WO2022027785A1 - 一种食品咀嚼效率的评估方法 - Google Patents

一种食品咀嚼效率的评估方法 Download PDF

Info

Publication number
WO2022027785A1
WO2022027785A1 PCT/CN2020/115217 CN2020115217W WO2022027785A1 WO 2022027785 A1 WO2022027785 A1 WO 2022027785A1 CN 2020115217 W CN2020115217 W CN 2020115217W WO 2022027785 A1 WO2022027785 A1 WO 2022027785A1
Authority
WO
WIPO (PCT)
Prior art keywords
sample
chewing
color
area
detected
Prior art date
Application number
PCT/CN2020/115217
Other languages
English (en)
French (fr)
Inventor
俞经虎
周星宇
于浩
钱善华
陈煜瑶
Original Assignee
江南大学
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 江南大学 filed Critical 江南大学
Publication of WO2022027785A1 publication Critical patent/WO2022027785A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/80—Geometric correction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment

Definitions

  • the invention relates to the technical field of food chewing degree detection, in particular to a method for evaluating food chewing efficiency.
  • Chewing efficiency refers to the degree to which a certain amount of food is chewed in a certain period of time.
  • the chewing activity is a complex activity completed under the control of the nervous system, the masticatory muscles contract, the jaws, teeth, etc. produce regular movements, and are completed with the assistance of the lips, cheeks and tongue. Therefore, studying the chewing efficiency of food is of great significance to food processing, denture planting and many other aspects.
  • Common food chewing efficiency includes subjective and objective methods. The subjective measurement method is to understand the chewing efficiency of the teeth through questionnaires and interviews with the testers. This method relies too much on the personal subjective feelings of the testers, the standards are not uniform, and the results are not accurate enough.
  • Objective test methods mainly include brittle food sieving and weighing method, peanut suspension colorimetry, hardened gelatin adsorption colorimetry and mixed test method. .
  • the present invention provides a method for assessing chewing efficiency of food.
  • the sample preparation is easy, the detection process is simple, the detection results are objective, and the accuracy is high.
  • the technical scheme of the present invention is as follows, a method for evaluating the chewing efficiency of food, which comprises the following steps:
  • S4 Press the chewed mixed sample into a sheet shape to obtain a sample piece to be tested; collect images of the front and back sides of the sample piece to be tested based on an image acquisition device, and record as: image of the sample to be tested;
  • S5 Perform de-distortion processing on the to-be-detected sample image to obtain a de-distorted to-be-detected sample image;
  • S9 Input the parameters to be detected into the trained chewing efficiency evaluation model for classification detection, and obtain a chewing efficiency evaluation result corresponding to the sample image to be detected.
  • the detection sample is made of food-grade paraffin as raw material, including the two-color rectangular wax flakes of color A and color B; the color of the two-color rectangular wax flakes is crossed and pressed into a rectangle;
  • the two-color rectangular wax sheet is a cuboid wax sheet of 10mm ⁇ 10mm ⁇ 2mm; the color A and the color B are set to red and green;
  • the measurement parameters include:
  • A represents the area of the sample with the hole removed
  • TA represents the sample area after chewing
  • OA represents the area of the sample before chewing
  • RA represents the color area of color A
  • GA represents the area of color B
  • L represents the length of the sample
  • W represents the sample Width
  • DR is the optical density of A color
  • DG is the optical density of B color
  • M is the average optical density
  • step S7 the discriminant function for the measurement parameter in the chewing efficiency evaluation model is:
  • x 1 is the ratio MIX of the area sum of color A and color B in the sample to the total area
  • x 2 is the length-width relationship LW of the sample after chewing
  • x 3 is the ductility FF of the sample after chewing
  • x 4 is The ratio of the total area of the sample after chewing to the total area of the sample before chewing TA'
  • x 5 The variance of the optical density of A color and the optical density of B color in the mixed area of the sample after chewing
  • x 6 is the ratio of the sample area to the total area after removing the holes TR;
  • step S8 based on the training data set, the obtained coefficient term and constant term of the discriminant equation Y are obtained to obtain the trained chewing efficiency evaluation model;
  • the parameters to be detected of the sample to be detected are substituted into the trained chewing efficiency evaluation model, that is, into the discriminant equation Y, and the calculated value is called the chewing efficiency index MEI; the higher the MEI score, the higher the sample. The poorer the mixing, the lower the chewing ability;
  • the image acquisition device includes: an image acquisition camera and a light source; the camera adopts an industrial camera, the light source adopts an LED light source, and the illumination mode is a light source illumination mode of backlight illumination and direct illumination;
  • step S5 the undistortion processing is performed on the to-be-detected sample image based on the traditional calibration method, and the specific steps include:
  • a2 Based on the image acquisition camera, the standard chessboard is photographed from different angles to obtain a reference picture;
  • a3 extracting reference feature points from the reference picture
  • a4 Calculate and obtain camera intrinsic parameters and distortion parameters corresponding to the image capture camera based on the reference feature point;
  • a5 Extract the R, G, B matrix of the sample image to be detected to obtain three two-dimensional matrices, which are recorded as the two-dimensional matrix of the sample to be detected;
  • a6 According to the camera internal parameters and distortion parameters corresponding to the image acquisition camera, perform a linear interpolation and de-distortion operation on the two-dimensional matrix of the three samples to be detected, to obtain a corrected image data matrix;
  • the prepared sample for detection is placed in a dry and low temperature environment for sealing and preservation; before step S2 is performed, the sample for detection is heated in a water bath at 37° C. for 1 min;
  • the classification of the chewing efficiency evaluation results includes: good, fair, and poor.
  • the method for evaluating the chewing efficiency of food is based on the acquisition of an image of a sample to be detected by an image acquisition device, extraction of parameters to be detected from the image of the sample to be detected, and input to a trained evaluation model of chewing efficiency for automatic identification and classification;
  • the process only requires technicians to compress the chewed mixed samples into thin slices, and the rest of the steps are completed automatically, which greatly reduces the dependence on the ability of technicians, and at the same time does not require the subjective judgment of the inspected personnel, the entire inspection process.
  • the test results are objective and accurate; the larger the amount of training samples of the chewing efficiency evaluation model constructed based on Fisher's discriminant analysis method, the richer the discriminant indicators, and the higher the accuracy of the discriminant equation, especially for the classification of food chewing efficiency
  • the technical scheme of the present invention uses two colors of food
  • the color mixing test material for chewing test is made of high-grade paraffin as the raw material, and the tableting process of the mixed sample after chewing is very simple.
  • the manual operation process is very simple, which reduces the possibility of operating errors and further ensures the accuracy of the evaluation results .
  • Fig. 1 is the scatter diagram of the first discriminant function Y 1 and the second discriminant function Y 2 in the embodiment;
  • Fig. 2 is the BA graph between the peanut median diameter and the MEI value in the contrast experiment.
  • the present invention is a method for evaluating the chewing efficiency of food, which comprises the following steps.
  • the sample for detection is made of food-grade paraffin, including two-color rectangular wax slices of color A and color B; the two-color rectangular wax slices are crossed and pressed into a rectangle; in the embodiment of the present invention, the two-color rectangular wax Wax slice 10mm ⁇ 10mm ⁇ 2mm cuboid wax slice; color A and color B are set to red and green; three green and two yellow wax slices are pressed and the size of the test sample is just right to be placed in the oral cavity; made
  • the detection sample is placed in a dry and low temperature environment and sealed and stored; before step S2, the detection sample is heated with a water bath at 37°C for 1 min; the paraffin material is insoluble in water and does not chemically react with the digestive enzymes in the oral cavity.
