A novel nondestructive detection approach for seed cotton lint percentage using deep learning

Background The lint percentage of seed cotton is one of the most important parameters for evaluating seed cotton quality and affects its price.The traditional measuring method of lint percentage is labor-intensive and time-consuming;thus,an efficient and accurate measurement method is needed.In recent years,classification-based deep learning and computer vision have shown promise in solving various classification tasks.Results In this study,we propose a new approach for detecting the lint percentage using MobileNetV2 and transfer learning.The model is deployed on a lint percentage detection instrument,which can rapidly and accurately determine the lint percentage of seed cotton.We evaluated the performance of the proposed approach using a dataset comprising 66924 seed cotton images from different regions of China.The results of the experiments showed that the model with transfer learning achieved an average classification accuracy of 98.43%,with an average precision of 94.97%,an average recall of 95.26%,and an average F1-score of 95.20%.Furthermore,the proposed classification model achieved an average accuracy of 97.22%in calculating the lint percentage,showing no significant difference from the performance of experts(independent-sample t-test,t=0.019,P=0.860).Conclusion This study demonstrated the effectiveness of the MobileNetV2 model and transfer learning in calculating the lint percentage of seed cotton.The proposed approach is a promising alternative to traditional methods,providing a rapid and accurate solution for the industry.

Experimental study on the compressive strength of grouted concrete block masonry based on nondestructive detection methods

Existing nondestructive detection methods were adopted to test the compressive strength of grouted concrete block masonry,i.e.the rebound method,pulling-out method and core drilling method were employed to test the strength of block,mortar and grouted concrete,respectively.The suitability of these methods for the testing of strength of grouted concrete block masonry was discussed,and the comprehensive strength of block masonry was appraised by combining existing nondestructive or micro-destructive detection methods.The nondestructive detection test on 25 grouted concrete block masonry specimens was carried out.Experimental results show that these methods mentioned above are applicable for the strength detection of grouted concrete block masonry.Moreover,the formulas of compressive strength,detection methods and proposals are given as well.

Application of ultrasonic CT method in nondestructive detection of interior defects in large scale concrete structural member of bridge

The ultrasonic computed tomography(USCT)method is derived from the basic principles of X-ray section scanning.This method records the arriving times of ultrasonic wave between the probes and the sources to calculate the elastic wave velocity values in the section using the arrival times.Through analyzed the distribution of elastic wave velocity in aim area,the information of the strength and the homogeneity of the investigated zone could be got indirectly.The authors introduced the operational principle of USCT and a practical case of using this method to detect the interior defects in large scale concrete structural member.Compared with other exploration methods,this method is more efficient and accurate.

Nondestructive detection of infertile hatching eggs based on spectral and imaging information

In order to quickly distinguish infertile eggs from fertile eggs,the hyperspectral imaging technology consisting of imaging and spectral information was used for detecting the fertile information of eggs.Before hatching eggs were incubated,a hyperspectral imaging system(wavelength between 400 to 1000 nm)was used to acquire the images one-by-one manually.The characteristic information of ratios of length to short axis,elongation,roundness and the ratios of the yolk area to the whole area was extracted based on the images.The normalization method was used as the spectral data preprocessing,and then 155 spectral characteristic variables were extracted from 520 nm waveband through the correlation coefficient method.Principal component analysis(PCA)method was adopted to reduce the dimensions of image-spectrum fusion information;the top six principal components were extracted.Support vector machine(SVM)method was used to establish classification of fertile and infertile eggs models,which are based on image,spectrum and image-spectrum fusion information respectively.The accuracy rates of the SVM models were 84.00%,90.00%and 93.00%respectively.The experimental results show that the model based on image-spectrum fusion information technology is superior to the single information model.Hyperspectral transmission imaging technology is effective and feasible to detect the fertile hatching eggs before incubation.

