CN105044022B - A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness - Google Patents
A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness Download PDFInfo
- Publication number
- CN105044022B CN105044022B CN201510477871.7A CN201510477871A CN105044022B CN 105044022 B CN105044022 B CN 105044022B CN 201510477871 A CN201510477871 A CN 201510477871A CN 105044022 B CN105044022 B CN 105044022B
- Authority
- CN
- China
- Prior art keywords
- hardness
- sample
- wheat
- data
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 241000209140 Triticum Species 0.000 title claims abstract description 66
- 235000021307 Triticum Nutrition 0.000 title claims abstract description 66
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 17
- 238000011156 evaluation Methods 0.000 title 1
- 238000001514 detection method Methods 0.000 claims abstract description 35
- 238000005516 engineering process Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 6
- 230000001066 destructive effect Effects 0.000 claims abstract description 5
- 230000003595 spectral effect Effects 0.000 claims description 30
- 238000012937 correction Methods 0.000 claims description 27
- 238000003062 neural network model Methods 0.000 claims description 13
- 238000001228 spectrum Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000005303 weighing Methods 0.000 claims description 7
- 238000004497 NIR spectroscopy Methods 0.000 claims description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 5
- 238000010298 pulverizing process Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 abstract description 8
- 235000013339 cereals Nutrition 0.000 abstract description 5
- 238000011897 real-time detection Methods 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 abstract 1
- 239000000463 material Substances 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 239000003153 chemical reaction reagent Substances 0.000 description 3
- 238000003801 milling Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 244000098345 Triticum durum Species 0.000 description 2
- 235000007264 Triticum durum Nutrition 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000002835 absorbance Methods 0.000 description 2
- 238000009395 breeding Methods 0.000 description 2
- 230000001488 breeding effect Effects 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007542 hardness measurement Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 229920000832 Cutin Polymers 0.000 description 1
- 229920002472 Starch Polymers 0.000 description 1
- 244000098338 Triticum aestivum Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000009659 non-destructive testing Methods 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
- 235000019698 starch Nutrition 0.000 description 1
- 239000008107 starch Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
本发明公开了一种基于近红外光谱技术快速无损检测小麦硬度的方法及应用，属于谷物质量检测技术领域。本发明所提供的方法是通过建立小麦硬度的近红外监测模型，利用近红外光谱仪扫描待检小麦样品，获取样品近红外光谱曲线，在对近红外光谱曲线进行数据处理后，再根据所得数据利用检测模型确定待检小麦样品的硬度。本发明所提供的方法具有成本低、无污染、检测速度快、客观性高，并可实现在线分析，不受地域限制，可以进行实时检测。
The invention discloses a method and application for fast and non-destructive detection of wheat hardness based on near-infrared spectrum technology, belonging to the technical field of grain quality detection. The method provided by the present invention is to establish a near-infrared monitoring model of wheat hardness, use a near-infrared spectrometer to scan a wheat sample to be inspected, and obtain a near-infrared spectrum curve of the sample. After data processing the near-infrared spectrum curve, use the The detection model determines the hardness of the wheat sample to be tested. The method provided by the invention has the advantages of low cost, no pollution, fast detection speed and high objectivity, and can realize on-line analysis without geographical restriction, and can perform real-time detection.
Description
技术领域technical field
本发明涉及一种基于近红外光谱技术快速无损检测小麦硬度的方法及应用，属于谷物质量检测技术领域。The invention relates to a method and application for fast and non-destructive detection of wheat hardness based on near-infrared spectroscopy, and belongs to the technical field of grain quality detection.
背景技术Background technique
小麦籽粒质地的软、硬是评价小麦加工品质和食用品质的一项重要指标，并与小麦育种和贸易价格等多方面密切相关，硬度是国内外小麦市场分类和定价的重要依据之一，也是各国的育种家重要的育种目标之一。小麦硬度被定义为破碎籽粒时所受到的阻力，即破碎籽粒时所需要的力。硬度是由胚乳细胞中蛋白质基质和淀粉之间的结合强度决定的，这种结合强度受遗传控制。小麦的制粉品质与籽粒硬度密切相关，小麦硬度的变化可使小麦制粉流程中各系统在制品数量和质量、各设备工作效率、加工动力消耗等产生很大变化。预先的测定原料小麦的硬度，对于及时调整制粉工艺流程和相应的技术参数，确定配麦方案、保持流程的物料平衡和生产稳定、提高生产效率等，都具有重要的技术指导意义。按照硬度的不同，国外将小麦分为三个明显的硬度等级：软麦、硬麦和硬粒小麦。The softness and hardness of wheat grain texture is an important indicator for evaluating wheat processing quality and eating quality, and is closely related to wheat breeding and trade prices. Hardness is one of the important basis for classification and pricing of wheat at home and abroad. One of the most important breeding goals for breeders. Wheat hardness is defined as the resistance to breaking the kernel, ie the force required to break the kernel. Firmness is determined by the strength of the bond between the protein matrix and starch in the endosperm cells, which is genetically controlled. Wheat milling quality is closely related to grain hardness, and changes in wheat hardness can cause great changes in the quantity and quality of WIP of each system in the wheat milling process, the working efficiency of each equipment, and processing power consumption. Pre-determining the hardness of raw wheat has important technical guiding significance for timely adjustment of milling process and corresponding technical parameters, determination of wheat blending scheme, maintenance of process material balance and production stability, and improvement of production efficiency. According to the difference in hardness, wheat is divided into three obvious hardness grades abroad: soft wheat, hard wheat and durum wheat.