  • the production process of samples for testing is relatively simple.
  • the square shape produces multiple stress points during chewing, which makes the sample more easily deformed.
  • the samples pressed with multiple layers of paraffin wax are more likely to mix during the chewing process, and the color mixing effect is better. good.
  • the paraffin material is softer than the natural experimental materials such as peanuts used in the prior art, which can meet the experimental needs of most oral restoration patients and the elderly with severe tooth loss. Compared with yeast sugar, the viscosity is lower, and during the experiment The problem of sticking to the teeth is less likely to occur.
  • S4 Wash and dry the mixed sample after chewing, and then press it into a sheet shape to obtain a sample piece to be tested; based on the image acquisition device, collect images of the front and back sides of the sample piece to be tested, and record as: image of the sample to be tested; image acquisition
  • the device includes: an image acquisition camera and a light source; the camera adopts an industrial camera, the light source adopts an LED light source, and the illumination method is a light source illumination method of backlight illumination and direct illumination; Small size and other characteristics; because the position of the object to be measured is relatively fixed in the later detection process, so in the technical solution of the invention, the fixed focal length mode (that is, the focal length of the lens remains unchanged during the shooting process) is used to improve the shooting quality and ensure that the image acquisition device captures The accuracy of the obtained sample images to be detected, thereby improving the accuracy of the final evaluation results.
  • the fixed focal length mode that is, the focal length of the lens remains unchanged during the shooting process
  • the sample for detection is a cube with a side length of 10 mm.
  • the YX6060 lens of Easyvxin is selected, the focal length of the lens ranges from 6 to 60 mm, and the maximum imaging size of the lens is 2/3".
  • the light source illumination mode of backlight illumination and direct illumination is selected.
  • the top light source is composed of 4 strip LED light sources of HF-FX160160 model, the irradiation method is direct irradiation, and the side light source model Exin-64LED, the irradiation method is backlight irradiation; backlight irradiation ensures a clear outline of the mixed sample after chewing, direct irradiation The irradiation ensures that the surface features of the mixed sample after chewing are obtained, thereby ensuring the accuracy of the image of the sample to be detected collected by the image acquisition device.
  • S5 Perform de-distortion processing on the image of the sample to be detected to obtain a de-distorted image of the sample to be detected.
  • optical axis of the lens is not perpendicular to the plane where the sensor is located, that is, the camera coordinate system is not parallel to the plane where the image coordinate system is located;
  • the lens cannot present a perfect pinhole image due to the manufacturing accuracy, resulting in the projection deviating from the projective straight line, at this time, it needs to be corrected by the distortion parameter;
  • the premise of subsequent analysis and measurement is to do a good job of camera calibration and de-distort the image;
  • the purpose of camera calibration is to determine the conversion matrix of the camera from three-dimensional space to two-dimensional image, so as to realize the conversion between pixels on the picture.
  • the spatial three-dimensional geometric information of the detection target is obtained from the corresponding relationship.
  • the camera calibration can also correct the distortion of the lens and the camera due to the accuracy error caused by manufacturing or installation.
  • Camera calibration is a key step in visual inspection.
  • the image of the sample to be detected is de-distorted, and the specific steps include:
  • a2 Based on the image acquisition camera, the standard chessboard is photographed from different angles to obtain reference pictures;
  • a4 Calculate the camera internal parameters and distortion parameters corresponding to the image acquisition camera based on the reference feature point;
  • a5 Extract the R, G, B matrices of the sample image to be detected to obtain three two-dimensional matrices, which are recorded as the two-dimensional matrix of the sample to be detected;
  • a6 According to the camera internal parameters and distortion parameters corresponding to the image acquisition camera, perform a linear interpolation and de-distortion operation on the two-dimensional matrix of the three samples to be detected, and obtain a corrected image data matrix;
  • the traditional calibration method is used to de-distort the image of the sample to be detected, and the accuracy of the processing result is high.
  • the collection object targeted by the technical solution of the present invention is very easy to obtain, and it is especially suitable for using the traditional calibration method. The more, the higher the accuracy of the calibration method.
  • the standard chessboard calibration template is first made as shown in the figure.
  • the chessboard is divided into 7 rows and 9 columns.
  • the size of a single chessboard is 50 ⁇ 50mm.
  • After printing, it is attached to a flat plate with high flatness, and the camera is used from different angles. Take 20 photos.
  • the standard deviation of the corner point calibration error of 0.71735pixel can meet the requirements of use.
  • the image of the sample to be detected is de-distorted.
  • Use Matlab software to read in the RGB pixel matrix of the sample image to be detected, extract the R, G, B matrices of the sample image to be detected respectively to obtain three two-dimensional matrices, use the im2double function to convert the data of each matrix, and use the im2double function to perform data conversion.
  • Internal parameters and distortion parameters perform linear interpolation and de-distortion operation on the matrix, and finally combine and output the three-dimensional matrix after image correction to obtain a color de-distorted sample image to be detected.
  • Image pro-Plus 6.0 software is used to extract relevant measurement parameters in the picture. Specific steps are as follows:
  • the measurement command under the Count menu can select the measurement parameters for the target area, including area, inclination, optical density, length and width.
  • View command can view and export measurement results;
  • the length and width of the chewed mixture sample the area of the unmixed red area (the range of the image R component is 150-250 area), the unmixed green area (the range of the image G component) in the sample 100-245 area), the total area of the mixture samples before and after removing the holes (the default thickness is less than 0.1mm is the hole, the color is close to white in the image, and the R, G, B component values are all greater than 250), and chewing After the sample red optical density and green optical density and other parameters.
  • Measurement parameters include:
  • A represents the area of the sample with the hole removed
  • TA represents the sample area after chewing
  • OA represents the area of the sample before chewing
  • RA represents the color area of color A
  • GA represents the area of color B
  • L represents the length of the sample
  • W represents the sample Width
  • DR is the optical density of A color
  • DG is the optical density of B color
  • M is the average optical density
  • Fisher discriminant analysis method is to train a large number of sample data with known classification, and then use several indicators of the sample as independent variables, establish a linear discriminant equation according to the principle of the smallest variance within the group, and bring the evaluation indicators of the samples to be evaluated into the discriminant Equations to complete the classification of new samples.
  • the discriminant function for the measurement parameters in the chewing efficiency evaluation model is:
  • x 1 is the ratio MIX of the area sum of color A and color B in the sample to the total area
  • x 2 is the length-width relationship LW of the sample after chewing
  • x 3 is the ductility FF of the sample after chewing
  • x 4 is The ratio of the total area of the sample after chewing to the total area of the sample before chewing TA'
  • x 5 The variance of the optical density of A color and the optical density of B color in the mixed area of the sample after chewing
  • x 6 is the ratio of the sample area to the total area after removing the holes tr.
  • the coefficient terms and constant terms of the discriminant equation Y are obtained, and a trained chewing efficiency evaluation model is obtained;
