Rapid Detection of Microorganisms Using Image Processing Parameters and Neural Network

A rapid and cost-effective technique for identification of microorganisms was explored using fluorescence microscopy and image analysis, and classification was done with trained neural network. The microorganisms used in this study are Bacillus thuringiensis (C399), Escherichia coli K12 (ATCC 10798), Lactobacillus brevis (LJH240), Listeria innocua (C366), and Staphylococcus epidermis (LJH343). After staining the microorganisms with fluorescent dyes [diamidino-2-phenyl-indole and acridine orange (AO)], images of the microorganisms were captured using a digital camera attached to a light microscope. Geometrical, optical, and textural features were extracted from the images using image analysis. Parameters extracted from images of microorganisms stained with AO gave better results for classification of the microorganisms. From these parameters, the best identification parameters that could classify the microorganisms with higher accuracy were selected using a probabilistic neural network (PNN). PNN was then used to classify the microorganisms with a 100% accuracy using nine identification parameters. These parameters are: 45° run length non-uniformity, width, shape factor, horizontal run length non-uniformity, mean gray level intensity, ten percentile values of the gray level histogram, 99 percentile values of the gray level histogram, sum entropy, and entropy. When the five microorganisms were mixed together then, also the PNN could classify the microorganisms with 100% accuracy using these nine parameters.     

Identification of multiple grain image samples from tray

This paper presents a methodology to identify the different grain types from image samples of tray containing multiple grains using colour and textural features. The multiple grain images are segmented into individual grain images. From these images, eighteen colour and twenty‐four textural features are obtained. A neural network model is implemented for identification of bulk food grains. Five different types of grains namely, alasandi, green gram, metagi, red gram and wheat commonly used in Indian food preparations are considered in this work. The maximum and minimum food grain identification accuracies observed in this work are 94% and 80% for wheat and alasandi, respectively. The work finds application in development of machine vision system for grain identification, classification and grading.     

Application of Image Texture Analysis for Evaluation of X-Ray Images of Fungal-Infected Maize Kernels

The feasibility of image texture analysis to evaluate X-ray images of fungal-infected maize kernels was investigated. X-ray images of maize kernels infected with Fusarium verticillioides and control kernels were acquired using high-resolution X-ray micro-computed tomography. After image acquisition and pre-processing, several algorithms were developed to extract image textural features from selected two-dimensional (2D) images of the kernels. Four first-order statistics (mean, standard deviation, kurtosis and skewness) and four grey level co-occurrence matrix (GLCM) features (correlation, energy, homogeneity and contrast) were extracted from the side, front and top views of each kernel and used as inputs for principal component analysis (PCA). The first-order statistical image features gave a better separation of the control from infected kernels on day 8 post-inoculation. Classification models were developed using partial least squares discriminant analysis (PLS-DA), and accuracies of 67 and 73% were achieved using first-order statistical features and GLCM extracted features, respectively. This work provides information on the possible application of image texture as method for analysing X-ray images.     

Textural analysis of hyperspectral images for improving contaminant detection accuracy

Previous studies demonstrated a hyperspectral imaging system has a potential for poultry fecal contaminant detection by measuring reflectance intensity. The simple image ratio at 565 and 517 nm images with optimal thresholding was able to detect fecal contaminants on broiler carcasses with high accuracy. However, differentiating false positives from real contaminants, especially cecal feces were challenging. Further image processing such as textural analysis in the spatial domain was able to reduce false positive errors. In this study, textural analysis of hyperspectral images was conducted to improve detection accuracy by reducing false positives. Specifically, textural analysis with co-occurrence matrix of hyperspectral images performed well to identify “true” contamination. In addition, co-occurrence matrix textural features including average, variance, entropy, contrast, correlation, moment of poultry hyperspectral images were investigated for selecting optimal features to represent contamination. Image pre-processing with co-occurrence textural analysis, specifically mean of co-occurrence textural feature from the matrix (0° angle and distance equals to one) followed by image ratio was able to improve fecal detection accuracy without additional optical filters that increase cost for system hardware of multispectral imaging system for on-line application. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture.     