利用传统方法测定小麦的硬度，均费工、费时或手续繁杂，且不能在线检测。近几年，随着工业、农业以及药学等行业对更快速、低消费以及无损检测的追求，光谱技术成为人们研究的重点，受到越来越多的重视，它的很多应用已经被列入美国谷物化学家协会标准方法(AACC)。近红外(NIR)检测因具有检测速度快、可实现对多项目同时检测、准确度高、样品代表性好和成本低等特点，使得近红外技术成为一种应用潜力巨大的绿色环保分析技术。应用NIR法测定小麦硬度国外已有报道，AACC推荐方法采用1680和2230nm两个波长点，并给出三个回归系数，但是它对所用仪器和样品粉碎物进行标准化处理，否则测定误差较大。所以目前实际使用NIR法测定小麦硬度的报道并不多，人们仍致力于研究其它方法测定小麦硬度，然而应用NIR法测定小麦其它品质以及对小麦按一定标准进行分类的技术和方法已经很成熟，并且已经能达到较高的精度。Using traditional methods to measure the hardness of wheat is labor-intensive, time-consuming or complicated procedures, and cannot be detected online. In recent years, with the pursuit of faster, lower consumption and non-destructive testing in industries such as industry, agriculture and pharmacy, spectroscopy technology has become the focus of people's research and has received more and more attention. Many of its applications have been listed in the United States. Association of Cereal Chemists Standard Method (AACC). Near-infrared (NIR) detection has the characteristics of fast detection speed, simultaneous detection of multiple items, high accuracy, good sample representativeness and low cost, making near-infrared technology a green and environmentally friendly analysis technology with great application potential. Application of NIR method to measure wheat hardness has been reported abroad. The AACC recommended method uses two wavelength points of 1680 and 2230nm, and gives three regression coefficients, but it standardizes the instruments used and the crushed samples, otherwise the measurement error will be large. Therefore, there are not many reports on the actual use of NIR method to measure wheat hardness. People are still committed to studying other methods to measure wheat hardness. However, the techniques and methods of using NIR method to measure other qualities of wheat and to classify wheat according to certain standards are very mature. And has been able to achieve higher precision.
发明内容Contents of the invention
为解决现有技术检测周期长、检测成本高效率低、操作繁琐的问题，本发明提供了一种基于近红外光谱技术快速无损检测小麦硬度的方法，所采取的技术方案如下：In order to solve the problems of long detection period, high detection cost, low efficiency, and cumbersome operation in the prior art, the present invention provides a method for rapid and non-destructive detection of wheat hardness based on near-infrared spectroscopy. The technical scheme adopted is as follows:
本发明的目的在于提供一种基于近红外光谱技术快速无损检测小麦硬度的方法。该方法是通过建立小麦硬度的近红外检测模型，利用近红外光谱仪扫描待检小麦样品，获取样品近红外光谱数据，在对近红外光谱数据进行数据处理后，再根据所得数据利用检测模型确定待检小麦样品的硬度。The object of the present invention is to provide a method for fast and non-destructive detection of wheat hardness based on near-infrared spectroscopy. The method is to establish a near-infrared detection model of wheat hardness, use a near-infrared spectrometer to scan the wheat sample to be inspected, and obtain the near-infrared spectral data of the sample. Check the hardness of wheat samples.