  • the preparation of the training data set passes through three types of dentition: 1. Bilateral dentition loss. 2. Unilateral dentition loss. 3. The dentition is complete.
  • the test samples were chewed at five chewing speeds (40mm/min, 60mm/min, 80mm/min, 100mm/min and 120mm/min) and two types of chewing deformations (80% and 99%), respectively, to obtain 30 groups of chews Post-mix samples, place all chewed post-mix samples in a ziplock bag.
  • the 30 groups of mixed samples after chewing were classified according to the degree of color mixing and the ductility of the samples after chewing. They were divided into three categories: good chewing effect, average effect and poor effect by observation method. Among them, the criteria of the group with good chewing effect are: two colors can be well mixed, there is almost no single color area, and the area changes greatly after chewing, and the ductility is good.
  • the sample standard of the group with poor chewing effect is: the mixed color area is small, there is a large area of monochromatic area, the area of the sample before and after chewing has no obvious change, and the ductility is poor; the standard of the sample of the general chewing effect group is between the good group and the poor group. between.
  • the chewed mixed samples sorted in the ziplock bag were pressed into sheets with a thickness of about 2 mm with a glass plate, and then the chewed mixed samples were placed under the camera lens.
  • the center of the light source to ensure that there is no shadow in the photo, collect the images of the front and back sides, and get 60 images; in the software, measure the front and back sides of each picture twice and take the average value to obtain 6 measurement parameters after color mixing, take the positive value
  • the sum of the negative parameters is used as the measurement parameter for this sample. The details are shown in Table 1 below: Sample measurement parameters:
  • the 30 mixed samples after mastication were divided into three grades: good chewing effect, average chewing effect and poor chewing effect.
  • the measurement parameters corresponding to the 30 mixed samples after mastication were selected as the discriminant factors, and a linear discriminant equation was established:
  • Y 2 3.710x 1 +0.068x 2 -0.021x 3 +10.144x 4 +0.013x 5 -0.6111x 6 +0.619
  • Fisher's discriminant method proposes that the discriminant equation is established when the sample covariance between the multi-class populations is the largest and the intra-group dispersion within the population is the smallest.
  • the eigenvalues and variance contribution rates of the two discriminant functions are obtained, as shown in Table 2: eigenvalues of the discriminant equation.
  • the variance contribution rate is the proportion of the eigenvalue in the sum of the eigenvalues, which can describe the interpretation of the discriminant equation for the sample.
  • Table 3 shows the central values of the two discriminant functions in various groups.
  • the discriminant function Y 1 is used as an example for illustration.
  • the mean of the function was -0.113, and the mean of the function in the group with poor chewing effect was 4.594. Therefore, for any sample, when its six discriminant factors are brought into the discriminant function Y 1 , it can be classified by comparing the distance between the function value and the function value at the center of the three categories.
  • Fig. 1 is a scatter diagram of the first discriminant function Y1 and the second discriminant function Y2 of three groups of chewing effects, through which the classification of the samples can be intuitively understood.
  • the samples in the good chewing effect group have better color mixing, and the ductility of the samples is significantly better than that of the other two groups. Therefore, the samples in the good group have better aggregation in the scatter plot, while the samples in the general group and the poor group can also be The best ones are concentrated near the center value of their respective groups, but the distance between the two groups is close, which is prone to misjudgment. When encountering such samples, it is necessary to combine two discriminant functions for classification.
  • the discriminant function Y 1 has a significantly stronger ability to classify samples than the discriminant function Y 2 ; therefore, in this embodiment, the first discriminant function Y 1 is used as an evaluation function for evaluating the chewing effect; that is, based on the discrimination
  • the coefficient term and constant term of the discriminant equation Y obtained by the function Y 1 are used to obtain a trained chewing efficiency evaluation model.
  • the calculation method of the mastication efficiency index MEI (Mastication efficiency index) of the trained mastication efficiency evaluation model is as follows:
  • the measurement parameters of 30 mixed samples after mastication are brought into the MEI and calculation formulas. It can be found that the distribution of MEI values in different mastication effect groups is significantly different, and the MEI values of the group with the best mastication effect are all less than is equal to -3, the MEI value of the group with poor chewing effect is greater than 3, while the MEI value of the group with normal effect is between -2 and 1, that is, the MEI value has a good evaluation effect on the chewing efficiency, and the higher the MEI score, The poorer the mixing degree of the sample, the lower the chewing ability, and conversely, the lower the MEI score, the more uniform the color mixing of the sample, and the higher the chewing ability.
  • S9 Input the parameters to be detected into the trained chewing efficiency evaluation model for classification detection, and obtain the chewing efficiency evaluation result corresponding to the sample image to be detected.
  • the classification of the chewing efficiency evaluation results includes: good, fair, and poor.
  • Substitute the parameters of the sample to be detected into the trained chewing efficiency evaluation model that is, into the discriminant equation Y whose number and constant terms have been determined, and the calculated value is called the chewing efficiency index MEI; MEI score The higher it is, the less well mixed the sample is and the lower the chewing ability.
  • a comparison test is carried out with a commonly used food chewing efficiency detection method: brittle food sieving and weighing method as a control.
  • the sieving method reflects the chewing ability of the oral cavity by sieving and weighing the chewed particles multiple times.
  • the peanut material selected in this experiment is simple and easy to obtain, and the measured value is stable, which can objectively and accurately reflect the chewing ability of the tested teeth.
  • serial number Sieving method (median particle size/g) Color mixing method (MEI value) serial number Sieving method (median particle size/g) Color mixing method (MEI value) 1 1.13 2.7 16 -4.44 1.3 2 1.35 4 17 3.49 3.6 3 1.26 4.5 18 -3.28 2.6 4 6.73 5.8 19 4.64 4.3 5 -3.66 1.5 20 4.21 4.1 6 -3.96 1.3 twenty one 1.27 1.3 7 4.18 4 twenty two -4.61 0.8 8 -1.83 2.7 twenty three -5.03 1.8 9 1.10 2.9 twenty four 5.56 4.6 10 -3.89 1.7 25 5.69 4.8 11 -1.79 2.6 26 4.36 4.2 12 2.84 4.3 27 -4.65 1.3 13 3.88 3.9 28 -3.91 1.2 14 -1.52 2 29 -1.45 3 15 -4.33 1 30 -3.94 1.7
  • the abscissa of the BA diagram is the median particle size of peanuts and the color mixing value MEI in the technical solution of the present invention.
  • the average value of , and the ordinate is the difference between the median particle size and the color mixing value MEI (sieving method-color mixing method). It can be seen from the BA diagram that the solid line marked as the average value in the middle is the average value of the difference between the two measurement methods.
  • the standard deviation sd of the difference is calculated as 2.653, then the consistency interval can be obtained as (8.1, -2.3), the upper and lower consistency limits are represented by the two dotted lines marked SD in the figure, and the error bar marked error is consistent Confidence intervals for sexual limits, and the dashed line labeled Difference represents a difference of 0 between the two measurements.
  • the measured value of the chewed sample by the sieving method is often larger than that of the technical solution of the present invention, and the y-coordinate value of most of the 30 points in the graph is positive.
  • the MEI value of the sample with good chewing effect is often negative, so the sample with better chewing effect is in The position in the BA diagram is closer to the upper left corner.
  • the median particle size and MEI value are both large and positive.
  • the measurement results of the two methods are also closer. At this time, the difference between the two methods fluctuates around the 0 line.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Public Health (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Geometry (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Spectrometry And Color Measurement (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

