CN109871887A - A kind of novel trench oil detection method and detection device based on SVM - Google Patents
A kind of novel trench oil detection method and detection device based on SVM Download PDFInfo
- Publication number
- CN109871887A CN109871887A CN201910084360.7A CN201910084360A CN109871887A CN 109871887 A CN109871887 A CN 109871887A CN 201910084360 A CN201910084360 A CN 201910084360A CN 109871887 A CN109871887 A CN 109871887A
- Authority
- CN
- China
- Prior art keywords
- oil
- svm
- sample
- conductivity
- acid value
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 42
- 239000002253 acid Substances 0.000 claims abstract description 35
- 238000012360 testing method Methods 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 10
- 239000003921 oil Substances 0.000 claims description 77
- 239000008162 cooking oil Substances 0.000 claims description 33
- 238000000034 method Methods 0.000 claims description 21
- 239000013598 vector Substances 0.000 claims description 13
- 238000001448 refractive index detection Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 5
- 230000007480 spreading Effects 0.000 claims description 5
- 238000003892 spreading Methods 0.000 claims description 5
- 239000000523 sample Substances 0.000 description 61
- 238000012706 support-vector machine Methods 0.000 description 36
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 6
- 230000009977 dual effect Effects 0.000 description 5
- 238000013507 mapping Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 239000002699 waste material Substances 0.000 description 3
- RTZKZFJDLAIYFH-UHFFFAOYSA-N Diethyl ether Chemical compound CCOCC RTZKZFJDLAIYFH-UHFFFAOYSA-N 0.000 description 2
- 239000008157 edible vegetable oil Substances 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 235000021190 leftovers Nutrition 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 229930195730 Aflatoxin Natural products 0.000 description 1
- XWIYFDMXXLINPU-UHFFFAOYSA-N Aflatoxin G Chemical compound O=C1OCCC2=C1C(=O)OC1=C2C(OC)=CC2=C1C1C=COC1O2 XWIYFDMXXLINPU-UHFFFAOYSA-N 0.000 description 1
- 241000287196 Asthenes Species 0.000 description 1
- FMMWHPNWAFZXNH-UHFFFAOYSA-N Benz[a]pyrene Chemical compound C1=C2C3=CC=CC=C3C=C(C=C3)C2=C2C3=CC=CC2=C1 FMMWHPNWAFZXNH-UHFFFAOYSA-N 0.000 description 1
- 238000005481 NMR spectroscopy Methods 0.000 description 1
- 208000034699 Vitreous floaters Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 239000005409 aflatoxin Substances 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004587 chromatography analysis Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000008367 deionised water Substances 0.000 description 1
- 229910021641 deionized water Inorganic materials 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000012153 distilled water Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012921 fluorescence analysis Methods 0.000 description 1
- 239000004519 grease Substances 0.000 description 1
- 229910001385 heavy metal Inorganic materials 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 150000002978 peroxides Chemical class 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 235000015277 pork Nutrition 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 210000001835 viscera Anatomy 0.000 description 1
Landscapes
- Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)
Abstract
The invention discloses a kind of novel trench oil detection method and detection device based on SVM, acquire conductivity, index of refraction and the acid value value of known oil sample, and the oil sample includes normal oil and improper oil；Conductivity, index of refraction and the acid value Value Data in S1 are generated into data file by host computer, which are divided into training set and test set；Host computer establishes SVM model, and training set importing SVM model is trained, and reuses test set and tests trained SVM model, judges the correctness of SVM model foundation；Oil sample to be measured is analyzed and processed using trained SVM model, exports processing result.
Description
Technical Field
The invention belongs to the technical field of illegal cooking oil detection, and particularly relates to a novel illegal cooking oil detection method and device based on an SVM (support vector machine).
Background
The swill-cooked dirty oil is a general concept, and is a general name for various inferior oils, generally including waste cooking oil, waste frying oil, food and waste grease produced by related enterprises.
Generally speaking, the illegal cooking oil can be divided into the following categories: the first is narrow-sense swill oil, namely oil obtained by simply processing and refining greasy floaters in a sewer or leftovers and leftovers (generally called swill) of hotels and restaurants; second, the inferior pork, pig's viscera, pigskin process and oil produced after refining; and thirdly, the oil used for frying the food is reused after the use frequency of the oil exceeds the specified requirement or the oil is reused after some new oil is added into the oil.