所述方法的步骤如下：The steps of the method are as follows:
1)利用近红外光谱仪扫描小麦样品，获得样品的近红外光谱数据；1) Utilize the near-infrared spectrometer to scan the wheat sample to obtain the near-infrared spectrum data of the sample;
2)利用硬度指数法测定步骤1)所述小麦样品的硬度，并建立小麦硬度数据库；2) Utilize the hardness index method to measure the hardness of the wheat sample described in step 1), and establish a wheat hardness database;
3)利用优化后的SPXY算法将步骤1)所得的近红外光谱数据划分为校正集和预测集，对校正集和预测集光谱数据进行预处理，获得预处理数据；3) using the optimized SPXY algorithm to divide the near-infrared spectral data obtained in step 1) into a correction set and a prediction set, and preprocessing the correction set and the prediction set spectral data to obtain preprocessed data;
4)以步骤3)所得预处理后的校正集光谱数据矩阵作为模型的输入，以校正集对应的小麦硬度数据作为模型的期望输出，建立初步径向基函数神经网络模型，再利用步骤3)所得预处理后预测集的数据对初步径向基函数神经网络模型进行验证和校正，最后获得检测模型；4) With the correction set spectral data matrix after the pretreatment gained in step 3) as the input of the model, with the wheat hardness data corresponding to the correction set as the expected output of the model, establish a preliminary radial basis function neural network model, and then use step 3) The data of the obtained pre-processed prediction set is verified and corrected for the preliminary radial basis function neural network model, and finally the detection model is obtained;
5)利用近红外光谱仪扫描待检小麦样品，获取样品近红外光谱数据，对所得近红外光谱数据进行预处理后，利用步骤4)所得的检测模型测定待测小麦样品的硬度。5) Use the near-infrared spectrometer to scan the wheat sample to be tested to obtain the near-infrared spectrum data of the sample. After preprocessing the obtained near-infrared spectrum data, use the detection model obtained in step 4) to determine the hardness of the wheat sample to be tested.
优选地，步骤1)所述利用近红外光谱仪扫描小麦样品，扫描范围为400-2498nm，分辨率为8nm，扫描次数为2次。Preferably, step 1) uses a near-infrared spectrometer to scan the wheat sample, the scanning range is 400-2498nm, the resolution is 8nm, and the number of scanning is 2 times.
优选地，步骤2)所述硬度指数法，是取24.99-25.01g样品，利用硬度测定仪粉碎并测定样品硬度，粉碎时间为50s，最后通过自动称量系统称量并计算最终结果或手工称量并根据以下公式计算结果：Preferably, the hardness index method in step 2) is to take a 24.99-25.01g sample, use a hardness tester to pulverize and measure the hardness of the sample, the pulverization time is 50s, and finally weigh and calculate the final result by an automatic weighing system or manually weigh and calculate the result according to the following formula:
式中，HI(％)表示校正至水分12％，环境温度25℃时的硬度指数，m1(g)表示粉碎后通过筛网的样品质量，w(％)表示样品的水分含量，k1表示水分校正系数，k2表示温度校正系数。In the formula, HI(%) represents the hardness index corrected to 12% moisture and the ambient temperature is 25°C, m 1 (g) represents the mass of the sample passing through the sieve after crushing, w(%) represents the water content of the sample, k 1 Indicates the moisture correction coefficient, k 2 represents the temperature correction coefficient.
优选地，步骤3)所述近红外光谱数据的预处理，是利用马氏距离法剔除光谱数据中的异常值，再利用优化后的SPXY算法对光谱数据进行划分。Preferably, the preprocessing of the near-infrared spectral data in step 3) is to use the Mahalanobis distance method to remove abnormal values in the spectral data, and then use the optimized SPXY algorithm to divide the spectral data.
优选地，步骤3)所述优化后的SPXY算法，是先将欧式距离最远的两点数据选进校正集，再利用SPXY方法从剩余样品中选取其余校正集数据。Preferably, the optimized SPXY algorithm described in step 3) first selects the data of the two points with the longest Euclidean distance into the calibration set, and then uses the SPXY method to select the rest of the calibration set data from the remaining samples.
优选地，步骤3)所述的预处理，为标准正态变量变换、一阶导数、二阶导数和连续投影算法处理光谱中的一种或几种。Preferably, the preprocessing described in step 3) is one or more of standard normal variable transformation, first derivative, second derivative and continuous projection algorithm processing spectrum.
优选地，步骤4)所述检测模型为采用newrbe函数创建的一个精密径向基函数神经网络(RBF)模型。RBF网络包括输入层、隐层和输出层，预处理后的校正集光谱数据矩阵作为模型的输入，以校正集对应的小麦硬度数据作为模型的期望输出，均方误差默认为0，SPREAD值取1200，隐层基函数采用高斯函数。径向基网络是一种局部逼近网络，且其一个最大优点是通过线性最小二乘法得到权值W，因此具有较快的处理速度。Preferably, the detection model in step 4) is a precise radial basis function neural network (RBF) model created by using the newrbe function. The RBF network includes an input layer, a hidden layer, and an output layer. The preprocessed correction set spectral data matrix is used as the input of the model, and the wheat hardness data corresponding to the correction set is used as the expected output of the model. The default mean square error is 0, and the SPREAD value is set to 1200, the hidden layer basis function adopts Gaussian function. The radial basis network is a kind of local approximation network, and one of its biggest advantages is that the weight W is obtained by the linear least square method, so it has a faster processing speed.
