WO2023015609A1 - 一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法 - Google Patents

一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法 Download PDF

Info

Publication number
WO2023015609A1
WO2023015609A1 PCT/CN2021/114199 CN2021114199W WO2023015609A1 WO 2023015609 A1 WO2023015609 A1 WO 2023015609A1 CN 2021114199 W CN2021114199 W CN 2021114199W WO 2023015609 A1 WO2023015609 A1 WO 2023015609A1
Authority
WO
WIPO (PCT)
Prior art keywords
meat
independent component
aoq
denoted
int
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.)
Ceased
Application number
PCT/CN2021/114199
Other languages
English (en)
French (fr)
Inventor
石吉勇
王月影
邹小波
黄晓玮
李志华
张新爱
张迪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to US17/606,053 priority Critical patent/US11940435B2/en
Priority to GB2115554.4A priority patent/GB2597160B/en
Publication of WO2023015609A1 publication Critical patent/WO2023015609A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Definitions

  • the invention relates to the technical field of identification of adulteration of meat products, in particular to an identification method of original cut meat and highly imitated meat based on the linear gradient feature of components.
  • Meat is an important source of high-quality dietary protein and plays an important role in consumers' daily diet.
  • natural and healthy food has gradually become the mainstream concept of consumers' diet, and raw meat from animal carcasses has become increasingly popular.
  • the appearance of raw cut meat has a specific texture and color, and its internal components and structure are formed through natural methods. It has the characteristics of high consumer recognition and high price.
  • the contradiction between supply and demand has become increasingly prominent, and there are more and more high-quality imitation meats with similar appearance and the same components as original cut meat.
  • high-quality imitation meat is mainly artificially produced, using minced meat and low-value meat as raw materials, imitating the texture and color of original cut meat, with the help of carrageenan, glutamyl transaminase or freezing Make it into chunks.
  • High imitation meat is not only low in cost, but also has similar or even the same appearance and components as the original cut meat. Unscrupulous merchants often sell it as natural original cut meat, which affects consumer rights and market order. Therefore, it is necessary to establish a method for identifying natural raw meat and artificial high-imitation meat.
  • the identification of the authenticity of meat is mainly based on the appearance characteristics and internal components.
  • Common methods include sensory methods, chromatography, spectroscopy and molecular biology methods.
  • the sensory method mainly identifies adulterated meat through the color, texture, smell, flavor and other characteristics of the meat sample; however, the artificial high-imitation meat can be similar to the original cut meat after being dyed with food dyes, modified with food additives, and adjusted in the order of placement.
  • the appearance characteristics of the sensory method lead to poor identification effect.
  • Chromatography and spectroscopy (such as patents CN106483221A, CN104132896A) are all based on the perspective of meat components, and qualitative identification is realized by comparison with standard substances.
  • the original cut meat and its imitation meat have similar or even the same appearance, components and biological characteristics, the formation process of the two is obviously different.
  • the original cut meat is formed from the natural growth and development of animal carcasses, and its local components have undergone a natural gradual process from scratch, from less to more.
  • the high imitation meat of the original cut meat is artificially spliced and recombined from the same kind of minced meat or low-value meat.
  • the rate of component change corresponding to natural growth. Therefore, the spatial variation rate of meat sample components can be used as the basis for identifying the original cut meat and its high-imitation meat.
  • the present invention utilizes the characteristics that visible/near-infrared hyperspectral signals are sensitive to meat component content and its spatial distribution, and proposes a method for identifying original cut meat and its highly imitated meat based on the linear array gradient characteristics of components .
  • the present invention provides a method for identifying original cut meat and its high-imitation meat based on the linear gradient characteristics of the components, so as to realize the rapid and accurate identification of the original cut meat and its high-imitation meat .
  • a method for identifying original cut meat and its highly imitated meat based on the linear gradient characteristics of the components which is characterized in that it includes two steps: construction of an identification model based on the linear gradient characteristics of the cross-sectional components, and identification of the meat samples to be tested:
  • Step 1 the construction of the identification model based on the linear gradient feature of the cross-section component includes the following process:
  • Process 1 longitudinally or cross-cutting the meat product to obtain a longitudinal or transverse section of the meat product;