一种食品咀嚼效率的评估方法，以两种颜色食品级石蜡为原材料制成了用于咀嚼试验的混色试验材料，基于图像采集装置获取待检测样本图像，基于Fisher判别分析方法构建的咀嚼效率评价模型，从待检测样本图像提取待检测参数，输入到训练好的咀嚼效率评价模型中，进行自动识别分类。该方法检测用样品制作容易，检测过程简单，检测结果客观，且准确率高。

Description

一种食品咀嚼效率的评估方法 技术领域
本发明涉及食品咀嚼程度检测技术领域，具体为一种食品咀嚼效率的评估方法。
背景技术
咀嚼效率是指在一定时间内将一定量食物嚼碎的程度。咀嚼活动是在神经系统的调控下，咀嚼肌收缩，颌骨、牙齿等产生规律性运动，并在唇、颊及舌的协助下完成的一项复杂的活动。因此，研究食品的咀嚼效率对食品加工，义齿种植等诸多方面具有重要意义。常见的食品咀嚼效率包括主观测定方法和客观测定方法。主观测定方法即通过问卷调查、对待检测者访问进而了解其牙齿的咀嚼效率，此方法过于依赖待检测者的个人主观感受，标准不统一，结果不够准确。客观实验法主要包括脆性食品过筛称重法、花生米悬浊液比色法、硬化明胶吸附比色法和混合试验法，但是这些方法同时存在检测过程复杂，导致临床应用范围受限等问题。
技术问题
为了解决现有的针对咀嚼效率检测方法中，主观测定方法存在准确性不够，客观实验法检测过程复杂导致临床应用范围受限的问题，本发明提供一种食品咀嚼效率的评估方法，其检测用样品制作容易，检测过程简单，检测结果客观，且准确率高。
技术解决方案
本发明的技术方案是这样的，一种食品咀嚼效率的评估方法，其包括以下步骤：
S1：制作检测用样品；
S2：对所述检测样品进行咀嚼测试；
S3：获取咀嚼后混合样本；
其特征在于，其还包括以下步骤：
S4：将所述咀嚼后混合样本做压制成薄片形状，获得待检测样本片；基于图像采集装置采集所述待检测样本片正反两侧图像，记做：待检测样本图像；
S5：对所述待检测样本图像进行去畸变处理，获得去畸变待检测样本图像；
S6：对所述去畸变待检测样本图像提取测量参数作为待检测参数；
S7：基于Fisher判别分析方法构建咀嚼效率评价模型；
S8：训练所述咀嚼效率评价模型，获得训练好的所述咀嚼效率评价模型；
S9：将所述待检测参数输入到训练好的咀嚼效率评价模型进行分类检测，获得所述待检测样本图像对应的咀嚼效率评价结果。
其进一步特征在于：
所述检测用样品以食品级石蜡为原料，包括颜色A、颜色B的双色矩形蜡片；将所述双 色矩形蜡片颜色交叉放置后压制成矩形；
所述双色矩形蜡片10mm×10mm×2mm的长方体蜡片；所述颜色A、颜色B设置为红色和绿色；
所述测量参数包括：
样品中A颜色区域和B颜色区域与总面积之比MIX：
Figure PCTCN2020115217-appb-000001
去除孔洞后样品面积与总面积之比TR：
Figure PCTCN2020115217-appb-000002
咀嚼后样品的长宽关系LW：
Figure PCTCN2020115217-appb-000003
咀嚼后样品的延展性FF：
Figure PCTCN2020115217-appb-000004
咀嚼后样品总面积与咀嚼前样品总面积之比TA’：
Figure PCTCN2020115217-appb-000005
咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差：
Figure PCTCN2020115217-appb-000006
其中：A代表除去孔洞部分样品面积，TA代表咀嚼后样品面积，OA为咀嚼前样品的面积，RA代表颜色A的色区域面积，GA代表颜色B的区域面积，L代表样品长度，W为样品宽度，DR为A颜色光密度，DG为B颜色光密度，M为平均光密度；
步骤S7中，所述咀嚼效率评价模型中对于所述测量参数的判别函数为：
Y＝c 1x 1+c 2x 2+c 3x 3+c 4x 4+c 5x 5+c 6x 6+a
式中：x 1为样品中颜色A、颜色B的区域之和与总面积之比MIX，x 2为咀嚼后样品的长宽关系LW、x 3为咀嚼后样品的延展性FF，x 4为咀嚼后样品总面积与咀嚼前样品总面积之比TA’，x 5咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差，x 6为去除孔洞后样品面积与总面积之比TR；
步骤S8中，基于训练数据集，得到的判别方程Y的系数项和常数项，获得训练好的所述咀嚼效率评价模型；
将被检测的样品的所述待检测参数，代入到训练好的所述咀嚼效率评价模型中，即代入到判别方程Y中，计算获得的值称为咀嚼效率指数MEI；MEI得分越高，样品混合程度越差，咀嚼能力越低；
所述图像采集装置包括：图像采集相机、光源；所述相机采用工业相机，所述光源采用LED光源，照射方式为背光照射加直接照射的光源照明方式；
步骤S5中，基于传统标定法对所述待检测样本图像进行去畸变处理，具体步骤包括：
a1：制作标准棋盘标定模板；
a2：基于所述图像采集相机对所述标准棋盘从不同角度拍摄，获取参照图片；
a3：对所述参照图片提取参照特征点；
a4：基于所述参照特征点计算获得所述图像采集相机对应的相机内参和畸变参数；
a5：提取所述待检测样本图像的R、G、B矩阵得到三个二维矩阵，记做待检测样本二维矩阵；
a6：根据所述图像采集相机对应的相机内参和畸变参数，对三个所述待检测样本二维矩阵进行线性插值去畸变运算，得到校正后图像数据矩阵；
a7：将三个所述校正后图像数据矩阵合并，得到所述去畸变待检测样本图像；
制成的所述检测用样品放置至于干燥低温环境中密封保存；在实施步骤S2之前，对所述检测用样品用37℃的水浴加热1min；
所述咀嚼效率评价结果的分类包括：良好、一般、较差。
有益效果
本发明提供的一种食品咀嚼效率的评估方法，基于图像采集装置获取待检测样本图像，从待检测样本图像提取待检测参数，输入到训练好的咀嚼效率评价模型中，进行自动识别分类；整个过程只需要技术人人员对咀嚼后混合样本做压制成薄片形状，其余的步骤都是自动实施完成，极大的减少了对技术人员能力的依赖，同时无需被检测人员的主观判断，整个检测过程简单、易实施，检测结果客观、准确；基于Fisher判别分析方法构建的咀嚼效率评价模型样品的训练量越大，判别指标越丰富，判别方程的准确率就越高，尤其适用于食品咀嚼效率分类检测这种训练样品采集范围广、可以持续采集数据的分类检测需求，咀嚼效率评价模型训练使用时间越久，分类精度越高，确保了评估结果的准确性；本发明技术方案，以两种颜色食品级石蜡为原材料制成了用于咀嚼试验的混色试验材料，对咀嚼后混合样本的进行压片处理，人工操作工序非常简单，降低了认为操作出错的可能性，进一步确保了评估结果的准确性。
附图说明
图1为实施例中，第一判别函数Y 1和第二判别函数Y 2的散点分布图；图2为对比实验中花生中值粒径和MEI值之间的BA图。
本发明的实施方式
本发明一种食品咀嚼效率的评估方法，其包括以下步骤。
S1：制作检测用样品；检测用样品以食品级石蜡为原料，包括颜色A、颜色B的双色矩形蜡片；将双色矩形蜡片颜色交叉放置后压制成矩形；本发明实施例中，双色矩形蜡片 10mm×10mm×2mm的长方体蜡片；颜色A、颜色B设置为红色和绿色；三个绿色的、两个黄色的蜡片压制成的检测用样品大小正适合放入口腔中；制成的检测用样品放置至于干燥低温环境中密封保存；在实施步骤S2之前，对检测用样品用37℃的水浴加热1min；石蜡材料不溶于水，不与口腔中的消化酶发生化学反应，能最大限度的避免唾液的干扰，且廉价易得。此外，检测用样品制作过程相对简单，方形的形状在咀嚼过程中产生多个受力点，样品更容易变形，而多层石蜡间隔压制的样品在咀嚼过程中也更容易发生混合，混色效果更好。而石蜡材料相对于现有技术中使用的花生等天然实验材料质地较软，可以满足大部分口腔修复患者和牙齿缺失严重的老年人的实验需求，相对于酵母糖黏性更低，实验过程中不容易发生附着在牙齿上的问题。
S2：对检测样品进行咀嚼测试。
S3：获取咀嚼后混合样本。