所述方法的具体步骤如下：The concrete steps of described method are as follows:
1)利用近红外光谱仪扫描小麦样品，扫描范围为400-2498nm，分辨率为8nm，扫描次数为2次，获得样品的近红外光谱曲线；1) Utilize the near-infrared spectrometer to scan the wheat sample, the scanning range is 400-2498nm, the resolution is 8nm, and the number of scans is 2 times to obtain the near-infrared spectrum curve of the sample;
2)取24.99-25.01g步骤1)所述小麦样品，利用硬度测定仪粉碎并测定样品硬度，粉碎时间为50s，最后通过自动称量系统称量并计算最终硬度，并建立小麦硬度数据库；2) Take 24.99-25.01g of the wheat sample described in step 1), use a hardness tester to pulverize and measure the hardness of the sample, the pulverization time is 50s, and finally weigh and calculate the final hardness by an automatic weighing system, and establish a wheat hardness database;
3)利用马氏距离法剔除光谱曲线中的异常值，再利用优化后的SPXY算法将所得的近红外光谱数据划分为校正集和预测集，利用标准正态变量变换优化校正集和预测集光谱数据，同时利用连续投影算法对校正集和预测集光谱数据进行压缩处理，获得预处理数据；所述优化后的SPXY算法，是先将欧式距离最远的两点数据选进校正集，在从剩余样品中选取其余校正集数据；3) Use the Mahalanobis distance method to remove the outliers in the spectral curve, and then use the optimized SPXY algorithm to divide the obtained near-infrared spectral data into a correction set and a prediction set, and use the standard normal variable transformation to optimize the correction set and prediction set spectrum At the same time, the continuous projection algorithm is used to compress the spectral data of the correction set and the prediction set to obtain preprocessed data; the optimized SPXY algorithm first selects the two point data with the farthest Euclidean distance into the correction set, and then from Select the remaining calibration set data from the remaining samples;
4)以所得预处理后的校正集光谱数据矩阵作为模型的输入，以校正集对应的小麦硬度数据作为模型的期望输出，均方误差默认为0，建立初步径向基函数神经网络模型，再利用步骤3)所得预处理后预测集的数据对初步径向基函数神经网络模型进行验证和校正，最后获得检测模型；4) Take the obtained preprocessed correction set spectral data matrix as the input of the model, take the wheat hardness data corresponding to the correction set as the expected output of the model, and the default mean square error is 0, establish a preliminary radial basis function neural network model, and then Utilize the data of the prediction set after the preprocessing of the gained step 3) to carry out verification and correction to the preliminary radial basis function neural network model, and finally obtain the detection model;
5)根据步骤1)所述方法利用近红外光谱仪扫描待检小麦样品，获取样品近红外光谱数据，再根据步骤3)所述方法对近红外光谱数据进行预处理后，利用步骤4)所得的检测模型测定待测小麦样品的硬度。5) Utilize the near-infrared spectrometer to scan the wheat sample to be inspected according to the method described in step 1), obtain the near-infrared spectrum data of the sample, and then preprocess the near-infrared spectrum data according to the method described in step 3), use the obtained The detection model determines the hardness of the wheat sample to be tested.
所述任一方法均可用于检测小麦的硬度。Either method can be used to detect the hardness of wheat.
本发明获得的有益效果如下：The beneficial effects that the present invention obtains are as follows:
1.本发明的方法具有成本低、无污染的特点。利用近红外光谱技术不需要任何试剂，不消耗样品，无需排放污染物。与常规方法相比，既能降低成本，又可以保护环境，是一种“绿色分析”技术。1. The method of the present invention has the characteristics of low cost and no pollution. The use of near-infrared spectroscopy does not require any reagents, does not consume samples, and does not emit pollutants. Compared with conventional methods, it can not only reduce costs, but also protect the environment. It is a "green analysis" technology.
2.本发明的方法的检测速度快，检测小麦种类多，使用小，常规测量耗时耗力。常规方法建模时样本数据量过大，存在较多的干扰因素，影响模型的稳定性，本发明通过对光谱数据进行预处理，压缩光谱数据来进一步的增强模型的稳定性。模型建立后，测量样品的时间只需几十秒钟，大幅度提高检测效率。2. The detection speed of the method of the present invention is fast, there are many types of wheat to be detected, the use is small, and conventional measurement is time-consuming and labor-intensive. When the conventional method is used for modeling, the amount of sample data is too large, and there are many interference factors, which affect the stability of the model. The present invention further enhances the stability of the model by preprocessing the spectral data and compressing the spectral data. After the model is established, it only takes tens of seconds to measure the sample, which greatly improves the detection efficiency.