Progress of Nondestructive Detection Technology on Remanufacturing Coatings

The defects of remanufacturing coatings,such as micro-cracks,micro-pores,oxide inclusion,and fatigue cracks producing in the service process have great influence on the qualities and lives of remanufacturing coatings.This paper summarizes several methods used for detecting coating defects,including ray method,ultrasonic method,eddy method,magnetic memory method,acoustic emission method.The advantages and limitations of the above methods are also discussed.The detection results by ray method are visualized,and it is easy to achieve qualitative,quantitative and locating detection,but this method has incipient fault and low detecting sensitivity.Ultrasonic detection can exactly locate defects,and it is sensitive to plane defects,but the detection has dead zones,and it is inconvenient to perform qualitative and quantitative measurement.Eddy method detects fast,but it only can detect conductive materials,and is difficult to achieve qualitative,quantitative and locating detection.Magnetic memory method is sensitive to ferromagnetic materials,but it can not detect nonferromagnetic materials and is difficult to attain quantitative measurement.Acoustic emission method has an extensive detection range and high sensitivity,but it has big noise interference,and is hard to achieve qualitative detection.In conclusion,it describes the application prospect of each method for detecting coating.

RESEARCH AND PRACTICE ON NONDESTRUCTIVE FLAW DETECTION INSTALLATION FOR WIRE-CORE BELT

Electromagnetic self-induction theory and computer are adopted and study of online monitoring technique for wire-core belt is conducted, the study shows that there is direct proportion between distance Ⅰ of broken ends and output volt Ⅴ, when Ⅰ ≥60 mm, Ⅴ keeps constantly, the running speed v of wire-core belt has no big effect on output volt Ⅴ, there is inverse proportion between the height h from probe to the surface of the belt and output volt Ⅴ, when h≥30 mm, Ⅴ tends to be zero. Based on the test result, on-line monitoring installation is developed, the practice proved that the accuracy of broken wire monitoring can be above 95%, the monitoring accuracy of joint twitch can be 0 .04 Ⅴ/mm.

High-sensitivity phase imaging eddy current magneto-optical system for carbon fiber reinforced polymers detection

This paper proposed a high-sensitivity phase imaging eddy current magneto-optical (PI-ECMO) system for carbon fiber reinforced polymer (CFRP) defect detection. In contrast to other eddy current-based detection systems, the proposed system employs a fixed position excitation coil while enabling the detection point to move within the detection region. This configuration effectively mitigates the interference caused by the lift-off effect, which is commonly observed in systems with moving excitation coils. Correspondingly, the relationship between the defect characteristics (orientation and position) and the surface vertical magnetic field distribution (amplitude and phase) is studied in detail by theoretical analysis and numerical simulations. Experiments conducted on woven CFRP plates demonstrate that the designed PI-ECMO system is capable of effectively detecting both surface and internal cracks, as well as impact defects. The excitation current is significantly reduced compared with traditional eddy current magneto-optical (ECMO) systems.

萘酚-苯并噻唑衍生物的合成及其琼脂糖凝胶对三文鱼新鲜度双通道无损检测

为了弥补单一比色法颜色区分不明显导致对鱼肉新鲜度的误判,本研究以萘酚为荧光团,通过与苯并噻唑盐缩合,合成萘酚-苯并噻唑衍生物(naphthol-benzothiazole derivative,HBO),其能够对有机胺比色和荧光双响应识别,并应用于三文鱼肉新鲜度的无损检测。HBO在EtOH/H2O(1∶9,V/V)体系中对12种胺都能产生明显响应,具有较低的检测限(精胺:0.4μmol/L),并可在活细胞中对精胺荧光成像。以琼脂糖为基质,将探针负载后制备传感凝胶HBO琼脂糖凝胶(HBO agarose gel,HBOAL),当凝胶与三文鱼鱼肉共同贮藏时,HBOAL日光下为浅红色,紫外光照射为黄色荧光时表明鱼肉已腐败。该凝胶对三文鱼鱼肉的3种等级颜色区分明显,评价结果准确,可用于实时监测三文鱼鱼肉的新鲜度。