  • the meat product includes original cut meat and its high imitation meat; including b original cut meat samples and c high imitation meat samples , according to the ratio of d:1, it is randomly divided into a correction set and a prediction set; the b, c, and d are all positive integers;
  • Process 2 Use the visible/near-infrared hyperspectral imaging system to image the cross-section of the processed meat in process 1 to obtain a three-dimensional hyperspectral image of M ⁇ N ⁇ W, where M and N represent the number of rows of image pixels at a single wavelength and the number of columns, W represents the number of wavelengths of the hyperspectral image;
  • Process 3 Perform independent component analysis on the 3D hyperspectral image corresponding to the cross-section of meat obtained in process 2, and obtain the first n independent component images corresponding to the hyperspectral image of the meat cross-section, which are respectively denoted as IC-1, IC-2, ... , IC-(n-1), IC-n, wherein n is a positive integer less than or equal to W;
  • Process 4 take the center position of the independent component image IC-i (i ⁇ [1,n]) obtained in process 3 as the center of the circle, denoted as O-i, and the coordinates are (Int(M/2), Int(N/2)) , define a circular area with a radius of r pixels as the feature extraction area, which is denoted as R-i; in the feature extraction area R-i, define O-i as the pole and the horizontal ray as the polar axis, denoted as Ox; take the pixel as the unit length, and The counterclockwise direction is a positive polar coordinate system.
  • Process 5 starting from the pole O-i in the independent component image IC-i in the process 4, connect the end point Q-i-j of the feature extraction line segment with a line with a width of 1 pixel, and obtain q feature extraction line segments in the independent component image IC-i , the feature extraction line segment is denoted as OQ-i-j;
  • Process 6 convert the image covered by q feature extraction line segments OQ-i-j in the independent component image IC-i in process 5 into a row vector, denoted as AOQ-i-j; see the pixel signal strength in the row vector AOQ-i-j Cheng is a function that changes with the position of the pixel point.
  • the first-order derivative vector corresponding to AOQ-i-j is denoted as AOQ'-i-j and the second-order derivative vector is denoted as AOQ”-i-j;
  • Process 7 Statistically calculate the linear array gradient features corresponding to the independent component image IC-i, denoted as Bi, take the value a as the judgment threshold, Bi is the absolute value of the q derivative vectors AOQ'-ij corresponding to the independent component image IC-i is greater than or equal to Threshold a and the number of elements whose AOQ”-ij is equal to 0, that is, satisfying the two conditions of
  • Step 8 Establish a 1*n matrix C to save the linear gradient feature B-i (i ⁇ [1,n]) corresponding to n independent component images, where the linear gradient feature B-i corresponding to the i-th independent component image is saved In C(1, i), it is used for the construction of the identification model; wherein the linear gradient feature data set corresponding to the correction set in the n independent component images is denoted as C_cal and the linear gradient feature data set corresponding to the prediction set is denoted as C_pre;
  • identification model construction use reference value 1 to represent the original cut meat, and reference value 0 to represent high imitation meat, construct the reference value data set Y_cal corresponding to the calibration set and the reference value data set Y_pre corresponding to the prediction set; combine process eight
  • the characteristic data set C_cal of the correction set and the characteristic data set C_pre of the prediction set are used to establish the identification model of the original cut meat and its high imitation meat;
  • the correct recognition rate R of the prediction set is used as an index to measure the performance of the identification model, R ⁇ 60%
  • Step 2 the identification of the meat samples to be tested, includes the following process:
  • the calibration set in process one contains Int[b*d/(d+1)] original cut meat samples and Int[c*d/(d+1)] high imitation meat samples; prediction set It contains a total of b-Int[b*d/(d+1)] original cut meat samples and c-Int[c*d/(d+1)] high imitation meat samples, where Int[b*d/( d+1)] and Int[c*d/(d+1)] respectively represent the rounding of b*d/(d+1) and c*d/(d+1).
  • both b and c mentioned in process one are positive integers greater than 30; the value of d is 1-5.
  • the step size ⁇ n in process three is 1.
  • max * ⁇ , where ⁇ [g%, h%], determine The maximum value of threshold a max
  • max *h%, the minimum value of decision threshold a a min
  • max *g%, step size ⁇ a
  • the corresponding identification model is denoted as Y op , where
  • the number of independent component images and the decision threshold corresponding to the model are denoted as the optimal number of independent component images n op and the optimal decision threshold a op .
  • the KNN algorithm is used to establish the identification model of the original cut meat and its high imitation meat as described in the ninth process.