3.本发明的方法客观性较高。传统小麦硬度测量法如角质率法等需要人为测量，误差较大，近红外检测方法基本上都是机器操作，基本上消除了人为误差。3. The method of the present invention has higher objectivity. Traditional wheat hardness measurement methods, such as cutin rate method, require manual measurement and have large errors. Near-infrared detection methods are basically machine-operated, which basically eliminates human errors.
4.可以实现在线分析。近红外检测方法不受地域限制，可以进行实时检测。4. Online analysis can be realized. The near-infrared detection method is not subject to geographical restrictions and can perform real-time detection.
附图说明Description of drawings
图1是本发明检测的流程图。Fig. 1 is the flow chart of the detection of the present invention.
图2是小麦样品近红外光谱图；Fig. 2 is near-infrared spectrogram of wheat sample;
(其中横坐标代表光谱的波长，纵坐标代表光谱的吸光度，每一条光谱对应每个样品的硬度)。(wherein the abscissa represents the wavelength of the spectrum, the ordinate represents the absorbance of the spectrum, and each spectrum corresponds to the hardness of each sample).
图3是连续投影算法选择的光谱点；Figure 3 is the spectral points selected by the continuous projection algorithm;
(其中横坐标代表波点，纵坐标是吸光度值，图中的方块代表筛选得到的波点位置)。(wherein the abscissa represents the wave point, the ordinate is the absorbance value, and the squares in the figure represent the wave point position obtained by screening).
图4是RBF网络结构示意图。Fig. 4 is a schematic diagram of the RBF network structure.
具体实施方式Detailed ways
下面结合具体实施例对本发明做进一步说明，但本发明不受实施例的限制。The present invention will be further described below in conjunction with specific examples, but the present invention is not limited by the examples.
以下实施例所用材料、试剂、方法和仪器，未经特殊说明，均为本领域常规材料、试剂、方法和仪器，本领域技术人员均可通过商业渠道获得。The materials, reagents, methods and instruments used in the following examples, unless otherwise specified, are conventional materials, reagents, methods and instruments in the art, and those skilled in the art can obtain them through commercial channels.
实施例1Example 1
1、本实施例所选用的样本是由黑龙江省农科院农产品质量安全研究所从全国各地小麦主产区收集的2013年小麦样品，共收集样本111个。1. The samples selected in this embodiment are the 2013 wheat samples collected by the Institute of Agricultural Product Quality and Safety of Heilongjiang Academy of Agricultural Sciences from the main wheat producing areas all over the country, and a total of 111 samples were collected.
利用近红外光谱分析仪获得光谱数据，具体过程为：扫描时温度控制在室温25℃左右；装样时样品装入样品杯后需将杯口刮平，尽量使样品表面与样品杯边缘向平；Use a near-infrared spectrometer to obtain spectral data. The specific process is as follows: when scanning, the temperature is controlled at room temperature at about 25°C; when loading the sample, the cup mouth needs to be scraped flat after the sample is loaded into the sample cup, so that the sample surface is as flat as possible with the edge of the sample cup;
将样品放入一个直径35mm，10mm深的原型小槽中进行扫描，扫描间隔是8nm，扫描2次，产生262个光谱点，如图2为4个小麦的近红外光谱图。Put the sample into a small prototype groove with a diameter of 35mm and a depth of 10mm for scanning. The scanning interval is 8nm, and the scanning is performed twice to generate 262 spectral points. Figure 2 shows the near-infrared spectra of 4 wheat.