The acid value, the peroxide value, the solvent residue, the content of heavy metal pollutants, aflatoxin b1, benzopyrene and the like of the illegal cooking oil are seriously exceeded, and the illegal cooking oil is very harmful to human bodies, but the judgment of the existing technical means on the illegal cooking oil is still to be improved.
The existing technical means mainly comprise a conductivity detection method, a refractive index detection method, a fluorescence analysis method, a chromatography method, a nuclear magnetic resonance identification method, a color development method, an acid value detection method, a water content determination method and the like. Wherein, researches show that the conductivity method can only be used for identifying oil samples with the swill oil content of more than 20 percent in the edible oil. Through experiments, the refractive index values of various greases are slightly different and are 1.4713-1.4768, and certain difficulties are caused if the greases are used for distinguishing oil sample types. Meanwhile, with the improvement of the production technology of the illegal cooking oil, the acid value of part of the illegal cooking oil meets the national standard of 4mg/g, so that the judgment of the illegal cooking oil by only measuring the acid value is not very accurate. Because various problems have certain defects, a method capable of detecting novel illegal cooking oil is urgently needed.
Disclosure of Invention
The purpose of the invention is as follows: for the problem that the recognition degree of the illegal cooking oil is not high when each method of a conductivity method, an acid value detection method and a refractive index detection method is used for independently detecting the illegal cooking oil, the invention provides a method for combining the three detection methods to improve the recognition degree of the illegal cooking oil, and meanwhile, aiming at the problem that the difference of refractive index values is small and the distinction is difficult, the invention provides a method for introducing the refractive index values into an SVM for training and constructing a classification hyperplane to detect the illegal cooking oil by taking the refractive index, the conductivity and the acid value as input matrix variables of the SVM.
The invention discloses a novel illegal cooking oil detection method based on an SVM (support vector machine), which comprises the following steps of:
s1; collecting the conductivity, refractive index and acid value of a known oil sample, wherein the oil sample comprises normal oil and abnormal oil;
s2: generating data files of the conductivity, the refractive index and the acid value data in the S1 by an upper computer, and dividing the data files into a training set and a testing set;
s3: the upper computer establishes an SVM model, a training set is led into the SVM model for training, the trained SVM model is tested by using a test set, and the accuracy of the establishment of the SVM model is judged;
s4: and analyzing and processing the oil sample to be detected by using the trained SVM model, and outputting a processing result.
Further, in S2, the data file includes conductivity, acid value, refractive index and corresponding tag variables of various oil samples, the tag variables including normal oil and abnormal oil.
Further, in S2, a training set and a testing set are randomly generated by a randderm function.
Further, in S3, the conductivity, acid value, and refractive index data input to the SVM model are subjected to data normalization.
Further, in S3, when training the SVM model, assume that the sample training set with size m { (x)i,yi) I ═ 1,2, …, m } is made up of two categories. If xiBelongs to the first class, then, it is recorded as yi1 is ═ 1; if xiBelong to the second category, then, remember yi-1; searching a classification hyperplane:
wherein ω is an n-dimensional vector and b is an offset;
defining a sample point xiSet of samples with minimum geometrical spacing to the classification hyperplane:
δ＝minδi(5)
the size of the geometric interval is related to the number N of false scores of the sample, and the relation is as follows
Wherein R ═ max | | | xi1,2, …, m, which is the value of the longest vector length in the sample;
selecting an optimal hyperplane under the condition of satisfying the formula (2), so that the maximum delta obtains the minimum error fraction N;
the optimal classification hyperplane is:
in the formula,to support vector sample weights, yiRepresenting the attribute of the training sample, and b is a parameter to be optimized;
wherein x represents x1，x2，…xlσ is a spreading constant of the kernel function.
The invention also discloses a novel illegal cooking oil detection device based on the SVM, which comprises a conductivity detection module for measuring the conductivity of the oil sample, an acid value detection module for measuring the acid value of the oil sample, a refractive index detection module for measuring the refractive index of the oil sample, a single chip microcomputer and an upper computer; the single chip microcomputer is connected with the conductivity detection module, the acid value detection module and the refractive index detection module; the upper computer is connected with the single chip microcomputer and used for receiving data sent by the single chip microcomputer, building and training an SVM model and judging whether the test oil sample is the illegal cooking oil or not.
Further, the upper computer divides the received data into a training set and a testing set, and trains and tests the established SVM model respectively.
Further, the optimal classification function in the SVM model is:
in the formula,to support vector sample weights, yiRepresenting the attribute of the training sample, and b is a parameter to be optimized;
wherein x represents x1，x2，…xlσ is a spreading constant of the kernel function.
