717.98 nm处的光谱反射率值小于0.6以及982.59 nm处的光谱反射率值大于0.52作为区分苹果正常区域和表面缺陷区域的阈值条件,进一步利用阈值分割方法对80个完好无损苹果和40个表面有缺陷苹果的正确识别率分别为97.5%和95%。表明高光谱成像技术结合图像分割技术可实现苹果表面缺陷的无损检测。     

加工番茄虫眼及霉变的可见近红外高光谱成像检测

为了探求一种快速有效识别虫眼和霉变加工番茄的无损检测方法,利用高光谱成像技术,从光谱和图像2个角度对其进行检测。先借助可见近红外高光谱成像系统获取408~1 013nm的加工番茄高光谱图像数据,提取并分析感兴趣区域的平均光谱曲线进行主成分分析,根据各波段权重系数优选了550,750,900nm 3个特征波长;然后通过特征波长下图像的主成分分析,选择缺陷部位与正常区域强度对照最明显的第一主成分图像,通过掩模、阈值处理和形态学开运算等图像处理方法对缺陷番茄进行检测判别。虫眼、霉变和正常三类番茄的识别率分别为93.3%,90%,100%。同时利用上述3个特征波长进行波段比图像运算,并选择波段比550nm/750nm图像进行缺陷识别,虫眼、霉变和正常三类加工番茄的识别率分别为93.3%,96.7%,100%。研究结果表明,二次主成分分析和波段比检测算法均可以有效地识别缺陷加工番茄。另外研究中仅选用了3个特征波段,数据量大大减少,为搭建开发适于加工番茄缺陷的多光谱在线检测系统提供了可能。     

基于可见/近红外高光谱成像技术的番茄表面农药残留无损检测

以番茄为研究对象,应用可见/近红外高光谱成像技术对水果表面农药残留的无损检测研究。用蒸馏水将嘧霉胺农药稀释成1∶20,1∶100,1∶500 3个梯度,将不同浓度的溶液分别滴到60个洗净的番茄表面,形成3×3矩阵。放置在通风阴凉处12h后,应用高光谱系统(400~1 000nm)采集光谱图像信息。利用主成分分析法获得主成分图像(PC),并根据第二主成分图像(PC-2)的权重系数选取特征波长564,809,967nm。采用波段比(564nm/809nm)结合适当的图像处理方法对番茄表面的农药残留进行检测。高浓度(1∶20,1∶100)农药点检测率为100%,而低浓度(1∶500)农药点的检测率为0。结果表明,高光谱成像技术对高浓度农药残留具有较好的检测效果。     

基于红外与可见光图像融合的苹果表面缺陷检测方法

目的：解决目前中国苹果分级分类大部分情况下仍需要进行人工筛选的问题。方法：采用基于多尺度变换的红外与可见光图像融合算法对所采集到的苹果的可见光图像和红外图像进行融合,得到缺陷特征更加直观的融合图像,对该图像进行图像的预处理操作得到二值化图像数据集,再采用卷积神经网络的AlexNet模型对之前的苹果表面缺陷数据集进行训练、验证和检测。结果：该检测方法在所制作的苹果表面缺陷数据集上对完好果、缺陷果、花萼/果梗、花萼/果梗加缺陷识别的平均准确度为99.0%,其中对花萼/果梗的识别准确率可达95.8%,对完好果、缺陷果和花萼/果梗加缺陷的识别准确率高达100%。结论：该方法对苹果表面缺陷的检测精度比较高,可以满足对苹果的在线分级的需求。     

高光谱图像技术检测苹果外部损伤

利用高光谱分选仪对苹果进行图像获取,基于光谱图像信息实现对苹果的外部损伤检测。由于光谱图像间的数据巨大,具有冗余性和相关性,通过主成分降维法提取640 nm波段下的图像作为特征信息进行外部损伤检测。光谱特征图像上光斑噪声的存在是造成损伤检测效果不佳的原因,分析光谱图像光斑的图像特征,利用二阶巴特沃斯高通滤波器抑除图像频域内的光斑信号,同时增强损伤区域的边缘图像信息。对比滤波前、后的特征图像损伤检测效果,滤波后的图像数据更能真实表现出样本自身的数据特点,结果表明:经过二阶巴特沃斯高通滤波器滤波后的苹果损伤检测效果明显,正确率达到95%。     