2、对经近红外光谱分析仪扫描后的小麦样品进行常规实验，准确称取制备好的样品(25.00±0.01)g；打开硬度测定仪端盖，将粉碎系统转子的一个型腔(两刀之间的凹部)向上对准进料口，关闭并锁好端盖；打开进料斗盖，将称取好的样品全部倒入进料斗中，关闭进料斗盖；开启测定仪，样品粉碎50s后，自动停机；待仪器停稳后打开端盖，小心将接料斗、筛网系统一起取出，按照仪器说明书的规定，将筛网上的留存物清扫干净。清扫中要防止筛网系统与接料斗分离，以免筛网上的留存物掉入接料斗中和(或)接料斗中的物质撒出；连同接料斗、筛网系统一起称量晒下物，扣除接料斗、筛网系统质量后得到筛下物质量m1，精确至0.01g；将仪器粉碎系统、接料斗、筛网系统等清扫干净，以备下次测定用。配备称量计算系统的仪器，称量后自动计算并打印出结果。未配备称量计算系统的，按公式(1)计算：2. Carry out a routine experiment on the wheat sample scanned by the near-infrared spectrometer, accurately weigh the prepared sample (25.00 ± 0.01) g; open the end cover of the hardness tester, and put a cavity (two knives) Align the feed port upwards, close and lock the end cover; open the feed hopper cover, pour all the weighed samples into the feed hopper, close the feed hopper cover; open the analyzer, the sample After crushing for 50 seconds, it will stop automatically; after the instrument has stopped, open the end cover, carefully take out the receiving hopper and the screen system together, and clean up the residue on the screen according to the provisions of the instrument manual. During cleaning, it is necessary to prevent the screen system from being separated from the receiving hopper, so as to prevent the remaining material on the screen from falling into the receiving hopper and (or) the material in the receiving hopper being spilled; After receiving the mass of the hopper and the sieve system, the mass of the under-sieve material m1 is obtained, accurate to 0.01g; clean the crushing system, hopper, and sieve system of the instrument for the next measurement. The instrument is equipped with a weighing calculation system, which automatically calculates and prints out the results after weighing. If it is not equipped with a weighing calculation system, it shall be calculated according to formula (1):
式中，HI(％)表示校正至水分12％，环境温度25℃时的硬度指数，m1(g)表示粉碎后通过筛网的样品质量，w(％)表示样品的水分含量，k1表示水分校正系数，k2表示温度校正系数。所有的实验步骤严格按照国家标准GB/T 21304执行，均由农科院专业研究人员完成。In the formula, HI (%) represents the hardness index corrected to 12% moisture and the ambient temperature is 25°C, m1 (g) represents the mass of the sample that passes through the sieve after crushing, w (%) represents the water content of the sample, and k1 represents the water content Correction coefficient, k2 represents the temperature correction coefficient. All the experimental steps are carried out in strict accordance with the national standard GB/T 21304, and are completed by professional researchers of the Academy of Agricultural Sciences.
3、采用马氏距离法挑选样本中的异常样品，并将其剔除，111个样品经处理后剩余108个样品，然后采用优化后的SPXY方法对整个样本集进行划分。最终选择84个具有代表性的样本作为校正集，用来建立模型，校正集有24个样品。3. Use the Mahalanobis distance method to select the abnormal samples in the sample and remove them. After 111 samples are processed, 108 samples remain, and then use the optimized SPXY method to divide the entire sample set. Finally, 84 representative samples are selected as the calibration set, which is used to build the model, and the calibration set has 24 samples.
4、利用标准正态变量变换SNV对校正集光谱数据进行优化，以消除校正散射的影响。4. The standard normal variable transformation SNV is used to optimize the spectral data of the calibration set to eliminate the influence of calibration scatter.
5、利用连续投影算法对光谱数据进行压缩处理，得到的波长点47个如图3所示。5. Using the continuous projection algorithm to compress the spectral data, 47 wavelength points are obtained as shown in Figure 3.
6、利用校正集样品建立径向基函数(RBF)神经网络模型(图4)，预处理后的光谱数据矩阵作为模型的输入，这些光谱所对应的样品的硬度作为模型的输出，均方误差默认为0，经过多次测试，当SPREAD值取1200时，模型的效果比较好,隐层基函数采用高斯函数，模型的判别系数R2为0.90，预测标准差SEP为3.02，相对分析误差RPD为3.11，模型的R2越趋近于1，模型的稳定性越好。6. Use the calibration set samples to establish a radial basis function (RBF) neural network model (Figure 4). The preprocessed spectral data matrix is used as the input of the model, and the hardness of the samples corresponding to these spectra is used as the output of the model. The mean square error The default is 0. After many tests, when the SPREAD value is 1200, the effect of the model is better. The hidden layer base function adopts Gaussian function. The discriminant coefficient R 2 of the model is 0.90, the prediction standard deviation SEP is 3.02, and the relative analysis error RPD is 3.11, the closer the R2 of the model is to 1, the better the stability of the model.
根据模型的判别系数、预测标准差、相对分析误差来不断调整模型的参数。The parameters of the model are constantly adjusted according to the discriminant coefficient, forecast standard deviation, and relative analysis error of the model.
实施例2Example 2
1、步骤1-3同实施例1中的步骤1-3相同。1. Steps 1-3 are the same as Steps 1-3 in Example 1.
2、利用未处理过的校正集样品的原始光谱建立径向基函数(RBF)神经网络模型，利用预测集样品来检验RBF模型，模型的R2为0.79，RPD为2.19，SEP为4.30。2. Establish a radial basis function (RBF) neural network model using the original spectrum of the unprocessed calibration set sample, and use the prediction set sample to test the RBF model. The R2 of the model is 0.79, the RPD is 2.19, and the SEP is 4.30.