Application of Visible and Near Infrared Hyperspectral Imaging to Differentiate Between Fresh and Frozen–Thawed Fish Fillets

The potential of visible and near infrared (VIS/NIR) hyperspectral imaging was investigated as a rapid and nondestructive technique to determine whether fish has been frozen–thawed. A total of 108 halibut (Psetta maxima) fillets were studied, including 48 fresh and 60 frozen–thawed (F-T) samples. Regarding the F-T samples, two speeds of freezing (fast and slow) were tested. The hyperspectral images of fillets were captured using a pushbroom hyperspectral imaging system in the spectral region of 380 to 1,030 nm. All images were calibrated for reflectance, followed by the minimum noise fraction rotation to reduce the noise. A region-of-interest (ROI) at the image center was selected, and the average spectral data were generated from the ROI image. Dimension reduction was carried out on the ROI image by principal component analysis. The first three principal components (PCs) explained over 98 % of variances of all spectral bands. Gray-level co-occurrence matrix analysis was implemented on the three PC images to extract 36 textural feature variables in total. Least squares-support vector machine classification models were developed to differentiate between fresh and F-T fish based on (1) spectral variables; (2) textural variables; (3) combined spectral and textural variables, respectively. Satisfactory average correct classification rate of 97.22 % for the prediction samples based on (3) was achieved, which was superior to the results based on (1) or (2). The results turned worse when different freezing rates were taken into consideration to classify three groups of fish. The overall results indicate that VIS/NIR hyperspectral imaging technique is promising for the reliable differentiation between fresh and F-T fish.     

Application of Atomic Force Microscopy on Rapid Determination of Microorganisms for Food Safety

ABSTRACT: Rapid detection and quantification of microorganisms is important for food quality, safety, and security. In this field, nanotechnology appears to be promising in its ability to characterize an individual microorganism and detect heterogeneous distribution of microbes in food samples. In this study, atomic force microscopy (AFM), a nanotechnology tool, was used to investigate Escherichia coli (E. coli) qualitatively and quantitatively. E. coli strains B and K12 were used as surrogates to represent pathogenic strains, such as E. coli O157: H7. The results from AFM were compared with those from scanning/transmission electron microscopy (SEM/TEM). The qualitative determination was obtained using morphology and characteristic parameters from AFM images, and the quantitative determination was obtained by calculating the microorganisms in AFM images. The results show that AFM provides a new approach for rapid determination of microorganisms for food safety.     

Classification of tea grains based upon image texture feature analysis under different illumination conditions

This paper discusses the role of illumination in discrimination of tea samples based upon textural features of tea granules. The images of tea granules were acquired using 3CCD color camera under Dual Ring light which consists of both Darkfield as well as Brightfield type of illumination. Ten graded tea samples were analyzed. Five textural features were ‘entropy’, ‘contrast’, ‘homogeneity’, ‘correlation’ and ‘energy’ obtained under both illuminations. The acquired textural features were subjected to principal component analysis (PCA). The results showed that best discrimination was obtained with Darkfield illumination with a variance of 96% whereas Brightfield illumination showed low discrimination with only 83% variance. Analysis of PCA biplot indicated correlations among graded tea samples and textural features. The study concludes that textural features may be used to estimate tea quality under Darkfield illumination being non-destructive and quick technique.     

高光谱亚像元分解预测花生中的黄曲霉毒素B1

黄曲霉毒素B1是一种剧毒、强致癌物质,具有紫外荧光特性。为研究高光谱成像技术对黄曲霉素的检测能力,在365 nm紫外灯下,通过高光谱成像系统采集5个浓度共250个花生籽粒样本33个波段(400~720 nm)的高光谱图像。提出一种基于高光谱亚像元分解丰度图像直方图量化特征预测黄曲霉毒素含量的方法。该方法首先通过N-FINDR端元提取方法获得黄曲霉毒素端元光谱,然后对高光谱图像进行非负矩阵分解(NMF),得到黄曲霉毒素丰度图像,对丰度图像构建直方图量化特征,使用偏最小二乘回归(PLS)和支持向量机回归(SVR)进行黄曲霉毒素丰度反演,五折交叉验法得到平均两种回归模型预测相对误差分别为29.95%和12.16%,RMSE最高为0.0306。本研究结果对农产品籽粒黄曲霉毒素光学快速检测具有积极意义。     