一种新型苹果果实特征提取算法

针对当前苹果特征提取常用的Hough算法存在运算复杂、实时性差等缺陷,提出了一种新型苹果果实特征提取算法。该算法利用一个滑动的高斯模板和苹果图像进行卷积运算提取苹果的圆形。试验及仿真结果表明,该方法可以实现单一、相邻和重叠3种情况下,苹果果实检测的高准确率,且在相邻并重叠的复杂情况下,其识别准确率也能达到94.1%,而在单个苹果的情况下,苹果果实的检测准确率可达96.6%,完全满足苹果实时、高效分级的需要。     

利用高光谱图像技术检测梨表面碰压伤的试验研究

以梨为研究对象,初步探讨了应用高光谱图像技术检测梨表面碰压伤的方法。采集梨在400～1 000nm范围的高光谱图像,应用主成分分析方法（PCA）获得主成分图像,根据第三主成分图像（PC-3）中各波长的权重,选出特征波长,分别是572nm、696nm和945nm。经过适当的图像处理方法对梨表面的碰压伤进行检测。检测结果表明,高光谱技术对检测梨表面碰压伤效果非常明显。     

基于高光谱成像技术无损检测苹果表面缺陷

基于高光谱成像技术结合模式识别,建立了苹果表面缺陷识别模型。首先,利用高光谱图像采集系统采集完好无损和表面有缺陷苹果的高光谱图像,提取感兴趣区域的平均光谱反射率;然后,比较标准正态变换(SNV)和多元散射校正(MSC) 2种光谱预处理方法对建模效果的影响,得出MSC为建模最优预处理方法。最后,采用主成分分析法选择累计贡献率超过99%的前5个主成分作为样本集特征光谱数据,分别建立了基于K最近邻(KNN)模式识别和偏最小二乘判别分析(PLS-DA)识别模型。结果表明:光谱经MSC预处理后,基于PLS-DA建立的识别模型对校正集和检验集识别率均达到100%,表明基于高光谱成像技术结合模式识别可实现苹果表面缺陷的无损检测。     

基于高光谱成像技术的水果品质无损检测

高光谱图像技术结合光谱技术与计算机图像技术两者的优点,可获得大量包含连续波长光谱信息的图像块,其图像信息可检测水果的外部品质,光谱信息则可用于水果内部品质的检测,达到根据水果内、外部综合品质进行分类的目的. 综述了国内外将该技术应用于水果品质检测方面的研究进展,提出了利用高光谱图像技术检测苹果轻微损伤的方法,利用500～900nm的高光谱图像数据,通过主成分分析提取547nm波长下的特征图像.     

高光谱图像法对稻谷贮藏中五种常见真菌生长拟合及区分

利用高光谱成像系统(HIS)获取稻谷贮藏中常见真菌(黑曲霉、米曲霉、杂色曲霉、构巢曲霉、桔青霉)在马铃薯葡萄糖琼脂板上培养期间的高光谱图像,波峰709 nm处的光谱值和全波段光谱值的第一主成分得分两种方法构建真菌Gompertz函数的生长模拟模型。Gompertz函数拟合结果显示,五种真菌基于全波段光谱值PCA分析后的第一主成分得分建立的生长拟合模型R2为0.1781~0.9501,基于波峰709 nm光谱值建立的拟合模型R2为0.9095~0.9679,效果明显优于第一主成分得分的建模效果。另外,主成分分析(PCA)结合偏最小二乘法判别分析(PLS-DA)可以区分五种不同菌种。其中,训练集和测试集中,PLS-DA模型对培养48 h的黑曲霉、米曲霉、构巢曲霉、桔青霉四种真菌及对照组的区分准确率为100%;而对杂色曲霉,训练集区分准确率为100%,测试集的区分率为33.33%。结果表明高光谱图像技术能够用来对真菌种类进行区分。     

Application of Near-Infrared Hyperspectral Imaging for Detection of External Insect Infestations on Jujube Fruit

A hyperspectral imaging system has been built for detecting external insect damage and acquiring reflectance images from jujubes in the near-infrared region of 900–1700 nm. Spectral information was extracted from each jujube, and six optimal wavelengths (987, 1028, 1160, 1231, 1285, and 1464 nm) were obtained using principal component analysis. The first principal component images (PC-1) using the selected six wavelengths were obtained for further image processing. The detection algorithm was then developed based on principal component analysis and two-band ratio (R1160/R1464) coupled with image subtraction algorithm (R1160-R1464). An identification accuracy of 93.1% for insect-infested jujubes and 100% classification rate for the intact ones were achieved. The results of this research demonstrated that it is feasible to discriminate insect-infested jujubes from intact jujubes using the near-infrared hyperspectral imaging technology.     