Has the advantages that: compared with the prior art, the invention has the following advantages:
1. for the oil sample with the conductivity only capable of detecting the swill oil content in the edible oil reaching more than 20%, the acid value measurement is inaccurate for distinguishing the novel swill oil at the present stage, and the refractive index value is too close to the problem of not distinguishing the swill oil, the invention introduces three input attributes into the SVM for training, constructs a classification hyperplane, and improves the recognition degree of the swill oil detection by a method combining various modes.
2. The measurement range is wider, the traditional illegal cooking oil detection method can only detect partial oil samples and cannot detect other special oil samples, and most of illegal cooking oil can be detected only by providing a plurality of different known oil samples and performing a plurality of groups of training.
3. The cost is low, and compared with other methods for detecting the illegal cooking oil, the instrument is lower in price.
Drawings
FIG. 1 is a novel illegal cooking oil detection device of the present invention;
FIG. 2 is a lower computer main flow chart;
FIG. 3 is a main flow chart of the host computer;
FIG. 4 is a structure of an SVM (support vector machine);
FIG. 5 is a flowchart of an SVM algorithm;
Detailed Description
The invention is further elucidated with reference to the drawing.
The invention discloses a novel illegal cooking oil detection method based on an SVM, which comprises the following specific steps:
s1; acquiring the conductivity, refractive index and acid value of a known oil sample through a sensor, wherein the known oil sample comprises normal oil and abnormal oil;
s2: the single chip microcomputer sends the collected data to an upper computer, and a data file is generated to comprise a training set and a testing set;
s3: importing the training set into an SVM model for training, testing the model by using the test set, and judging the correctness of model establishment;
s4: and analyzing and processing the oil sample to be detected by using the trained model, and outputting a processing result.
In S3, when creating/training the SVM model, assume a training set of samples of size m { (x)i,yi) I ═ 1,2, …, m } is made up of two categories. If xiBelonging to the first category. Then remember yi1 is ═ 1; if xiBelong to the second category, then, remember yiIs-1. Searching a classification hyperplane:
ωx+b＝0 (1)
where ω is an n-dimensional vector and b is an offset.
The classification hyperplane can correctly divide the samples into two types, namely, the samples of the same type are divided on one side of the classification hyperplane, and the oil samples can be divided into normal oil and abnormal oil (illegal cooking oil). Namely, it is
Defining a sample point xiThe interval to the classification hyperplane indicated by the formula (1) is
εi＝yi(ωxi+b)＝|ωxi+b| (3)
To simplify the calculation, ω and b in the formula (2) are normalized byAndinstead of the original ω and b, the normalized interval is defined as the geometric interval, i.e. as
At the same time, a sample set is defined in which the geometric separation of the sample points from the classification hyperplane is minimal (i.e., the distance to the classification hyperplane is closest). Namely, it is
δ＝minδi(5)
The size of the geometric interval is related to the number N of false scores of the sample, and the relation is as follows
Wherein R ═ max | | | xiI 1,2, …, m, which is the longest vector length in the sample.
According to the formula (6), the maximum value of the number N of false scores is determined by delta, namely the larger the delta is, the smaller the N is, and under the condition that the formula (2) is met, an optimal hyperplane is selected from countless classification hyperplanes, so that the delta is the maximum, namely the distance between a sample set and the classification hyperplane, namely the distance between the hyperplanes is the maximum.
The optimal classification hyperplane can be obtained by solving the following quadratic optimization problem:
the formula is when the interval ε ═ ω xiWhen + b | ═ 1, the distance between two sample points isUnder the condition of satisfying the formula (2)Find the optimal classification hyperplane to makeMaximum, i.e. minimumThus, the compound was obtained.
This equation can be obtained by solving the saddle points of the Lagrangian function, i.e.
Wherein, aiMore than or equal to 0, i ═ 1, 2., m, lagrange coefficients.
Here the original problem is converted into a dual problem according to the Lagrangian dual theory, i.e.