Detection of Fecal Residue on Poultry Carcasses by Laser-Induced Fluorescence Imaging

ABSTRACT:  Feasibility of fluorescence imaging technique for the detection of diluted fecal matters from various parts of the digestive tract, including colon, ceca, small intestine, and duodenum, on poultry carcasses was investigated. One of the challenges for using fluorescence imaging for inspection of agricultural material is the low fluorescence yield in that fluorescence can be masked by ambient light. A laser-induced fluorescence imaging system (LIFIS) developed by our group allowed acquisition of fluorescence from feces-contaminated poultry carcasses in ambient light. Fluorescence emission images at 630 nm were captured with 415-nm laser excitation. Image processing algorithms including threshold and image erosion were used to identify fecal spots diluted up to 1: 10 by weight with double distilled water. Feces spots on the carcasses, without dilution and up to 1: 5 dilutions, could be detected with 100% accuracy regardless of feces type. Detection accuracy for fecal matters diluted up to 1: 10 was 96.6%. The results demonstrated good potential of the LIFIS for detection of diluted poultry fecal matter, which can harbor pathogens, on poultry carcasses.     

A real-time computer vision-based platform for fabric inspection part 1: algorithm

Developing an efficient real-time detection algorithm is quite important for an automated inspection system. This paper presents a practical method based on local singular value decomposition (SVD) and normalised cross-correlation (NCC) for real-time defect detection in woven fabrics. As fabric-textured images exhibit high periodicity among the repeated sub-patterns, non-defective or normal image samples (image patches) can be efficiently approximated as a linear combination of the basis vectors (BVs) obtained via SVD. Since these BVs are recovered from normal samples, they will only capture the key structural features of the non-defective images. When using the BVs to model new samples, we can expect defective or abnormal samples with structural features not found in normal cases will incur substantial approximation errors. Therefore, complex defect detection can be converted to a template matching problem, where the robust NCC is utilised to measure disparities between the original and its approximation for defect identification. Experimental results on various real-world fabrics exhibit accurate defect detection with low false alarm rate, and we also conduct a comparison with a feature extraction-based method to further confirm the effectiveness of our algorithm.     

Citrus canker detection using hyperspectral reflectance imaging and PCA-based image classification method

Citrus canker is one of the most devastating diseases that threaten marketability of citrus crops. Technologies that can efficiently identify citrus canker would assure fruit quality and safety and enhance the competitiveness and profitability of the citrus industry. This research was aimed to investigate the potential of using hyperspectral imaging technique for detecting canker lesions on citrus fruit. A portable hyperspectral imaging system consisting of an automatic sample handling unit, a light source, and a hyperspectral imaging unit was developed for citrus canker detection. The imaging system was used to acquire reflectance images from citrus samples in the wavelength range between 400 and 900 nm. Ruby Red grapefruits with normal and various diseased skin conditions including canker, copper burn, greasy spot, wind scar, cake melanose, and specular melanose were tested. Hyperspectral reflectance images were analyzed using principal component analysis (PCA) to compress the 3-D hyperspectral image data and extract useful image features that could be used to discriminate cankerous samples from normal and other diseased samples. Image processing and classification algorithms were developed based upon the transformed images of PCA. The overall accuracy for canker detection was 92.7%. Four optimal wavelengths (553, 677, 718, and 858 nm) were identified in visible and short-wavelength near-infrared region that could be adopted by a future multispectral imaging solution for detecting citrus canker on a sorting machine. This research demonstrated that hyperspectral imaging technique could be used for discriminating citrus canker from other confounding diseases.     