  • the present invention utilizes the sensitivity of visible/near-infrared spectral signals to changes in meat components, and the advantage that spectral scanning technology can continuously acquire optical signals of samples in the spatial dimension, characterizes the variation law of characteristic components in meat products from the spatial dimension, and Further, based on the spectral difference of the components in the region of interest in the hyperspectral image, combined with the derivation to characterize the change rate of the characteristic components, the identification of the original cut meat and its high imitation meat is realized.
  • a method for identifying original cut meat and its highly imitated meat based on the linear gradient characteristics of the components which is characterized in that it includes two steps: construction of the identification model based on the gradient characteristics of the cross-section components and identification of the meat samples to be tested:
  • Step 1 the construction of the identification model based on the gradient characteristics of the cross-section components includes the following processes:
  • the meat sample includes original cut meat and its high imitation meat; specifically select 80 original cut meat samples and 80
  • the high imitation meat samples were randomly divided into calibration set and prediction set according to the ratio of 3:1.
  • the calibration set contained 60 original cut meat samples and 60 high imitation meat samples
  • the prediction set contained 20 original cut meat samples and 20 high imitation meat samples;
  • Process 2 Use the visible/near-infrared hyperspectral imaging system to image the cross-section of the meat to obtain a 1850 ⁇ 1700 ⁇ 618 three-dimensional hyperspectral image, where 1850 and 1700 represent the number of rows and columns of image pixels at a single wavelength, and 618 Indicates the number of wavelengths of the hyperspectral image;
  • Process 3 Perform independent component analysis on the hyperspectral image corresponding to the cross-section of meat obtained in process 2, and obtain the first n independent component images corresponding to the hyperspectral image of the meat cross-section; denoted as IC-1, IC-2, ..., IC -(n-1), IC-n;
  • Process 4 take the hyperspectral independent component map IC-1 of raw cut meat obtained in process 3 as an example, take the center position of IC-1 as the center of the circle, denote it as O-i, the coordinates are (925, 850), and the defined radius is 850
  • the circular area of the pixel is used as the feature extraction area, which is denoted as R-1; in the feature extraction area R-1, it is defined as O-1 as the pole and the horizontal ray as the polar axis, denoted as Ox, the unit length is pixel, and the inverse
  • the clockwise direction is a positive polar coordinate system, using this polar coordinate system to determine the end points of 6 feature extraction line segments Q-1-j(850, 360*(j-1)/6) on the independent component image IC-1, where j ⁇ [1,6];
  • Process 5 Starting from the pole O-1 in the independent component image IC-1, use a line connection feature with a width of 1 pixel to extract the end point Q-1-j of the line segment, and obtain 6 poles in the independent component image IC-1 Feature extraction line segment, the feature extraction line segment is recorded as OQ-1-j, j ⁇ [1, 6], the schematic diagram of the feature extraction line segment is shown in Figure 3;
  • Process six convert the image covered by the first feature extraction line segment OQ-1-1 in the independent component image IC-1 into a row vector, which is recorded as AOQ-1-1; the row vector AOQ
  • the pixel signal strength in -1-1 is regarded as a function that changes with the position of the pixel point.
  • the first-order derivative vector corresponding to AOQ-1-1 is recorded as AOQ'-1-1 and the second-order derivative.
  • the vector is denoted as AOQ"-1-1.
  • the graph corresponding to the first derivative vector AOQ'-1-1 of the meat sample OQ-1-1 when j 1;
  • Process seven counting the linear array gradient feature B-1 corresponding to the independent component image IC-1, using the value a as the judgment threshold, and B-1 is the six derivative vectors AOQ'-1-j( Including AOQ'-1-1, AOQ'-1-2, AOQ'-1-3, AOQ'-1-4, AOQ'-1-5, AOQ'-1-6) the absolute value is greater than or equal to the judgment Threshold a and AOQ"-1-1 is equal to the number of elements of 0.
  • the decision threshold a
  • Step 8 Establish a 1*n matrix C to store the linear gradient features B-i(i ⁇ [1,n]) corresponding to n independent component images, where the linear gradient feature B-i(i ⁇ [1,n]) corresponding to the first independent component image corresponds to 1 is stored in C(1, 1) for the construction of the identification model; wherein the linear gradient feature data set corresponding to the correction set in the n independent component images is denoted as C_cal and the linear gradient feature data set corresponding to the prediction set is denoted as C_pre;
  • identification model construction use reference value 1 to represent the original cut meat, and reference value 0 to represent high imitation meat, construct the reference value data set Y_cal corresponding to the calibration set and the reference value data set Y_pre corresponding to the prediction set; combine process eight
  • Y op F (4, 0.00183) (X)
  • Step 2 the identification of the meat sample to be tested includes the following processes:

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)

Abstract

一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法。利用可见/近红外光谱信号对肉类组分变化的敏感性，以及光谱扫描技术可以在空间维度连续获取样品光学信号的优势，从空间维度表征肉品内部特征组分的变化规律，并进一步依据高光谱图像中感兴趣区域的组分光谱差异，结合求导表征特征组分变化率，构建了鉴别模型，能够根据自然渐变的原切肉与人工重组的高仿肉间特征组分渐变特征的差异进行有效区分，实现对原切肉及其高仿肉的鉴别。

Description

一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法 技术领域
本发明涉及肉品掺假鉴别技术领域，具体涉及一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法。
背景技术
肉类是优质膳食蛋白的重要来源，在消费者的日常饮食中具有重要地位。近年来，天然、健康逐渐成为消费者饮食的主流观念，来源于动物胴体的原切肉日益受到追捧。原切肉的外观具有特定的纹理和色泽，其内部组分及结构均通过自然的方式形成，具有消费者认可度高、售价高等特点。然而，随着原切肉消费需求的不断攀升导致供需矛盾日益突出，与原切肉外观类似、组分相同的高仿肉日益增多。同原切肉的自然形成方式不同，高仿肉主要通过人工的方式，以碎肉、低值肉为原料，仿照原切肉的纹理和色泽，借助卡拉胶、谷氨酰转氨酶或冻结的方式使之成块。高仿肉不仅成本低，而且具有与原切肉相似、甚至相同的外观和组分，不良商家往往将其冒充天然原切肉销售，对消费者权益和市场秩序造成影响。因此，有必要建立天然原切肉及人工高仿肉的鉴别方法。
肉品真伪的鉴别主要从外观特征和内部组分两个角度出发，常见的方法有感官法、色谱法、光谱法和分子生物学方法。感官法主要通过肉样的颜色、纹理、气味、风味等特征鉴别掺假肉品；然而，人工高仿肉通过食品染料染色、食品添加剂改性、摆放顺序调节后可拥有与原切肉相似的外观特征，导致感官法的鉴别效果较差。色谱法、光谱法(如专利CN106483221A，CN104132896A)都是基于肉品组分的角度，通过与标准物质的对比实现定性鉴别。分子生物学方法(CN103361422A)虽然检测结果可靠，但存在成本高、耗时长、操作复杂等缺陷。这几类鉴别方法仅能实现不同种类肉的掺假鉴别，当使用同类肉进行重组时，无法实现对应高仿肉的鉴别。
原切肉及其高仿肉虽然具有相似、甚至相同的外观、组分及生物学特征，但两者的形成过程具有明显差异。原切肉形成于自然生长发育而成的动物胴体，其局部组分经历了从无到有、从少到多的自然渐变过程。原切肉的高仿肉是由同种类碎肉、或低值肉经人工拼接重组而成，碎肉与碎肉间隙处的组分没有经历自然渐变过程，其对应特征组分变化率往往大于自然生长对应的组分变化率。因此，肉样组分的空间变化率可成为鉴别原切肉及其高仿肉的依据。鉴于此，本发明利用可见/近红外高光谱信号对肉品组分含量及其空间分布敏感的特性，提出了一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法。
发明内容
针对现有技术的不足，本发明的提供了一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，以实现对原切肉及其高仿肉的快速、准确鉴别。
一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于包含基于截面组分线阵渐变特征的鉴别模型构建、待测肉品样品鉴别两个步骤：
步骤一，所述基于截面组分线阵渐变特征的鉴别模型构建包含以下过程：
过程一，对肉品进行纵切或横切，得到肉品的纵向或横向截面；所述肉品包括原切肉及其高仿肉；包括b个原切肉样品，c个高仿肉样品，按照d：1的比例随机分为校正集和预测集；所述b、c、d均为正整数；
过程二，利用可见/近红外高光谱成像系统对过程一处理后肉品的截面进行成像，得到M×N×W的三维高光谱图像，其中M和N分别表示单个波长下图像像素的行数和列数，W表示高光谱图像的波长数；
过程三，对过程二获得肉品截面对应的三维高光谱图像进行独立分量分析，得到肉品截面高光谱图像对应的前n个独立分量图像，分别记为IC-1，IC-2，……，IC-(n-1)，IC-n，其中n为小于等于W的正整数；
过程四，以过程三得到的独立分量图像IC-i(i∈[1，n])的中心位置为圆心，记为O-i，坐标为(Int(M/2)，Int(N/2))，定义半径为r个像素的圆形区域作为特征提取区域，记为R-i；在特征提取区域R-i内定义以O-i为极点、以水平射线为极轴，记为Ox；以像素为单位长度、以逆时针方向为正的极坐标系，利用该极坐标系在独立分量图像IC-i上确定q个特征提取线段终点，记为Q-i-j，坐标为(r,360*(j-1)/q)，其中j∈[1，q]；所述Int(M/2)表示对M/2取整，Int(N/2)表示对N/2取整；其中q为大于1的整数；其中r为不大于min(Int(M/2)，Int(N/2))的正整数，M和N分别表示单个波长下图像像素的行数和列数；
过程五，以过程四中独立分量图像IC-i中的极点O-i为起点，分别用宽度为1个像素的线条连接特征提取线段终点Q-i-j，得到独立分量图像IC-i中的q条特征提取线段，特征提取线段记为OQ-i-j；