实施例3Example 3
1、步骤1-3同实施例1中的步骤1-3相同。1. Steps 1-3 are the same as Steps 1-3 in Example 1.
2、利用标准正态变量变换SNV对所有样品的光谱数据进行优化，以消除校正散射的影响，再通过数学计算法得到二阶导数光谱。2. Optimize the spectral data of all samples by using the standard normal variable transformation SNV to eliminate the influence of corrected scattering, and then obtain the second derivative spectrum through mathematical calculation.
3、利用校正集样品建立偏最小二乘(PLS)模型，利用预测集样品来验证并校正PLS模型，模型的R2为0.85，RPD为2.57，SEP为3.66。3. The partial least squares (PLS) model was established using the calibration set samples, and the prediction set samples were used to verify and correct the PLS model. The R 2 of the model was 0.85, the RPD was 2.57, and the SEP was 3.66.
实施例4Example 4
1、步骤1-3同实施例1中的步骤1-3相同。1. Steps 1-3 are the same as Steps 1-3 in Example 1.
2、利用标准正态变量变换SNV对所有样品的光谱数据进行优化，以消除校正散射的影响，再通过数学计算法得到一阶导数光谱。2. Optimize the spectral data of all samples by using the standard normal variable transformation SNV to eliminate the influence of corrected scattering, and then obtain the first-order derivative spectrum through mathematical calculation.
3、利用连续投影算法对原始数据进行压缩处理。3. Use the continuous projection algorithm to compress the original data.
4、利用校正集样品建立BP神经网络模型，利用预测集样品来验正并校正BP模型，模型的R2为0.60，RPD为1.59，SEP为5.92。4. Establish the BP neural network model by using the calibration set samples, and use the prediction set samples to verify and correct the BP model. The R 2 of the model is 0.60, the RPD is 1.59, and the SEP is 5.92.
实施例5Example 5
建立好检测模型后，发明人重新到市场上购置并抽取4种小麦，每种6个样品，共计24个样品。利用实施例1-4所构建的检测模型进行硬度测定。近红外扫描及硬度指数法测定等过程均与实施例1中所列方法相同，最终检测结果如表1所示。从表中数据可以看出，实施例1中建立的RBF神经网络模型的预测性能是最好的，误差最小。After establishing the detection model, the inventor went back to the market to purchase and sample 4 kinds of wheat, each with 6 samples, totaling 24 samples. The hardness measurement was carried out using the detection model constructed in Examples 1-4. The processes of near-infrared scanning and hardness index method determination are the same as those listed in Example 1, and the final detection results are shown in Table 1. It can be seen from the data in the table that the prediction performance of the RBF neural network model established in Example 1 is the best, and the error is the smallest.
表1不同检测模型的检测结果Table 1 Detection results of different detection models
表2不同模型判别系数、相对分析误差、预测标准差Table 2 Discrimination coefficient, relative analysis error, and prediction standard deviation of different models
从表2可以看出，实施例1判别系数R2最接近1，相对分析误差值RPD最大，预测标准差SEP最小，由此可见实施例1的模型预测效果很好，可以满足于实际应用。It can be seen from Table 2 that the discriminant coefficient R2 of Example 1 is the closest to 1, the relative analysis error value RPD is the largest, and the prediction standard deviation SEP is the smallest. It can be seen that the model prediction effect of Example 1 is very good, which can meet the practical application.