Visual detection of apple bruises using AdaBoost algorithm and hyperspectral imaging

Hyperspectral imaging technique (400–1000 nm) was used for rapid and nondestructive recognition of bruises of apples. A total of 324 hyperspectral images were collected from 108 Fuji apples and the average spectral reflectance was extracted from the region of interest (ROI) of each image. The classification results of AdaBoost for the data pretreated by various existing methods were compared. Then, the correlation-based feature selection (CFS) algorithm was used to obtain characteristic wavelengths for reducing data redundancy. After pretreating with multiplicative scatter correction (MSC) and CFS, the average accuracy of the selected wavelengths was 97.63%. Then, an image processing algorithm based on the characteristic wavelengths selected before was proposed for the visual discrimination of bruises. This algorithm performed independent component analysis (ICA) transformation of the selected wavelengths, and chose the third component image of the ICA transform, then used adaptive threshold segmentation to obtain the bruise region of apples. The results showed that hyperspectral imaging technology could discriminate apple bruise, and this study can help to develop an online apple bruises detection system.     

Detection of Fecal Contamination on Cantaloupes Using Hyperspectral Fluorescence Imagery

To determine whether detection of fecal contamination on cantaloupes is possible using fluorescence imaging, hyperspectral images of cantaloupes artificially contaminated with a range of diluted bovine feces were acquired from 425 to 774 nm in responses to ultraviolet‐A (320 to 400 nm) excitation. Evaluation of images at emission peak wavelengths indicated that 675 nm exhibited the greatest contrast between feces contaminated and untreated surface areas. Two‐band ratios compared with the single‐band images enhanced the contrast between the feces contaminated spots and untreated cantaloupe surfaces. The 595/655‐nm, 655/520‐nm, and 555/655‐nm ratio images provided relatively high detection rates ranging from 79% to 96% across all feces dilutions. However, both single band and ratio methods showed a number of false positives caused by such features as scarred tissues on cantaloupes. Principal component analysis (PCA) was performed using the entire hyperspectral images data; 2nd and 5th principal component (PC) image exhibited differential responses between feces spots and false positives. The combined use of the 2 PC images demonstrated the detection of feces spots (for example, minimum level of 16‐μg/mL dry fecal matter) with minimal false positives. Based on the PC weighing coefficients, the dominant wavelengths were 465, 487, 531, 607, 643, and 688 nm. This research demonstrated the potential of multispectral‐based fluorescence imaging for online applications for detection of fecal contamination on cantaloupes.     

Development of algorithms for detecting citrus canker based on hyperspectral reflectance imaging

BACKGROUND: Automated discrimination of fruits with canker from other fruit with normal surface and different type of peel defects has become a helpful task to enhance the competitiveness and profitability of the citrus industry. Over the last several years, hyperspectral imaging technology has received increasing attention in the agricultural products inspection field. This paper studied the feasibility of classification of citrus canker from other peel conditions including normal surface and nine peel defects by hyperspectal imaging. RESULTS: A combination algorithm based on principal component analysis and the two‐band ratio (Q_687/630) method was proposed. Since fewer wavelengths were desired in order to develop a rapid multispectral imaging system, the canker classification performance of the two‐band ratio (Q_687/630) method alone was also evaluated. The proposed combination approach and two‐band ratio method alone resulted in overall classification accuracy for training set samples and test set samples of 99.5%, 84.5% and 98.2%, 82.9%, respectively. CONCLUSION: The proposed combination approach was more efficient for classifying canker against various conditions under reflectance hyperspectral imagery. However, the two‐band ratio (Q_687/630) method alone also demonstrated effectiveness in discriminating citrus canker from normal fruit and other peel diseases except for copper burn and anthracnose. Copyright © 2011 Society of Chemical Industry     

基于改进CNN和数据扩充的苹果表面缺陷检测

目的:提高苹果表面缺陷的检测准确率和效率。方法:基于改进卷积神经网络(CNN)和数据扩充建立苹果表面缺陷检测方法。改建CNN的拓扑结构,并将其用于苹果表面缺陷检测;利用条件生成对抗网络,合成表面无缺陷和有缺陷苹果图像,实现图像数据扩充和提高改进CNN的苹果表面缺陷的识别性能;通过模型剪枝,合理权衡苹果表面缺陷的检测准确率、检测时间及节能限制,以提高所提方法的实用性。结果:当改进CNN中的解释层选用2 048个解释性神经元时,平均检测准确率最高;条件生成对抗网络增强了苹果图像数据集的多样性;随着增强图像数在测试数据集中占比的增加,所提方法对苹果表面缺陷的检测准确率不断升高;当剪枝后的模型尺寸占原始模型尺寸的百分比从100%降至50%时,可以以6.96%的准确率损失将苹果表面缺陷的检测效率提升1倍。结论:试验方法有望在苹果生产和加工过程中实现自动化缺陷检测。     