过程六，分别将过程五中独立分量图像IC-i中的q条特征提取线段OQ-i-j覆盖的图像转化为行向量，记为AOQ-i-j；将行向量AOQ-i-j中的像素信号强度看成是随像素点位置变化的函数，对行向量求导后得到AOQ-i-j对应的一阶导数向量记为AOQ'-i-j和二阶导数向量记为AOQ”-i-j；
过程七，统计独立分量图像IC-i对应的线阵渐变特征，记为B-i，以数值a作为判定阈值，B-i为独立分量图像IC-i对应q条导数向量AOQ'-i-j的绝对值大于等于阈值a且AOQ”-i-j等 于0的元素个数，即同时满足|AOQ'-i-j|≥a和AOQ”-i-j＝0两个条件；所述判定阈值a＝|AOQ'-i-j| max*η，其中0＜η＜1；|AOQ'-i-j| max为独立分量图像IC-i对应q条特征提取线段的一阶导数向量AOQ'-i-j中元素绝对值的最大值；
过程八，建立1*n的矩阵C用于保存n个独立分量图像对应的线阵渐变特征B-i(i∈[1，n])，其中第i个独立分量图像对应的线阵渐变特征B-i保存于C(1，i)中用于鉴别模型的构建；其中n个独立分量图像中校正集对应的线阵渐变特征数据集记为C_cal和预测集对应的线阵渐变特征数据集记为C_pre；
过程九，鉴别模型构建：以参考值1代表原切肉、以参考值0代表高仿肉，构建校正集对应的参考值数据集Y_cal和预测集对应的参考值数据集Y_pre；结合过程八中的校正集的特征数据集C_cal和预测集的特征数据集C_pre，建立原切肉及其高仿肉的鉴别模型；以预测集的正确识别率R作为衡量鉴别模型性能的指标，R≥60％时，对应的鉴别模型记为Y，其中Y＝F (n，a)(X)，X为每个样品全部独立分量图像对应的线阵渐变特征数据集，n为小于等于W的正整数，a作为判定阈值。
步骤二，待测肉品样品的鉴别，包含以下过程：
S1，采集待测肉品样品对应的高光谱图像，根据步骤一过程一到过程八描述的步骤提取待测肉品样品对应的线阵渐变特征数据集，记为C_uk；
S2，将C_uk代入步骤一的过程九中鉴别模型Y＝F (n，a)(X)内，计算待测样品对应的参考值；当模型预测的参考值为1时，待测样品为原切肉；当模型预测的参考值为0时，待测样品为高仿肉，实现待测肉品样品的鉴别。
优选的，过程一中所述校正集共包含Int[b*d/(d+1)]个原切肉样品和Int[c*d/(d+1)]个高仿肉样品；预测集共包含b-Int[b*d/(d+1)]个原切肉样品和c-Int[c*d/(d+1)]个高仿肉样品，其中Int[b*d/(d+1)]、Int[c*d/(d+1)]分别表示对b*d/(d+1)和c*d/(d+1)取整。
优选的，过程一中所述b，c均为大于30的正整数；d的取值为1～5。
优选的，过程三中所述独立分量图像个数n的范围优化及步长Δn：设定独立分量图像个数n的最大值为n max＝Int(W*p)，其中p∈[0.5％，20％]，Int(W*p)表示对W*p取整；独立分量图像个数n的最小值n min＝1，n的优化范围为n min～n max；步长Δn记为f，其中f为正整数。
优选的，过程三中所述步长Δn为1。
优选的，过程七中所述判定阈值a的取值范围优化及其步长Δa：设定判定阈值a＝|AOQ'-i-j| max*η，其中η∈[g％，h％]，判定阈值a的最大取值a max＝|AOQ'-i-j| max*h％，判定阈值a的最小取值a min＝|AOQ'-i-j| max*g％，步长Δa＝|AOQ'-i-j| max*k％，其中g，h，k均为正整数， 且g＜h。
优选的，过程九中当正确识别率R取最大值时，对应的鉴别模型记为Y op，其中
Figure PCTCN2021114199-appb-000001
该模型对应的独立分量图像个数、判定阈值记为最优独立分量图像个数n op和最优判定阈值a op。
优选的，过程九中所述建立原切肉及其高仿肉的鉴别模型运用的是KNN算法。
上述步骤中使用的符号*表示乘以的含义；步骤中所出现的字母i，均满足i∈[1，n]，其中n为小于等于W的正整数；j∈[1，q]，其中q为大于1的整数。
本发明的有益效果：
本发明利用可见/近红外光谱信号对肉类组分变化的敏感性，以及光谱扫描技术可以在空间维度连续获取样品光学信号的优势，从空间维度表征肉品内部特征组分的变化规律，并进一步依据高光谱图像中感兴趣区域的组分光谱差异，结合求导表征特征组分变化率，实现对原切肉及其高仿肉的鉴别。
肉类组分与光谱信号之间具有线性关系，而求导是表征变化趋势的重要方式，导数反映了原函数的变化率。因此，即使高仿肉与原切肉的颜色相近、组分相似，本发明提供的技术方案仍能够根据自然渐变的原切肉与人工重组的高仿肉间特征组分渐变特征的差异进行有效区分。
附图说明
图1为实施例1中当n＝4时，肉品样品的高光谱独立分量图分别为IC-1、IC-2、IC-3、IC-4。
图2为实施例1中当i＝1，q＝6时，从肉品样品独立分量图IC1中特征提取线段的示意图。
图3为实施例1中当i＝1，j＝1时，肉品样品OQ-1-1的行向量AOQ-1-1对应的曲线图。
图4为实施例1中当i＝1，j＝1时，肉品样品OQ-1-1的一阶导数向量AOQ'-1-1对应的曲线图。
具体实施方式
下面结合一些具体实施例为本发明进一步详细说明，但本发明的保护范围不仅限于这些实施例。
实施例1：
一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于包含基于截面组分渐变特征的鉴别模型构建和待测肉品样品鉴别两个步骤：
步骤一，所述基于截面组分渐变特征的鉴别模型构建包含以下过程：