A fast approach for finding design repeat in textile rotary printing for fault detection

Speed is the most critical factor for real-time defect detection in fabric rotary printing systems. Most of the methods proposed in the literature work by acquiring an error-free image and registering this image with subsequent images captured during the production. In these methods, major time-consuming problem is to find the design repeat in the sample image before further processing. We proposed a fast method for finding the design repeat. The method is tuned up according to the textile printing domain. During the printing process, the fabric moves in a horizontal direction, so the image registration can be restricted to the same direction. To further speed up the process, the image registration method is applied using few initial pixel columns of the reference image with the sample image. This bunch of selected columns of the reference image is matched with the same number of columns selected from the sample image by moving this bunch on the sample image column by column. The maximum matching position is marked as a start of the design repeat. Since the repeat size is always fixed, we can extract the complete design pattern from an acquired image for further processing of fault detection. Experimental results on different fabric designs using the above-mentioned method are very promising.     

Using Wavelet Textural Features of Visible and Near Infrared Hyperspectral Image to Differentiate Between Fresh and Frozen–Thawed Pork

In this study, wavelet textural analysis was applied to hyperspectral images in the visible and near-infrared (VIS/NIR) region (400–1,000 nm) for differentiation between fresh and frozen–thawed pork. The spectral data of acquired hyperspectral images were analyzed using partial least squares (PLS) regression and five wavelengths (462, 488, 611, 629, and 678 nm) were selected as the feature wavelengths by the regression coefficients from the PLS model. The fourth-order daubechies wavelet (“db4”) was used to serve as the wavelet mother function for wavelet textural extraction of the feature images at the above selected feature wavelengths with the wavelet decomposition level from 1 to 4. Four textural features were calculated in the horizontal, vertical, and diagonal orientations at each level. Forty-eight textural features were extracted from each feature image and used to differentiate between fresh and frozen–thawed pork samples by least-squares support vector machine (LS-SVM) model. Wavelet texture extracted from all five feature images at first decomposition level was identified as optimal wavelet texture combination, resulting in the highest classification accuracy for the LS-SVM models (98.48 % for the training set and 93.18 % for the testing set). Based on the texture combination, the quality attributes of pork meat could be predicted with correlation coefficients of calibration (r _c ) of 0.982 and 0.913, and correlation coefficients of prediction (r _p ) of 0.845 and 0.711 for pH and thawing loss, respectively. The results showed the possibility of developing a fast and reliable hyperspectral system for discrimination between fresh and frozen–thawed pork samples based on wavelet texture in the VIS/NIR wavelength range.     

基于高光谱图像特征融合的榛子水分含量测定

采用高光谱图像技术对榛子水分含量进行快速无损检测。采集200个榛子在400~1 000 nm波段的高光谱图像,提取榛子图像区域的平均光谱信息。利用K-S算法划分样品验证集和预测集,使用四种预处理方法对光谱进行预处理。通过竞争自适应加权算法（Competitive Adaptive Reweighted Sampling,CARS）和逐次投影法（Successive Projection Algorithm,SPA）进行光谱特征的提取;灰度共生矩阵法（Gray Level Co-occurrence Matrix,GLCM）提取图像的纹理特征;分别建立基于光谱特征,图像纹理特征以及两者串联融合的偏最小二乘回归（Partial Least Squares Regression,PLSR）和支持向量回归（Support Vector Regression,SVR）模型对榛子水分进行预测。结果表明,CARS和SPA算法能够有效选择特征波长并提升预测性能;图像特征能够对榛子水分进行预测,基于主成分图像提取的图像特征信息建立的模型预测效果更好。光谱图像特征融合能明显提高对榛子水分含量预测的准确率,CARS提取的特征波段结合主成分图像的纹理特征显示出了更好的效果,SVR模型的RMSECV为0.03,RC 为0.97,RMSEP为0.04,RP为0.96。利用高光谱图像和纹理特征能够对榛子水分进行有效预测,为榛子水分含量检测提供了新的方法。     