基于可见-近红外高光谱技术对鸡种蛋性别鉴定的研究

目的 为实现鸡种蛋胚胎性别的无损检测,提出了基于可见-近红外高光谱检测海兰褐鸡种蛋胚胎性别的方法。方法 通过分析种蛋0~14 d大头部位的400~1000 nm波段下的光谱,建立基于偏最小二乘判别分析（PLS-DA）和支持向量机（SVM）的种蛋性别判别模型,比较不同孵育天数下的模型判别率,优选出最佳的检测天数;通过分析四种不同的预处理算法,选出最佳的鸡种蛋胚胎高光谱预处理方法,最后构建基于全波段和特征波段光谱信息的判别模型,并对结果进行比较。结果 基于PLS-DA和SVM的模型在第9 d的预测集结果达到最高,分别为80%和82.5%。主成分分析（PCA）结果表明,雄雌种蛋光谱信息可以进行区分;变量标准化（SNV）为最佳预处理方法;全波段相对于连续投影算法（SPA）、竞争性自适应重加权算法（CARS）选择特征波长的模型更优,建模集、预测集准确率分别为90%和85%。结论 研究结果表明可见-近红外高光谱技术可以快速、较准确、无损检测海兰褐种蛋胚胎性别,该技术为褐壳种蛋胚胎性别鉴定实现在线检测提供了一定的理论基础。     

Classification of Black Beans Using Visible and Near Infrared Hyperspectral Imaging

A rapid and non-destructive method based on the visible and near infrared hyperspectral imaging technique in the wavelength range of 390–1050 nm was investigated for discriminating the varieties of black beans. In total, 300 samples of three varieties were scanned by the visible and near infrared hyperspectral imaging system, and hyperspectral data were analyzed by spectral and image processing technique respectively. A successive projection algorithm was used to obtain 13 characteristic wavelengths (504, 507, 512, 516, 522, 529, 692, 733, 766, 815, 933, 998, and 1000 nm) for spectral analysis. After the processing of successive projection algorithm, optimal image selection was carried out by principal component analysis based on the characteristic wavelengths. The first principal component image was used for the image analysis, whose contribution rate was over 98.34%. Gray level co-occurrence matrix analysis from first principal component image was applied to extract image features including 16 textural features and six morphological features. In this study, partial least squares-discriminate analysis, support vector machine, and K-nearest neighbors were used for model establishments, respectively, based on spectral feature, image feature, and the combination of spectral and image features. The results show that the best correct discrimination rate of 98.33% was achieved by applying combined spectral and image features. The study demonstrated that visible and near infrared hyperspectral imaging technique was potential for rapid classification of black beans, and the performance of the classification model can be improved by the feature combination.     

基于高光谱成像技术的大米溯源研究

利用高光谱成像技术提取大米的光谱信息进行大米产地溯源研究。采用X-Y距离样本集算法（SPXY）进行训练集和测试集的划分,将1 000颗大米样本中800个为训练集,剩下200个为测试集。并采用主成分分析（PCA）法提取相关性较强的主成分光谱信息,进行数据降维。基于主成分分析法提取前4个主成分,并在贡献率最高的第4主成分基础上,结合支持向量机算法（SVM）建立大米产地溯源预测模型。研究得出训练集准确率可达96%,测试集平均准确率为79%。通过训练集和测试集的实验结果表明,高光谱成像技术可以对大米产地进行溯源,为大米产地快速、无损检测提供了一定思路和参考。     

基于高光谱成像技术与GBDT的大豆品种无损鉴别

大豆的品种直接关系到大豆制品的质量和出油率,目前主要采用对大豆中蛋白质及脂肪等含量的检测来实现对大豆品种的鉴别。这种鉴别方式破坏了大豆本质,并且存在检测费用高、效率低、精度差的问题。本文基于高光谱成像技术和机器学习理论,研究了大豆品种无损快速鉴别方法。采集并建立了4个品种(每个品种200粒,共计800粒)大豆的高光谱原始图像及光谱数据集。研究了利用归一化、均值中心化、小波变换、S-G平滑滤波以及矢量归一化对采集到的高光谱数据进行滤波去噪预处理,建立了基于KNN、RF及GBDT的大豆种粒无损检测模型。实验对比得出,利用主成分分析结合GBDT的检测模型精度最高,识别准确率可达99.58%,结果表明,利用高光谱成像技术结合主成分分析的GBDT 算法模型能够有效消除噪声干扰的影响,实现对大豆种粒快速、准确的无损检测,并对其他农作物的品种检测具有一定的参考意义。