过程一，对肉品样品进行纵切或横切，得到肉品样品的纵向或横向截面；所述肉品样品 包括原切肉及其高仿肉；具体选择80个原切肉样品和80个高仿肉样品，按照3：1的比例随机分为校正集和预测集，其中校正集共包含60个原切肉样品和60个高仿肉样品，预测集共包含20个原切肉样品和20个高仿肉样品；
过程二，利用可见/近红外高光谱成像系统对肉品截面进行成像，得到1850×1700×618的三维高光谱图像，其中1850和1700分别表示单个波长下图像像素的行数和列数，618表示高光谱图像的波长数；
过程三，对过程二获得肉品截面对应的高光谱图像进行独立分量分析，得到肉品截面高光谱图像对应的前n个独立分量图像；记为IC-1，IC-2，……，IC-(n-1)，IC-n；
所述独立分量图像个数n的范围优化及步长Δn：设定独立分量图像个数n的最大值为n max＝Int(618*p)，其中p∈[0.5％，20％]，Int(618*p)表示对618*p取整，取P为0.7％，n max＝6；独立分量图像个数n的最小值n min＝1，优化范围为1～6；取f为1，步长Δn＝1，其中独立分量图像n的可能取值为[1、2、3、4、5、6]；n取4时，如图1所示；
过程四，以过程三得到的原切肉的高光谱独立分量图IC-1为例，取IC-1的中心位置为圆心，记为O-i，坐标为(925，850)，定义半径为850个像素的圆形区域作为特征提取区域，记为R-1；在特征提取区域R-1定义以O-1为极点、以水平射线为极轴，记为Ox、以像素为单位长度、以逆时针方向为正的极坐标系，利用该极坐标系在独立分量图像IC-1上确定6个特征提取线段终点Q-1-j(850，360*(j-1)/6)，其中j∈[1，6]；
过程五，以独立分量图像IC-1中的极点O-1为起点，分别用宽度为1个像素的线条连接特征提取线段终点Q-1-j，得到独立分量图像IC-1中的6条特征提取线段，特征提取线段记为OQ-1-j，j∈[1，6]，特征提取线段的示意图如图3所示；
过程六，以j＝1为例，将独立分量图像IC-1中的第1条特征提取线段OQ-1-1覆盖的图像转化为行向量，记为AOQ-1-1；将行向量AOQ-1-1中的像素信号强度看成是随像素点位置变化的函数，对行向量求导后得到AOQ-1-1对应的一阶导数向量记为AOQ'-1-1和二阶导数向量记为AOQ”-1-1。其中图3为当i＝1，j＝1时肉品样品OQ-1-1的行向量AOQ-1-1对应的曲线图；图4为当i＝1，j＝1时肉品样品OQ-1-1的一阶导数向量AOQ'-1-1对应的曲线图；
过程七，统计独立分量图像IC-1对应的线阵渐变特征B-1，以数值a作为判定阈值，B-1为独立分量图像IC-1对应的6条导数向量AOQ'-1-j(其中包括AOQ'-1-1、AOQ'-1-2、AOQ'-1-3、AOQ'-1-4、AOQ'-1-5、AOQ'-1-6)的绝对值大于等于判定阈值a且AOQ”-1-1等于0的元素个数。所述判定阈值a＝|AOQ'-i-j| max*η，其0＜η＜1；
所述判定阈值a的取值范围优化及步长Δa：设定判定阈值a＝|-0.0122|*η，其中η∈[5％，40％]，|-0.0122|为独立分量图像IC-i对应全部特征提取线段的一阶导数向量AOQ'-i-j中元素 绝对值的最大值；取k＝5，判定阈值a的最大取值a max＝0.00488，判定阈值a的最小取值a min＝0.00061，步长Δa＝0.00061，其中判定阈值a的可能取值为[0.00061、0.00122、0.00183、0.00244、0.00305、0.00366、0.00427、0.00488]；
过程八，建立1*n的矩阵C用于保存n个独立分量图像对应的线阵渐变特征B-i(i∈[1，n])，其中第1个独立分量图像对应的线阵渐变特征B-1保存于C(1，1)中用于鉴别模型的构建；其中n个独立分量图像中校正集对应的线阵渐变特征数据集记为C_cal和预测集对应的线阵渐变特征数据集记为C_pre；
过程九，鉴别模型构建：以参考值1代表原切肉、以参考值0代表高仿肉，构建校正集对应的参考值数据集Y_cal和预测集对应的参考值数据集Y_pre；结合过程八中的校正集的特征数据集C_cal和预测集的特征数据集C_pre，基于KNN算法建立原切肉及其高仿肉的鉴别模型；对应的鉴别模型记为Y，其中Y＝F (n，a)(X)，X为每个样品全部独立分量图像对应的线阵渐变特征数据集。
以预测集的正确识别率R作为衡量鉴别模型性能的指标，当述R≥60％时符合检测要求；当正确识别率R＝98.75％时为最大值，对应的鉴别模型为最优模型，记为Y op，其中Y op＝F (4，0.00183)(X)，该模型对应的独立分量图像个数、判定阈值记为最优独立分量图像个数n op＝4和最优判定阈值a op＝0.00183。
步骤二，所述待测肉品样品鉴别包含以下过程：
S1，采集待测肉品样品对应的高光谱图像，根据步骤一过程一到过程八描述的全部过程提取得到待测肉品样品对应的线阵渐变特征数据集，记为C_uk；
S2，将C_uk代入步骤一的过程九中鉴别模型Y op＝F (4，0.00183)(X)内，计算待测样品对应的参考值；当模型预测的参考值为1时，待测样品为原切肉；当模型预测的参考值为0时，待测样品为高仿肉，实现待测肉品样品的鉴别。
说明：以上实施例仅用以说明本发明而并非限制本发明所描述的技术方案；因此，尽管本说明书参照上述的各个实施例对本发明已进行了详细的说明，但是本领域的普通技术人员应当理解，仍然可以对本发明进行修改或等同替换；而一切不脱离本发明的精神和范围的技术方案及其改进，其均应涵盖在本发明的权利要求范围内。