基于LDA_SVM的小麦质地检测方法研究

研究基于透射光图像的小麦质地检测方法,使用工业相机采集14种小麦种子的透射光图像,通过图像处理技术获取整粒小麦、胚乳和种胚代表性区域,并提取对应区域的颜色特征数据。分别运用PCA和LDA进行数据降维,并将降维前后的数据与支持向量机(SVM)、K近邻算法(KNN)和决策树模型(DT)3种分类器相结合建立分类模型,对不同品种小麦质地进行分类识别研究。结果表明：利用图像处理技术提取透射光全部特征,建立的LDA＿SVM模型分类正确率可以达到97%以上,证明透射光图像下通过机器学习对不同质地小麦快速分类鉴别是可行的。     

基于卷积神经网络的蓝印花布纹样基元分类

为更好地数字化传承与创新传统的蓝印花布纹样,并能单独提取构成纹样的图案基元并进行分类,提出一种基于卷积神经网络的纹样基元分类方法。首先,对采集的128张蓝印花布图像进行纹样基元提取,形成图像样本库,共21 212张。其次,从库中随机选取80%的图像样本作为训练集,20%作为测试集,利用5×5卷积核对训练样本进行卷积操作,将得到的特征图进行池化。通过3层卷积、3层池化及2层全连接层计算后,利用Softmax分类器得到12种分类结果。最后,通过基元样本的学习获取最佳网络模型参数,并取得较理想的分类结果。结果表明:提出的卷积神经网络模型对12种纹样基元的平均分类准确率达99.61%,检测平均准确率达98.5%,为蓝印花布纹样的研究提供了新思路。     

Evaluation of the effect of moisture content on cereal grains by digital image analysis

Physical appearance and kernel morphology significantly affect the grade of a harvested crop in addition to other factors such as test weight, percentage of foreign matter and constituent components. Moisture content of grain can potentially affect the physical appearance and kernel morphology. The objective of this study was to evaluate the effect of moisture content on the classification capability of colour, morphology and textural features of imaged grains. Colour images of individual kernels and bulk samples of three grain types, namely Canada Western Amber Durum (CWAD) wheat, Canada Western Red Spring (CWRS) wheat and barley were acquired using a machine vision system. The grain kernels were conditioned to 12%, 14%, 16%, 18% and 20% moisture contents before imaging. Previously developed algorithms were used to extract 123 colour, 56 textural features from bulk sample images and 123 colour, 56 textural, 51 morphological features from individual kernel images. The extracted features were analysed for the effect of moisture content. Statistical classifiers and a back propagation neural network model were used for classifying the grain bulk at different moisture contents. The colour and textural features of bulk grain images were affected by the moisture content more than that of the single kernel images.     

基于图像处理的小麦容重检测方法研究

为了实现图像处理技术对小麦容重影响因素的分析和容重的准确识别,研究了一种基于小麦图像特征和模式识别的小麦容重检测方法。采集不同容重小麦完整籽粒和籽粒横切面图像,对图像进行中值滤波、形态学运算、图像分割等处理,提取原图像与处理后图像的形态、颜色和纹理共3大类44个特征参数。最后采用逐步判别分析对提取的特征参数进行筛选,建立线性参数统计分类器和BP神经网络模型实现小麦不同容重的检测。结果表明,与小麦横切面图像特征相比,小麦完整籽粒图像的特征参数能更好的反映不同容重的差异;2种分类器对基于完整籽粒图像的小麦容重整体识别率均在95%以上。研究结果证明将图像处理技术应用于小麦容重检测识别是可行的。     

基于图像处理的白酒酒花轮廓检测

依据"看花摘酒"的传统经验,采用机器视觉代替人眼,通过CCD获取摘酒酒花的视频图像,并截取不同酒度酒花图像进行直方图均衡化、图像腐蚀等图像预处理,消除了高光噪声的影响,然后采用不同边缘检测算法对酒花轮廓进行了对比研究,采用OTSU算法与Canny边缘检测算法相结合的方法,较好地实现酒花与背景的分割,提取清晰的酒花边缘轮廓,通过对大清花与小清花图像的模式识别,为摘酒自动化提供了有效分级依据。该智能化的分级摘酒方法,能够提高分级摘酒工艺的稳定性和准确性,易于实现分级摘酒工序的智能自动化。