抹茶品质指标的可见近红外光谱检测研究

目的建立适用于抹茶品质的可见近红外(visible-nearinfrared,Vis-NIR)光谱快速无损检测模型以实现多种品质指标的定量分析。方法通过Vis-NIR获取抹茶样本的光谱数据,使用一阶导数(first derivative,1^(st))光谱预处理方法,最后采用自助软收缩法(bootstrapping soft shrinkage,BOSS)、迭代变量子集优化法(iterative variable subset optimization,IVSO)和竞争性自适应重加权采样法(competitive adaptive reweighted sampling,CARS)筛选光谱特征变量,构建抹茶品质指标的偏最小二乘(partial least square,PLS)预测模型,探究光谱信息与茶多酚、游离氨基酸、酚氨比、咖啡碱和可溶性糖之间的定量关系。结果构建的Vis-NIR的CARS-PLS预测模型在抹茶品质指标含量预测方面均获得了最佳结果,预测相关系数(correlation coefficient in the prediction set,Rp)分别为0.9227、0.8906、0.9243、0.9381和0.9522;预测均方根误差(root mean square error in the prediction set,RMSEP)分别为0.867、0.337、0.557、0.216和0.440。结论本研究采用的Vis-NIR光谱技术综合了可见光、短波近红外和长波近红外的优势,在快速无损预测多种抹茶品质指标方面具有良好应用潜力,为抹茶品质的快速无损高效检测提供理论依据和技术支撑。

城镇聚乙烯燃气管道内检测机器人技术研究进展

国内城镇聚乙烯燃气管道主要采用埋地安装方式。在长期使用过程中,为及时发现管道老化、内部缺陷问题,需要使用管道机器人进行全面检测。文中按照机器人的运动类型分类介绍了国内外管道机器人的研究进展;介绍了聚乙烯管道的特性以及现有聚乙烯管道检测技术的发展现状;研究了目前城镇聚乙烯管道检测存在的主要问题,最后提出实现机器人与多种检测技术联合使用将对未来城镇聚乙烯燃气管道内检测的发展有重要意义。

空间偏移拉曼光谱分析技术及其在食品次表层检测中的应用

空间偏移拉曼光谱(SORS)是一种应用于浑浊/分层介质深层无损检测的新型拉曼光谱分析技术。分层物质不同深度的拉曼信号随传输距离增加对探测器的贡献产生规律性变化,SORS技术通过采集偏移激光入射点不同距离的拉曼光谱,结合最佳偏移距离信号增强、次表层谱峰识别、次表层信号分离等方法降低表层的干扰,获取纯净的次表层物质拉曼信号,然后借助拉曼的分子特异性实现无损的次表层物质定性和定量检测。介绍了空间偏移拉曼光谱技术的基本概念及测量技术,详细阐述了空间偏移拉曼光谱数据的分析方法及该技术在食品次表层检测中的应用,讨论了空间偏移拉曼光谱技术的限制条件并展望了该领域的未来发展。

激光波长对拉曼光谱检测地质成分的影响

随着激光技术的不断发展,以激光为光源的拉曼光谱检测技术由于其快速、无损、无接触探测等优势,已成为地质样品成分鉴别的一种重要途径。由于激光波长是样品拉曼效应的重要决定因素之一,进一步明确其在地质样品成分检测中的影响,能够为开展行星矿物成分研究提供重要参考。基于自主建立的532 nm/785 nm双色光源激光拉曼光谱探测系统,开展了不同激光波长对地质样品成分拉曼光谱的影响研究,获得了包括硝酸根、碳酸根、磷酸根、硅酸根等分子基团以及硫化物、氧化物等多种地质样品主要成分的拉曼特征光谱。通过对比表明,532 nm激发光具有更高的光子能量,能够检测更多的样品成分,但荧光效应较强,785 nm激发光存在拉曼信号强度较低问题,但是具有很好的荧光抑制效果,可根据实际样品种类进行最佳激发光波长的选择。

基于ZEMAX的柚果内部品质光谱无损检测光路仿真与试验

[目的]解决柚果Citrus maxima在无损检测中采集的光谱信息较弱、柚果内部品质检测不准确的问题。[方法]采用ZEMAX软件对光谱无损检测光路进行光学仿真,并搭建实物装置进行测试与验证。为进一步验证该光路对柚果内部品质的无损检测效果,采集400~1 100 nm波段132个柚果光谱数据,理化测定其可溶性固形物含量并进行训练建模和测试验证。[结果]最优光路参数为:灯珠数量为7、柚果与光源间距为2 cm,探测器与柚果间距为0.5 cm。通过Savitzky-Golay (SG)平滑和标准正态变换(Standard normal variate,SNV)预处理,竞争自适应重加权采样(Competitive adaptive reweighted sampling,CARS)特征选择,建立柚果糖度偏最小二乘回归(Partial least squares regression,PLSR)预测模型,训练集R2和RMSE分别为0.81和0.85,测试集R2和RMSE分别为0.81和0.89。[结论]本研究得到的柚果光谱无损检测光路参数可用于柚果内部品质的无损检测,检测效果较好,也为其他大尺寸水果内部无损检测提供了参考。