Claims (9)

  1. 一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，具体步骤如下：
    步骤一，所述基于截面组分线阵渐变特征的鉴别模型构建包含以下过程：
    过程一，对肉品进行纵切或横切，得到肉品的纵向或横向截面；所述肉品包括原切肉及其高仿肉；包括b个原切肉样品，c个高仿肉样品，按照d：1的比例随机分为校正集和预测集；所述b、c、d均为正整数；
    过程二，利用可见/近红外高光谱成像系统对过程一处理后肉品的截面进行成像，得到M×N×W的三维高光谱图像，其中M和N分别表示单个波长下图像像素的行数和列数，W表示高光谱图像的波长数；
    过程三，对过程二获得肉品截面对应的三维高光谱图像进行独立分量分析，得到肉品截面高光谱图像对应的前n个独立分量图像，分别记为IC-1，IC-2，……，IC-(n-1)，IC-n，其中n为小于等于W的正整数；
    过程四，以过程三得到的独立分量图像IC-i的中心位置为圆心，记为O-i，坐标为(Int(M/2)，Int(N/2))，其中i∈[1，n]；定义半径为r个像素的圆形区域作为特征提取区域，记为R-i；在特征提取区域R-i内定义以O-i为极点、以水平射线为极轴，记为Ox；以像素为单位长度、以逆时针方向为正的极坐标系，利用该极坐标系在独立分量图像IC-i上确定q个特征提取线段终点，记为Q-i-j，坐标为(r,360*(j-1)/q)，其中j∈[1，q]；所述Int(M/2)表示对M/2取整，Int(N/2)表示对N/2取整；其中q为大于1的整数；其中r为不大于min(Int(M/2)，Int(N/2))的正整数；
    过程五，以过程四中独立分量图像IC-i中的极点O-i为起点，分别用宽度为1个像素的线条连接特征提取线段终点Q-i-j，得到独立分量图像IC-i中的q条特征提取线段，特征提取线段记为OQ-i-j；
    过程六，分别将过程五中独立分量图像IC-i中的q条特征提取线段OQ-i-j覆盖的图像转化为行向量，记为AOQ-i-j；将行向量AOQ-i-j中的像素信号强度看成是随像素点位置变化的函数，对行向量求导后得到AOQ-i-j对应的一阶导数向量记为AOQ'-i-j和二阶导数向量记为AOQ”-i-j；
    过程七，统计独立分量图像IC-i对应的线阵渐变特征，记为B-i，以数值a作为判定阈值，B-i为独立分量图像IC-i对应q条导数向量AOQ'-i-j的绝对值大于等于阈值a且AOQ”-i-j等于0的元素个数，即同时满足|AOQ'-i-j|≥a和AOQ”-i-j＝0两个条件；所述判定阈值a＝|AOQ'-i-j| max*η，其中0＜η＜1；|AOQ'-i-j| max为独立分量图像IC-i对应q条特征提取线段的一阶导数向量AOQ'-i-j中元素绝对值的最大值；
    过程八，建立1*n的矩阵C用于保存n个独立分量图像对应的线阵渐变特征B-i(i∈[1，n])，其中第i个独立分量图像对应的线阵渐变特征B-i保存于C(1，i)中用于鉴别模型的构建；其中n个独立分量图像中校正集对应的线阵渐变特征数据集记为C_cal和预测集对应的线阵渐变特征数据集记为C_pre；
    过程九，鉴别模型构建：以参考值1代表原切肉、以参考值0代表高仿肉，构建校正集对应的参考值数据集Y_cal和预测集对应的参考值数据集Y_pre；结合过程八中的校正集的特征数据集C_cal和预测集的特征数据集C_pre，建立原切肉及其高仿肉的鉴别模型；以预测集的正确识别率R作为衡量鉴别模型性能的指标，R≥60％时，对应的鉴别模型记为Y，其中Y＝F (n,a)(X)，X为每个样品全部独立分量图像对应的线阵渐变特征数据集，n为小于等于W的正整数，a作为判定阈值；
    步骤二，待测肉品样品的鉴别，包含以下过程：
    S1，采集待测肉品样品对应的高光谱图像，根据步骤一过程一到过程八描述的步骤提取待测肉品样品对应的线阵渐变特征数据集，记为C_uk；
    S2，将C_uk代入步骤一的过程九中鉴别模型Y＝F (n,a)(X)内，计算待测样品对应的参考值；当模型预测的参考值为1时，待测样品为原切肉；当模型预测的参考值为0时，待测样品为高仿肉，实现待测肉品样品的鉴别。
  2. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程一中所述校正集共包含Int[b*d/(d+1)]个原切肉样品和Int[c*d/(d+1)]个高仿肉样品；预测集共包含b-Int[b*d/(d+1)]个原切肉样品和c-Int[c*d/(d+1)]个高仿肉样品，其中Int[b*d/(d+1)]、Int[c*d/(d+1)]分别表示对b*d/(d+1)和c*d/(d+1)取整。
  3. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程一中所述b，c均为大于30的正整数；d的取值为1～5。
  4. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程三中所述独立分量图像个数n的范围优化及步长Δn：设定独立分量图像个数n的最大值为n max＝Int(W*p)，其中p∈[0.5％，20％]，Int(W*p)表示对W*p取整；独立分量图像个数n的最小值n min＝1，n的优化范围为n min～n max；步长Δn记为f，其中f为正整数。
  5. 根据权利要求4所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，所述步长Δn为1。
  6. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程七中所述判定阈值a的取值范围优化及其步长Δa：设定判定阈值 a＝|AOQ'-i-j| max*η，其中η∈[g％，h％]，判定阈值a的最大取值a max＝|AOQ'-i-j| max*h％，判定阈值a的最小取值a min＝|AOQ'-i-j| max*g％，步长Δa＝|AOQ'-i-j| max*k％，其中g，h，k均为正整数，且g＜h。
  7. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程九中当正确识别率R取最大值时，对应的鉴别模型记为Y op，其中
    Figure PCTCN2021114199-appb-100001
    Figure PCTCN2021114199-appb-100002
    该模型对应的独立分量图像个数、判定阈值记为最优独立分量图像个数n op和最优判定阈值a op。
  8. 根据权利要求1所述的基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法，其特征在于，过程九中所述建立原切肉及其高仿肉的鉴别模型运用的是KNN算法。
  9. 根据权利要求1～8任一所述的方法用于原切肉和/或高仿肉鉴别的用途。
PCT/CN2021/114199 2021-08-10 2021-08-24 一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法 Ceased WO2023015609A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US17/606,053 US11940435B2 (en) 2021-08-10 2021-08-24 Method for identifying raw meat and high-quality fake meat based on gradual linear array change of component
GB2115554.4A GB2597160B (en) 2021-08-10 2021-08-24 Method for identifying raw meat and high-quality fake meat based on gradual linear array change of component

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110914828.8 2021-08-10
CN202110914828.8A CN113567359B (zh) 2021-08-10 2021-08-10 基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法

Publications (1)

Publication Number Publication Date
WO2023015609A1 true WO2023015609A1 (zh) 2023-02-16

Family

ID=78171098

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/114199 Ceased WO2023015609A1 (zh) 2021-08-10 2021-08-24 一种基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法

Country Status (2)