S4：将咀嚼后混合样本进行清洗和干燥处理，然后压制成薄片形状，获得待检测样本片；基于图像采集装置采集待检测样本片正反两侧图像，记做：待检测样本图像；图像采集装置包括：图像采集相机、光源；相机采用工业相机，光源采用LED光源，照射方式为背光照射加直接照射的光源照明方式；工业相机相比较民用相机来说，具有稳定性强、工作时间长、体积小等特点；因后期检测过程中，待测物体的位置相对固定，因此发明技术方案中，使用固定焦距模式(即拍摄过程中镜头的焦距不变)来提高拍摄质量，确保图像采集装置采集到的待检测样本图像的准确性，进而提高最终评估结果的准确性。
本发明实施例中，检测用样品是边长为10mm的正方体，考虑到后期实验中可能对多个材料块同时咀嚼，为了有效地完成检测工作，相机的视野范围暂定为l×l＝60×60mm，检测精度需要达到p＝0.1mm，因此可以确定相机的分辨率应达到：
r·r＝(L/p)·(L/p)＝(60/0.1)×(60/0.1)＝600×600
本发明实施例中，选择了Easyvxin的YX6060镜头，该镜头的焦距范围为6～60mm，镜头最大成像尺寸2/3″。
因为经过咀嚼后混合样本表面平滑，在后续的参数提取过程中，材料的侧向轮廓和表面的颜色分布都非常重要；因此本发明技术方方案中，选择背光照射加直接照射的光源照明方式。顶部光源由4根HF-FX160160型号的条状led光源组成，照射方式为直接照射，侧向光源型号Exin-64LED，照射方式为背光照射；背光照射确保得到咀嚼后混合样本的清晰的轮廓，直接照射确保得到咀嚼后混合样本的表面特征，进而确保图像采集装置采集到的待检测样本图像的准确性。
S5：对待检测样本图像进行去畸变处理，获得去畸变待检测样本图像。
在本发明技术方案中的食品咀嚼效率评价过程中需要提取咀嚼后混合样本的尺寸和像素信息，但是由于镜头的制造和安装问题，往往会造成以下两点误差：
1.镜头的光轴不垂直于传感器所在的平面，即相机坐标系与图像坐标系所在平面不平行；
2.畸变现象，镜头因制造精度原因无法呈现完美的小孔成像，导致投影偏离射影直线，此时需要通过畸变参数来校正；
因此进行后续分析测量的前提，是做好相机标定工作，对图像进行去畸变处理；相机标定的目的是确定相机从三维空间到二维图像的转换矩阵，来实现从图片上像素点之间的对应关系中获得检测目标的空间三维几何信息，同时相机标定也可以对镜头和相机因制造或安装中产生的精度误差进行畸变矫正，相机标定是视觉检测的关键步骤。
本发明技术方案中，基于传统标定法对待检测样本图像进行去畸变处理，具体步骤包括：
a1：制作标准棋盘标定模板；
a2：基于图像采集相机对标准棋盘从不同角度拍摄，获取参照图片；
a3：对参照图片提取参照特征点；
a4：基于参照特征点计算获得图像采集相机对应的相机内参和畸变参数；
a5：提取待检测样本图像的R、G、B矩阵得到三个二维矩阵，记做待检测样本二维矩阵；
a6：根据图像采集相机对应的相机内参和畸变参数，对三个待检测样本二维矩阵进行线性插值去畸变运算，得到校正后图像数据矩阵；
a7：将三个校正后图像数据矩阵合并，得到去畸变待检测样本图像；
本发明技术方案中采用传统标定法对待检测样本图像进行去畸变处理，处理结果准确率高，同时，本发明技术方案针对的采集对象，非常容易获得，特别适合采用传统标定法这种参照数据越多，准确率越高的标定方法。
具体实施的时候，首先制作标准棋盘标定模板如图，棋盘共分7行9列，其中单个棋格尺寸50×50mm，打印好后将其附在平面度高的平板上，利用相机从不同角度拍摄照片20张。利用Matlab中的Toolbox_calibration工具箱进行标定，对每一张照片按照顺时针方向手动提取4个特征点，输入棋盘格子的实际尺寸，而后点击运行标定程序，系统会对20张照片自动进行迭代计算，结果收敛后输出相机内参和畸变参数到结果文件，得到的内参数和畸变参数。经检查，角点标定误差的标准差0.71735pixel能够满足使用要求。
得到的内参数和畸变参数后，对待检测样本图像进行去畸变处理。利用Matlab软件读入待检测样本图像的RGB像素矩阵，分别提取待检测样本图像的R、G、B矩阵得到三个二维矩阵，通过im2double函数对每个矩阵进行数据转换，根据标定得到的相机内参和畸变参数，对矩阵进行线性插值去畸变运算，最后将图片校正后的三维矩阵合并输出，得到彩色的去畸变待检测样本图像。
S6：对去畸变待检测样本图像提取测量参数作为待检测参数；
本实施例中，利用Image pro-Plus 6.0软件对图片中相关测量参数进行提取。具体步骤如下：
(1)导入图片，标定图像，将坐标由pixel转换成实际长度，通过enhance命令增强图像对比度，使显示更加鲜明；
(2)确定目标区域范围，通过count菜单下segmentation命令可以按照颜色自动区分目标区域与背景区域，结果准确度较高。手动提取样品的轮廓线，可以去除咀嚼后样品中厚度很薄区域及孔洞；
(3)选择测量参数，Count菜单下的measurement命令可以选择对目标区域的测量参数，主要包括面积、倾角、光密度、长度和宽度。View命令可以查看和导出测量结果；
通过以上几步操作，可以测量得到咀嚼后混合物样本的长度及宽度，样品中未混色的红色区域(图像R分量的范围为150～250区域)的面积、未混色绿色区域(图像G分量的范围为100～245区域)的面积、去除孔洞前后混合物样品的总面积(默认厚度小于0.1mm处为孔洞，在图像中表现为颜色接近白色，R、G、B分量值均大于250)，以及咀嚼后样品红色光密度和绿色光密度等参数。
测量参数包括：
样品中A颜色区域和B颜色区域与总面积之比MIX：
Figure PCTCN2020115217-appb-000007
去除孔洞后样品面积与总面积之比TR：
Figure PCTCN2020115217-appb-000008
咀嚼后样品的长宽关系LW：
Figure PCTCN2020115217-appb-000009
咀嚼后样品的延展性FF：
Figure PCTCN2020115217-appb-000010
咀嚼后样品总面积与咀嚼前样品总面积之比TA’：
Figure PCTCN2020115217-appb-000011
咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差：
Figure PCTCN2020115217-appb-000012
其中：A代表除去孔洞部分样品面积，TA代表咀嚼后样品面积，OA为咀嚼前样品的面积，RA代表颜色A的色区域面积，GA代表颜色B的区域面积，L代表样品长度，W为样品宽度，DR为A颜色光密度，DG为B颜色光密度，M为平均光密度；
S7：基于Fisher判别分析方法构建咀嚼效率评价模型；
Fisher判别分析方法是通过对大量分类已知的样本数据进行训练，然后以样品的几个指标作为自变量，根据组内方差最小的原则建立线性判别方程，将待评价样品的评价指标带入判别方程，完成新样品的分类工作。
咀嚼效率评价模型中对于测量参数的判别函数为：
Y＝c 1x 1+c 2x 2+c 3x 3+c 4x 4+c 5x 5+c 6x 6+a
式中：x 1为样品中颜色A、颜色B的区域之和与总面积之比MIX，x 2为咀嚼后样品的长宽关系LW、x 3为咀嚼后样品的延展性FF，x 4为咀嚼后样品总面积与咀嚼前样品总面积之比TA’，x 5咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差，x 6为去除孔洞后样品面积与总面积之比TR。
S8：训练咀嚼效率评价模型，获得训练好的咀嚼效率评价模型；
基于训练数据集，得到的判别方程Y的系数项和常数项，获得训练好的咀嚼效率评价模型；
本实施例中，训练数据集的准备通过三类牙颌：1.双侧牙列缺失。2.单侧牙列缺失。3.牙列完整。分别在五种咀嚼速度(40mm/min、60mm/min、80mm/min、100mm/min和120mm/min)和两种咀嚼形变(80％和99％)下对检测用样品进行咀嚼得到30组咀嚼后混合样本，将所有咀嚼后混合样本放入自封袋。