多传感器数据融合技术在管道无损检测中的应用

管道是石油和天然气最经济高效的运输方式之一,确保管道的安全性和完整性至关重要,管道无损检测是油气管道安全稳定运行的有力保障。在管道检测数据处理方面,低计算成本、高精度的多传感器数据融合模型在未来管道缺陷识别量化和轮廓重构方面具有广阔的应用前景。对多传感器数据融合技术及其在管道检测领域的应用进行了综述,讨论了多传感器数据融合技术在复合检测技术、管道缺陷识别以及量化方面的应用,对存在的问题和未来的研究方向进行了分析。

利用便捷式可见-近红外光谱仪和机器学习分辨霉变小麦及霉变程度

[目的/意义]可见-近红外光谱可对小麦霉变情况快速无损检测,但是高分辨率光谱仪价格高、体积大,不利于在农业环境中推广,因此通过对低分辨率光谱数据进行优化处理,以期接近高分辨率光谱仪分辨霉变小麦的效果。[方法]使用可见-近红外农产品检测仪(型号VNIAPD,分辨率1.6 nm)和复享光纤光谱仪(型号SIN02040,分辨率0.19 nm)采集100份小麦样本的新鲜状态以及不同霉变状态的光谱数据。首先对SINO2040光谱进行裁剪,让其和VNIAPD波长保持一致,均为640~1 050 nm;然后对其使用标准差标准化(Standard Deviation Normalization,SDN)、标准正态变换(Standard Normal Variation,SNV)、均值中心化(Mean Centrality,MC)、一阶导数(First-order Derivatives,1ST)、Savitzky-Golay平滑(Savitzky-Golay Smoothing,SG)、多元散射校正(Multiple Scattering Correction,MSC)等多种预处理方法处理并使用离群点检测算法(Local Outlier Factor,LOF)筛选出离群点并剔除;其次使用连续投影算法(Sequential Projection Algorithm,SPA)和最小绝对收缩和选择算法(Least Absolute Shrinkage and Selection Operator,LASSO)对预处理后的光谱进行特征波长提取;最后分别采用K近邻算法(K-Nearest Neighbor,KNN)、支持向量机(Support Vector Machine,SVM)、随机森林(Random Forests,RF)和朴素贝叶斯(Na?ve-Bayes)、后向传播神经网络(Back Propagation Neural Network,BPNN)、深度神经网络(Deep Neural Networks,DNN)6种算法对特征波长光谱进行建模分析,从而分辨霉变小麦以及区分霉变程度。[结果和讨论]BPNN、DNN两种神经网络模型的测试集准确率均可达到100%,但是建模时间长,模型内存大;而KNN、SVM、RF和Na?ve-Bayes浅层模型的测试集准确率为93.18%~100%,建模速度快、模型内存小。本研究光谱仪VNIAPD在光学参数(光学分辨率1.6 nm)低于SINO2040的光学参数(光学分辨率0.19 nm)且成本更低的情况下,检测准确率到达同一水平。[结论]本研究通过对比光谱数据的不同预处理方法从而找出了对应算法的最佳数据优化选择,使低分辨率光谱仪VNIAPD检测霉变小麦性能可以追平高分辨率光谱仪SINO2040,为基于可见-近红外光谱的小麦霉变低成本无损检测提供了新选择。

基于多项式混沌展开法的涡流无损检测高效元模型辅助探测概率的分析

探测概率对于量化涡流无损检测系统的检测能力非常重要。模型辅助探测概率的参数需要大量试验或仿真数据才可以确定,往往难以实现。应用基于退化核函数加速的边界元法的数值模型,提出使用基于普通最小二乘法的多项式混沌展开算法的元模型提升三维涡流无损检测问题探测概率的效率。通过有限截面线圈检测金属板面槽的算例,引入线圈位置和提离距离为不确定传播的参数,测试结果表明,该元模型预测的模型参数与基于退化核函数加速的边界元法物理模型计算的模型参数相对误差在1%以内,极大降低了所需的计算开销。