Country Link
CN (1) CN113567359B (zh)
WO (1) WO2023015609A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11940435B2 (en) * 2021-08-10 2024-03-26 Jiangsu University Method for identifying raw meat and high-quality fake meat based on gradual linear array change of component
CN115980005B (zh) * 2022-12-27 2025-01-21 中国农业科学院农产品加工研究所 一种基于超弱发光技术鉴别菜肴原料肉真伪的方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080199080A1 (en) * 2007-02-21 2008-08-21 Board Of Regents Of University Of Nebraska System and method for analyzing material properties using hyperspectral imaging
CN104132896A (zh) * 2014-07-15 2014-11-05 山东国家农产品现代物流工程技术研究中心 一种快速鉴别掺假肉的方法
CN106841052A (zh) * 2016-12-29 2017-06-13 中国科学院遥感与数字地球研究所 一种快速检测食用肉类新鲜度的方法和系统
CN107271375A (zh) * 2017-07-21 2017-10-20 石河子大学 一种羊肉品质指标的高光谱图像检测方法
TW202100999A (zh) * 2019-06-21 2021-01-01 財團法人食品工業發展研究所 使用近紅外光譜的肉品品質檢測方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103760110A (zh) * 2013-05-23 2014-04-30 山东商业职业技术学院 一种快速鉴别不同动物来源肉的方法
CN108956604B (zh) * 2018-05-29 2021-11-23 江苏大学 一种基于高光谱图像技术识别中华绒螯蟹品质的方法
CN110490895B (zh) * 2019-08-20 2022-02-11 浙江大学 一种基于高光谱图像处理的提高肉源真伪鉴别正确率方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080199080A1 (en) * 2007-02-21 2008-08-21 Board Of Regents Of University Of Nebraska System and method for analyzing material properties using hyperspectral imaging
CN104132896A (zh) * 2014-07-15 2014-11-05 山东国家农产品现代物流工程技术研究中心 一种快速鉴别掺假肉的方法
CN106841052A (zh) * 2016-12-29 2017-06-13 中国科学院遥感与数字地球研究所 一种快速检测食用肉类新鲜度的方法和系统
CN107271375A (zh) * 2017-07-21 2017-10-20 石河子大学 一种羊肉品质指标的高光谱图像检测方法
TW202100999A (zh) * 2019-06-21 2021-01-01 財團法人食品工業發展研究所 使用近紅外光譜的肉品品質檢測方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SUN ZONGBAO, LIANG LI-MING; LI JUN-KUI; ZOU XIAO-BO; LIU XIAO-YU; WANG TIAN-ZHEN: "Identification of Chilled and Frozen-Thawed Salmon Based on Hyperspectral Imaging Technology", SPECTROSCOPY AND SPECTRAL ANALYSIS, BEIJING DAXUE CHUBANSHE, CN, vol. 40, no. 11, 30 November 2020 (2020-11-30), CN , pages 3530 - 3536, XP093034322, ISSN: 1000-0593, DOI: 10.3964/j.issn.1000-0593(2020)11-3530-07 *

Also Published As

Publication number Publication date
CN113567359B (zh) 2022-05-20
CN113567359A (zh) 2021-10-29

Similar Documents

Publication Publication Date Title
Barbin et al. NIR hyperspectral imaging as non-destructive evaluation tool for the recognition of fresh and frozen–thawed porcine longissimus dorsi muscles
CN117115077B (zh) 一种湖泊蓝藻水华检测方法
Yang et al. Improving vegetation segmentation with shadow effects based on double input networks using polarization images
CN113567359B (zh) 基于组分线阵渐变特征的原切肉及其高仿肉的鉴别方法
CN114399807B (zh) 一种基于图像转换的跨光谱人脸识别方法及监控设备
Asmara et al. Chicken meat freshness identification using the histogram color feature
Ghasemi-Varnamkhasti et al. Application of image analysis combined with computational expert approaches for shrimp freshness evaluation
CN105931223A (zh) 基于波段比方法的高光谱反射图像中玉米胚部的分割方法
CN104331877A (zh) 融合色调梯度的彩色图像边缘检测方法
Yang et al. Rapid determination of biogenic amines in cooked beef using hyperspectral imaging with sparse representation algorithm
Mendoza et al. Automated prediction of sensory scores for color and appearance in canned black beans (Phaseolus vulgaris L.) using machine vision
CN108195771A (zh) 一种海洋高光谱遥感影像目标识别方法
CN119251704A (zh) 一种基于机器学习的湖泊遥感影像藻华覆盖度反演方法
CN120877162B (zh) 一种基于神经动力学模型的无人机多尺度作物检测方法
Misimi et al. Quality grading of Atlantic salmon (Salmo salar) by computer vision
US11940435B2 (en) Method for identifying raw meat and high-quality fake meat based on gradual linear array change of component
KR102596206B1 (ko) 영상 분석을 이용한 건해삼 등급 분류 시스템
CN113340914A (zh) 一种基于高光谱遥感的水体富营养化状态测算方法
CN106934760B (zh) 一种面向类别的高光谱数据彩色可视化方法
CN109785342A (zh) 一种畜禽肉质均匀度的分析方法
Ji et al. Infrared and visible image fusion of generative adversarial network based on multi-channel encoding and decoding
Pertiwi et al. Fish freshness detection using UV light based on deep neural network
CN105447891A (zh) 冷冻肉新鲜程度的评估方法及系统
Wu et al. Research on terahertz image analysis of thin-shell seeds based on semantic segmentation
CN112001288B (zh) 一种单幅多光谱遥感图像深灰色飞机快速检测方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21953220

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21953220

Country of ref document: EP

Kind code of ref document: A1