对30组咀嚼后混合样本按照其混色程度，咀嚼后样品的延展性等参数进行分类，通过观察法对其分成咀嚼效果良好、效果一般以及效果较差三类。其中咀嚼效果良好组标准为：两种颜色能很好的混合，几乎没有单颜色区域，并且咀嚼后面积变化大，延展性很好。咀嚼效果较差组的样品标准为：混色区域小，出现大面积单色区域，咀嚼前后样品的面积无明显的变化，延展性差；咀嚼效果一般组样品的标准介于良好组和较差组之间。
对自封袋内分类好的咀嚼后混合样本用玻璃板压成厚度约2mm的片状，然后将片状咀嚼后混合样本放置在相机镜头下，拍摄时在暗光环境下，将样品放置在环形光源的中心，保证照片中无阴影，采集其正反面的图像，得到60张图像；在软件中对每一张图片正反两面测量2次取平局值，得到6个混色后测量参数，取正反面参数之和作为该样本的测量参数。具体如下面表1：样本测量参数：
表1：样本测量参数
Figure PCTCN2020115217-appb-000013
将30个咀嚼后混合样本分成咀嚼效果良好、咀嚼效果一般和咀嚼效果较差三个等级，选取30个咀嚼后混合样本对应的测量参数作为判别因子，建立线性判别方程：
Y＝c 1x 1+c 2x 2+c 3x 3+c 4x 4+c 5x 5+c 6x 6+a
本实施例中，采用SPSS软件的Fisher判别分析功能对咀嚼效果进行分析，得到的判别方程的系数项和常数项，进而建立起两个判别函数为：
Y 1＝6.695x 1+0.705x 2-0.028x 3-0637x 4+0.056x 5+-2.995x 6-4.157
Y 2＝3.710x 1+0.068x 2-0.021x 3+10.144x 4+0.013x 5-0.6111x 6+0.619
Fisher判别法提出，当多类总体之间的样本协方差最大，总体内的组内离差最小时，此时的判别方程成立。求得两个判别函数的特征值和方差贡献率，如表2：判别方程特征值，方差贡献率即为该特征值在特征值总和中的占比，能够描述判别方程对于样品的解释量。
表2：判别方程特征值
函数 特征值 方差占比 累计占比
Y 1 14.616 96.3 96.3
Y 2 0.563 3.7 100.0
表3为两个判别函数在各类组别中的中心值大小，以判别函数Y 1为例进行说明，判别函数Y 1在咀嚼效率良好组的函数平均值为-4.084，在咀嚼效果一般组的函数平均值为-0.113，在咀嚼效果较差组的函数平均值为4.594。因此，对于任意样品，将其6个判别因子带入判别函数Y 1时，可以通过比较函数值与三类中心处函数值的距离将其归类。
表3：各类中心处的判别函数值
Figure PCTCN2020115217-appb-000014
图1为三组咀嚼效果第一判别函数Y 1和第二判别函数Y 2的散点分布图，通过该图可以直观地了解样品的分类情况。其中咀嚼效果良好组的样品混色程度佳，样品的延展性明显优于其他两个组，因此在散点图中良好组的样本聚集性更好，而一般组和较差组的样本虽然也能较好的集中在各自组别的中心值附近，但两组组间距离较近，容易发生误判，遇到此类样品时需要结合两个判别函数进行分类。从图中也能发现判别函数Y 1对样品的分类判别能力明显强于判别函数Y 2；因此本实施例中，将第一判别函数Y 1用作评价咀嚼效果的评价函数；即，基于判别函数Y 1得到的判别方程Y的系数项和常数项，获得训练好的咀嚼效率评价模型。
本实施例中，训练好的咀嚼效率评价模型咀嚼效率指数MEI(Mastication efficiency index)的计算方法如下：
MEI＝6.695x 1+0.705x 2-0.028x 3+-0.637x 4+0.056x 5+-2.995x 6-4.157
本实施例中，将30个咀嚼后混合样本的测量参数带入到MEI、计算公式中，可以发现，不同咀嚼效果分组的MEI值分布呈明显差异化，咀嚼效果最好组的MEI值均小于等于-3，咀嚼效果较差分组的MEI值均大于3，而效果一般组的MEI值在-2到1之间，即：MEI值对咀嚼效率有很好的评价效果，MEI得分越高，样品混合程度越差，咀嚼能力越低，相反，MEI得分越低，样品混色均匀，咀嚼能力更高。
S9：将待检测参数输入到训练好的咀嚼效率评价模型进行分类检测，获得待检测样本 图像对应的咀嚼效率评价结果。本实施例中，咀嚼效率评价结果的分类包括：良好、一般、较差。将被检测的样品的待检测参数，代入到训练好的咀嚼效率评价模型中，即代入到数项和常数项已经确定的判别方程Y中，计算获得的值称为咀嚼效率指数MEI；MEI得分越高，样品混合程度越差，咀嚼能力越低。
为了验证本发明技术方案的分类检测效果，以常用的的食品咀嚼效率检测方法：脆性食品过筛称重法做对照，进行对比试验。筛分法通过对咀嚼后颗粒进行多次筛分称重来反映口腔的咀嚼能力，本实验中选用的花生米材料简单易得，测量值稳定，能够客观准确的反映测试牙齿的咀嚼能力。
挑选大小和形状相似的熟花生米作为实验样品。实验前对每一颗花生米进行称重测量并记录，(称重仪器：乐祺圆盘电子天平秤，量程：300g精度：0.01g)，为了降低误差，花生米的称重测量在防风罩内进行。利用义齿实验平台分别对本实验中的混色材料和筛分法中的花生样品进行咀嚼。通过混色法得到30个石蜡样本的混色值MEI，通过筛分法得到相同实验条件下30个花生样本的中值粒径。用Excel软件整理实验数据，BA(Bland-Altman图法)图验证两种测量方法的一致性，SPSS19.0软件对数据做进一步的统计学分析，详细如表4：试验样本的中值粒径和MEI值。
表4：试验样本的中值粒径和MEI值
序号 筛分法(中值粒径/g) 混色法(MEI值) 序号 筛分法(中值粒径/g) 混色法(MEI值)
1 1.13 2.7 16 -4.44 1.3
2 1.35 4 17 3.49 3.6
3 1.26 4.5 18 -3.28 2.6
4 6.73 5.8 19 4.64 4.3
5 -3.66 1.5 20 4.21 4.1
6 -3.96 1.3 21 1.27 1.3
7 4.18 4 22 -4.61 0.8
8 -1.83 2.7 23 -5.03 1.8
9 1.10 2.9 24 5.56 4.6
10 -3.89 1.7 25 5.69 4.8
11 -1.79 2.6 26 4.36 4.2
12 2.84 4.3 27 -4.65 1.3
13 3.88 3.9 28 -3.91 1.2
14 -1.52 2 29 -1.45 3
15 -4.33 1 30 -3.94 1.7
用BA图判定本发明技术方案的混色法和筛分法测量结果的一致性情况，如图2所示，BA图的横坐标为花生的中值粒径和本发明技术方案中的混色值MEI的平均值，纵坐标为中值粒径与混色值MEI之差(筛分法-混色法)。由BA图可知，中间标记为平均值的实线为两种测量方法差值的均值
Figure PCTCN2020115217-appb-000015
差值的标准差sd经过计算为2.653，那么可以得到一致性区间为(8.1，-2.3)，上下一致性界限由图中标记为SD的两条虚线来表示，标记为误差的误差条为一致性界限的置信区间，标记为差值的虚线代表两种测量方法差值为0。
由BA图中可以看出，筛分法对于咀嚼后样本的测量值往往大于本发明技术方案，图中表现为30个点中大多数的y坐标值为正。这是因为在两种测量方法中，均是得分越低咀嚼效果越好，且在本发明技术方案中，咀嚼效果良好样本的MEI值往往为负值，因此咀嚼效果越好的样本，其在BA图中位置越靠近左上角。相反咀嚼效率较差的样本，其中值粒径和MEI 值都很大，且均为正值，自然两种方法的测量结果也更加接近，此时两种方法的差值在0线附近波动。从BA图总体来看，30个样本的数据点均落在了一致性界限之内，因此可以说明两种方法测量结果具有一致性，说明本发明技术方案的混色测量法可以替代脆性称重法进行咀嚼效率的测定。在获得同等水平的检测结果的基础上，本发明技术方案对操作人员个人能力依赖更少，整个检测过程简单、易实施，实施效率更高。