虽然本发明已以较佳的实施例公开如上，但其并非用以限定本发明，任何熟悉此技术的人，在不脱离本发明的精神和范围内，都可以做各种改动和修饰，因此本发明的保护范围应该以权利要求书所界定的为准。Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person familiar with this technology can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore The scope of protection of the present invention should be defined by the claims.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510477871.7A CN105044022B (en) | 2015-08-06 | 2015-08-06 | A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510477871.7A CN105044022B (en) | 2015-08-06 | 2015-08-06 | A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105044022A CN105044022A (en) | 2015-11-11 |
| CN105044022B true CN105044022B (en) | 2018-09-21 |
Family
ID=54450740
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510477871.7A Expired - Fee Related CN105044022B (en) | 2015-08-06 | 2015-08-06 | A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105044022B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106803009A (en) * | 2015-11-26 | 2017-06-06 | 北京光巨力信息技术有限公司 | A kind of near infrared no-wound blood glucose meter correcting algorithm based on manifold learning and RBF neural |
| CN106485049B (en) * | 2016-09-09 | 2019-03-26 | 黑龙江大学 | A kind of detection method of the NIRS exceptional sample based on Monte Carlo cross validation |
| CN108318445A (en) * | 2018-04-10 | 2018-07-24 | 江苏大学 | A kind of near infrared technology qualitative discrimination wheat whether the detection method of heated denaturalization |
| CN108872143A (en) * | 2018-05-22 | 2018-11-23 | 南京农业大学 | A kind of wheat infection head blight level detection method based near infrared spectrum |
| CN108875913B (en) * | 2018-05-30 | 2021-09-17 | 江苏大学 | Tricholoma matsutake rapid nondestructive testing system and method based on convolutional neural network |
| CN109580525A (en) * | 2018-12-03 | 2019-04-05 | 益海嘉里（兖州）粮油工业有限公司 | A kind of detection method of quick predict wheat baking quality |
| CN109636074A (en) * | 2019-02-01 | 2019-04-16 | 中国农业大学 | The near-infrared method for quick predicting of pig digestible energy and metabolic energy in a kind of sorghum |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6100526A (en) * | 1996-12-30 | 2000-08-08 | Dsquared Development, Inc. | Grain quality monitor |
| US6596930B1 (en) * | 1998-05-22 | 2003-07-22 | The United States Of America As Represented By The Secretary Of Agriculture | Modification of cereal grain hardness via expression of puroindoline protein |
| CN101713731B (en) * | 2009-11-09 | 2012-02-29 | 北京中医药大学 | A method for identifying the coating quality of pharmaceutical preparations |
| CN103149174A (en) * | 2013-03-19 | 2013-06-12 | 周口市农业科学院 | Method for rapidly detecting yellow pigment content in wheat grain |
| CN104020135B (en) * | 2014-06-18 | 2016-09-21 | 中国科学院重庆绿色智能技术研究院 | Calibration model modeling method based near infrared spectrum |
| CN104749126A (en) * | 2015-02-28 | 2015-07-01 | 华北水利水电大学 | Wheat hardness prediction method based on near infrared hyperspectral image analysis |
-
2015
- 2015-08-06 CN CN201510477871.7A patent/CN105044022B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN105044022A (en) | 2015-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105044022B (en) | A kind of method and application based on near-infrared spectrum technique Fast nondestructive evaluation wheat hardness | |
| CN105181643B (en) | A kind of near infrared detection method of rice quality and application | |
| Osborne et al. | Single‐kernel characterization principles and applications | |
| CN108663339A (en) | Corn online test method of going mouldy based on spectrum and image information fusion | |
| CN107515203A (en) | Quantitative analysis of rice single grain amylose content by near-infrared technology | |
| CN110672546A (en) | A modeling method of distiller's grains model based on portable near-infrared spectrometer | |
| CN108760647A (en) | A kind of wheat content of molds line detecting method based on Vis/NIR technology | |
| CN106226267B (en) | A near-infrared method for measuring the color value of dried pepper | |
| CN107655851A (en) | A kind of method based on near-infrared spectrum technique quick detection lysine content | |
| CN109540837A (en) | The method that near-infrared quickly detects Boehmeria nivea leaves wood fibre cellulose content | |
| CN105044024A (en) | Method for nondestructive testing of grape berries based on near infrared spectrum technology | |
| CN109520965A (en) | A method of lysine content is detected based near infrared spectrum characteristic extractive technique | |
| CN111665217A (en) | Near infrared spectrum detection method for sucrose content of soybean seeds | |
| CN110231306A (en) | A kind of method of lossless, the quick odd sub- seed protein content of measurement | |
| Hao et al. | Online accurate detection of soluble solids content in navel orange assisted by automatic orientation correction device | |
| CN110231302A (en) | A kind of method of the odd sub- seed crude fat content of quick measurement | |
| Symons et al. | Measurement of hard vitreous kernels in durum wheat by machine vision | |
| WO2020248961A1 (en) | Method for selecting spectral wavenumber without reference value | |
| CN107991265A (en) | A kind of wheat flour Rubus biflorus Buch quick determination method based on information fusion | |
| CN108717001B (en) | A kind of detection method of wheat seed hardness | |
| CN108169168A (en) | Test and analyze rice grain protein content mathematical model and construction method and application | |
| CN109932336A (en) | A kind of method for quick identification of wholemeal | |
| Fei et al. | The rapid determination of the fatty acid content of rice by combining hyperspectral imaging and integrated learning models | |
| Sinelli et al. | Prediction of semolina technological quality by FT-NIR spectroscopy | |
| Chen et al. | Prediction of milled rice grades using Fourier transform near-infrared spectroscopy and artificial neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Ye Dandan Inventor after: Sun Laijun Inventor after: Liu Yangyang Inventor after: Hui Guangyan Inventor after: Wang Jianan Inventor before: Sun Laijun Inventor before: Hui Guangyan Inventor before: Wang Jianan |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180921 Termination date: 20210806 |