枇杷糖度无损检测及可视化研究

糖度是评价枇杷内部品质的重要指标。采集枇杷高光谱图像并提取不同感兴趣区域的平均光谱,分析光谱提取区域对枇杷糖度预测精度的影响规律。使用标准正态变换对原始光谱作预处理,分别利用竞争性自适应重加权算法和连续投影算法(Successive Projection Algorithm, SPA)筛选特征光谱,为枇杷糖度建立多元线性回归(Multi Linear Regression, MLR)预测模型。最后采用伪彩色技术实现枇杷糖度含量分布的可视化。结果表明:以整个样本为感兴趣区域,经连续投影算法筛选特征光谱建立的多元线性回归模型具有最好的预测性能(决定系数R_(p)^(2)=0.822,均方根误差RMSE_(p)=0.435,剩余预测偏差RPD=2.407),基于SPA-MLR模型计算枇杷每个像素点的糖度,生成可视化分布图。表明采用高光谱成像技术可实现枇杷糖度的快速无损检测及可视化研究。

基于磁导率无损检测的双U型探头仿真优化研究

基于磁导率无损检测原理,设计了一种双U型检测探头。以20钢试件为研究对象,通过模拟仿真分析了双U型探头在检测过程中的关键影响参数与检测灵敏度的关系,对关键影响参数进行了优化选取,又通过试验研究了拉应力对检测灵敏度的影响。结果表明,检测灵敏度随着激励电压幅值的升高而增大,随着激励频率的提高先增大后趋于稳定,随着激励线圈匝数的增加而下降,随着拉应力的增大而减小,双U型探头相比于传统U型探头具有更广泛的分析检测能力。该仿真研究为后续探头的选取制作提供了设计理念和理论依据。

Nondestructive determination of soluble solids and firmness in mix-cultivar melon using near-infrared CCD spectroscopy

Nondestructive evaluation of melon quality is in great need of comprehensive study.Soluble solids content(SSC)and firmness are the two indicators of melon internal quality that mostly a®ect consumer acceptance.To provide guidance for fruit classification,internal quality standards was preliminarily established through sensory test,as:Melon with SSC over 12
Brix,firmness 4–5.5 kgf·cm^(-2)2 were considered as satisfactory class sample;and SSC over 10
Brix,¯rmness 3.5–6.5 kgf·cm^(-2) as average class sample.The near infrared(NIR)nondestructive detection program was set as spectra collected from the stylar-end,Brix expressed by the average SSC of inner and outer mesocarp,each cultivar of melon was detected with its own optimum integration time,and the second derivative algorithm was used to equalize them.Using wavelength selected by genetic algorithms(GA),a robust SSC model of mix-cultivar melon was established,the root mean standard error of cross-validation(RMSECV)was 0.99 and the ratio performance deviation(RPD)nearly reached 3.0,which almost could meet the accuracy requirement of 1.5
Brix.Firmness model of mix-cultivar melon was acceptable but inferior.

应用高光谱技术及MLSPTSVM模型检测热损伤大豆

进口大豆在运输过程中极易因储藏温度过高而造成热损伤,加剧大豆蛋白及油脂的品质恶化,对大豆质量造成影响。利用高光谱图像技术和多元最小二乘递归投影孪生支持向量机(MLSPTSVM)对大豆的热损伤进行检测。应用高光谱图像采集系统在400~1000 nm范围内获取正常大豆、轻度热损伤、重度热损伤大豆的光谱图像。采用多种预处理方法进行光谱预处理,对预处理方法提高模型检测性能的有效性进行分析。结果表明,多元散射校正预处理搭配线性核的MLSPTSVM模型、原始光谱数据搭配非线性核的MLSPTSVM模型均能达到100%检测准确率,相较于经典检测模型具有显著优势。在实验样本数量大幅减少的情况下,应用线性核的模型检测准确率仍能达到100%。因此,结合MLSPTSVM模型的高光谱图像检测方法可有效地提高热损伤大豆检测精度,且具有良好的鲁棒性。