Claims (10)

  1. 一种食品咀嚼效率的评估方法，其包括以下步骤：
    S1：制作检测用样品；
    S2：对所述检测样品进行咀嚼测试；
    S3：获取咀嚼后混合样本；
    其特征在于，其还包括以下步骤：
    S4：将所述咀嚼后混合样本做压制成薄片形状，获得待检测样本片；基于图像采集装置采集所述待检测样本片正反两侧图像，记做：待检测样本图像；
    S5：对所述待检测样本图像进行去畸变处理，获得去畸变待检测样本图像；
    S6：对所述去畸变待检测样本图像提取测量参数作为待检测参数；
    S7：基于Fisher判别分析方法构建咀嚼效率评价模型；
    S8：训练所述咀嚼效率评价模型，获得训练好的所述咀嚼效率评价模型；
    S9：将所述待检测参数输入到训练好的咀嚼效率评价模型进行分类检测，获得所述待检测样本图像对应的咀嚼效率评价结果。
  2. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：所述检测用样品以食品级石蜡为原料，包括颜色A、颜色B的双色矩形蜡片；将所述双色矩形蜡片颜色交叉放置后压制成矩形。
  3. 根据权利要求2所述一种食品咀嚼效率的评估方法，其特征在于：所述双色矩形蜡片10mm×10mm×2mm的长方体蜡片；所述颜色A、颜色B设置为红色和绿色。
  4. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：所述测量参数包括：
    样品中A颜色区域和B颜色区域与总面积之比MIX：
    Figure PCTCN2020115217-appb-100001
    去除孔洞后样品面积与总面积之比TR：
    Figure PCTCN2020115217-appb-100002
    咀嚼后样品的长宽关系LW：
    Figure PCTCN2020115217-appb-100003
    咀嚼后样品的延展性FF：
    Figure PCTCN2020115217-appb-100004
    咀嚼后样品总面积与咀嚼前样品总面积之比TA’：
    Figure PCTCN2020115217-appb-100005
    咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差：
    Figure PCTCN2020115217-appb-100006
    其中：A代表除去孔洞部分样品面积，TA代表咀嚼后样品面积，OA为咀嚼前样品的面积，RA代表颜色A的色区域面积，GA代表颜色B的区域面积，L代表样品长度，W为样品宽度，DR为A颜色光密度，DG为B颜色光密度，M为平均光密度。
  5. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：步骤S7中，所述咀嚼效率评价模型中对于所述测量参数的判别函数为：
    Y＝c 1x 1+c 2x 2+c 3x 3+c 4x 4+c 5x 5+c 6x 6+a
    式中：x 1为样品中颜色A、颜色B的区域之和与总面积之比MIX，x 2为咀嚼后样品的长宽关系LW、x 3为咀嚼后样品的延展性FF，x 4为咀嚼后样品总面积与咀嚼前样品总面积之比TA’，x 5咀嚼后样品中混合区域里A颜色光密度和B颜色光密度的方差，x 6为去除孔洞后样品面积与总面积之比TR。
  6. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：步骤S8中，基于训练数据集，得到的判别方程Y的系数项和常数项，获得训练好的所述咀嚼效率评价模型；将被检测的样品的所述待检测参数，代入到训练好的所述咀嚼效率评价模型中，即代入到判别方程Y中，计算获得的值称为咀嚼效率指数MEI；MEI得分越高，样品混合程度越差，咀嚼能力越低。
  7. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：所述图像采集装置包括：图像采集相机、光源；所述相机采用工业相机，所述光源采用LED光源，照射方式为背光照射加直接照射的光源照明方式。
  8. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：步骤S5中，基于传统标定法对所述待检测样本图像进行去畸变处理，具体步骤包括：
    a1：制作标准棋盘标定模板；
    a2：基于所述图像采集相机对所述标准棋盘从不同角度拍摄，获取参照图片；
    a3：对所述参照图片提取参照特征点；
    a4：基于所述参照特征点计算获得所述图像采集相机对应的相机内参和畸变参数；
    a5：提取所述待检测样本图像的R、G、B矩阵得到三个二维矩阵，记做待检测样本二维矩阵；
    a6：根据所述图像采集相机对应的相机内参和畸变参数，对三个所述待检测样本二维矩阵进行线性插值去畸变运算，得到校正后图像数据矩阵；
    a7：将三个所述校正后图像数据矩阵合并，得到所述去畸变待检测样本图像。
  9. 根据权利要求3所述一种食品咀嚼效率的评估方法，其特征在于：制成的所述检测用样品放置至于干燥低温环境中密封保存；在实施步骤S2之前，对所述检测用样品用37℃的水浴加热1min。
  10. 根据权利要求1所述一种食品咀嚼效率的评估方法，其特征在于：所述咀嚼效率评价结果的分类包括：良好、一般、较差。
PCT/CN2020/115217 2020-08-04 2020-09-15 一种食品咀嚼效率的评估方法 WO2022027785A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010773952.2A CN111862245B (zh) 2020-08-04 2020-08-04 一种食品咀嚼效率的评估方法
CN202010773952.2 2020-08-04

Publications (1)

Publication Number Publication Date
WO2022027785A1 true WO2022027785A1 (zh) 2022-02-10

Family

ID=72953595

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/115217 WO2022027785A1 (zh) 2020-08-04 2020-09-15 一种食品咀嚼效率的评估方法

Country Status (2)

Country Link
CN (1) CN111862245B (zh)
WO (1) WO2022027785A1 (zh)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110495959A (zh) * 2019-09-21 2019-11-26 苏州欢益医疗科技有限公司 一种智能医疗美牙种牙方法及辅助诊断设备
CN111175191A (zh) * 2020-03-26 2020-05-19 江南大学 一种用于食品流变性检测的仿生咀嚼装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5843567B2 (ja) * 2011-10-26 2016-01-13 株式会社ロッテ キシリトールガム咀嚼力判定用カラースケール
CN104062223B (zh) * 2013-03-22 2016-12-28 华中农业大学 柑橘咀嚼性测定方法
DE102016107689A1 (de) * 2016-04-26 2017-10-26 Bredent Medical Gmbh & Co. Kg Verfahren zur Auswertung eines Kaufunktionstests

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110495959A (zh) * 2019-09-21 2019-11-26 苏州欢益医疗科技有限公司 一种智能医疗美牙种牙方法及辅助诊断设备
CN111175191A (zh) * 2020-03-26 2020-05-19 江南大学 一种用于食品流变性检测的仿生咀嚼装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
VACCARO GUSTAVO, PELÁEZ JOSÉ IGNACIO, GIL-MONTOYA JOSÉ ANTONIO: "A novel expert system for objective masticatory efficiency assessment", PLOS ONE, vol. 13, no. 1, 31 January 2018 (2018-01-31), pages e0190386, XP055894599, DOI: 10.1371/journal.pone.0190386 *

Also Published As

Publication number Publication date
CN111862245B (zh) 2022-11-04
CN111862245A (zh) 2020-10-30

Similar Documents

Publication Publication Date Title
Halazonetis et al. Novel software for quantitative evaluation and graphical representation of masticatory efficiency
Carter et al. Automated quantification of dental plaque accumulation using digital imaging
CN111879735B (zh) 一种基于图像的大米外观品质检测方法
WO2016179981A1 (zh) 一种新生儿黄疸的自动检测方法及系统
CN106875393A (zh) 一种基于图像分析的肿瘤细胞dna含量检测方法
CN110490826B (zh) 一种照相机中雾滴样本图像处理方法
CN106937109B (zh) 低成本判断摄像头分辨率水平的方法
CN105160346A (zh) 一种基于纹理和分布特征的舌苔腐腻识别方法
CN109827957A (zh) 一种基于计算机视觉的水稻叶片spad值估测方法及系统
CN104198325A (zh) 基于计算机视觉的烟丝中梗丝比例测定方法
CN113570538B (zh) 一种叶片rgb图像偏态分布参数信息采集及分析方法
Harris et al. Optimising the measurement of bruises in children across conventional and cross polarized images using segmentation analysis techniques in Image J, Photoshop and circle diameter measurements
WO2022027785A1 (zh) 一种食品咀嚼效率的评估方法
CN110009609A (zh) 一种快速检测黄粒米的方法
CN109781730B (zh) 一种快速识别垩白米的方法
Gigilashvili et al. The impact of optical and geometrical thickness on perceived translucency differences
CN108451501A (zh) 一种基于像素分析的鲜红斑痣颜色和面积评估方法
CN100410971C (zh) 数字图像颜色分析系统的分析方法
CN113838081A (zh) 一种基于机器视觉判别烤烟烟叶颜色均匀度的方法和装置
CN104000668B (zh) 一种可摘局部义齿铸造金属支架适合性的定量评价方法
CN106872471A (zh) 稻米垩白检测方法
CN113838123B (zh) 基于图像处理的烟丝形貌特征的测量方法
Storlie et al. Growth analysis of whole plants using video imagery
CN104165696A (zh) 一种材料表面颜色特征在线自动检测方法
CN112329537A (zh) 一种基于yolov3算法的粪便有形成分的检测方法

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: 20948801

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: 20948801

Country of ref document: EP

Kind code of ref document: A1