It is specifically mentioned that since the problem studied by the present invention is non-linear, by non-linear mapping Φ: rd→ H, the samples of the original input space are mapped into a high-dimensional feature space H, converting the nonlinear problem into a linear problem. And constructing an optimal classification hyperplane in the high-dimensional feature space H, wherein the calculation amount is huge because the dot product needs to be calculated in the high-dimensional feature space. Thus, the dot product operation is replaced by a kernel function, i.e. K (x)i，xj) In place of xixjThe kernel function used is the radial basis kernel function (RBF):
wherein x represents x1，x2，…xiσ is a spreading constant of the radial basis function. Mapping to high-dimensional featuresAfter the space, the corresponding dual problem becomes
Meanwhile, there may be a few samples as outliers, which causes the problem of linear divisibility to be changed into the problem of linear inseparability, so the rule needs to modify the outlier sample points, i.e., introduce a slack variable, modify the optimization objective and the constraint term, and convert the optimization objective and the constraint term into the problem of linear divisibility. Here, the constraint term and optimization objective of equation 7 are modified, i.e.
Due to the complexity of the computation, it is converted here into a dual problem, the other being unchanged except that the constraints become
And C is a penalty factor, and is used for controlling the penalty degree of the misclassified samples and realizing the compromise between the proportion and the complexity of the misclassified samples. The larger C is, the larger the punishment is, the more the weight occupied by the wrong sample is considered, and if C is smaller, the weight can be ignored.
Finally, the dual problem of mapping to a high-dimensional feature space and introducing a relaxed variable becomes
The optimal solution obtained by the solution is set asThen the optimum ω*And b*Is composed of
Wherein x isrAnd xsA pair of support vectors in either of two categories.
The final optimal classification function is:
wherein,for the parameter to be optimized, the physical meaning is the weight of the support vector sample, yiAnd representing the attributes of the training samples, namely positive samples or negative samples, which are kernel functions for calculating inner products, and b is a parameter to be optimized.
Final classification hyperplane:
it is noted here that converting a non-linear problem into a linear problem, as it is said when introducing a relaxation variable, is different from mapping samples into a high-dimensional space. When a relaxation variable is introduced to convert the nonlinear problem into a linear problem, the original problem is linearly separable, and the problem is changed into the nonlinear problem only because of the abnormality of a few sample points. When the samples are mapped to the high-dimensional space, the samples are inseparable in the original low-dimensional space, no matter how to search the classification plane, a large number of outliers always exist, and at the moment, the kernel function is used for mapping the samples to the high-dimensional space, and although the result is still inseparable, the result is closer to a linear separable state (namely, a state of approximate linear separable is achieved) than that in the original space. I.e. to convert the non-linear problem into a linear problem. For better searching for hyperplane, the effect will be very significant if the outlier is treated with the relaxation variable.
And importing the training set into an SVM model for training, importing the test set into the SVM model for testing, and judging the correctness of model establishment. After the model is built, the unknown oil sample is sent to the device to know the type of the unknown oil sample.
As shown in fig. 1, the novel illegal cooking oil detection device based on the SVM comprises a conductivity detection probe 1, an acid value detection probe 2, a refractive index sample tank 3, a conductivity detection module 4, an acid value detection module 5, a refractive index detection module 6, an MSP430 single chip microcomputer 7, a display screen 8 and an upper computer 9.
The conductivity detection probe 1 is connected with the conductivity detection module 4 and is used for measuring the conductivity of the oil sample. After each measurement, the conductivity detection probe 1 needs to be cleaned by deionized water, and the position of the probe needs to be noticed, so that the probe needs to be completely immersed in the solution. Before measuring the conductivity of the water phase of the oil sample, because the viscosity of part of the oil sample is overlarge, proper petroleum ether needs to be added to reduce the viscosity of the oil sample, so that the conductive substance can be leached by water more quickly and completely, and the error of conductivity measurement is reduced.
The acid value detection probe 2 is connected with the acid value detection module 5 and is used for measuring the acid value of the oil sample, and the acid value detection module 5 needs to be calibrated before measurement. Meanwhile, after each sample is measured, the acid value detection probe 2 is cleaned by distilled water.
The refractive index detection sample groove 3 is connected with the refractive index detection module 6 and used for measuring the refractive index of the oil sample, and after data is measured once, the refractive index detection sample groove 3 is washed by clear water.
The conductivity detection module 4, the acid value detection module 5 and the refractive index detection module 6 are connected with the MSP430 singlechip 7, and detection data are sent to the singlechip through serial ports.
The display screen 8 is connected with the MSP430 singlechip 7 and is used for displaying the detection result.
The upper computer 9 is connected with the MSP430 single chip microcomputer 7 and used for receiving data sent by the single chip microcomputer, establishing/training an SVM model, judging whether the test oil sample is illegal cooking oil or not and sending the result to the lower computer.
As shown in fig. 2, the specific process of the lower computer part is as follows: the method comprises the steps of firstly, respectively measuring the conductivity, the acid value and the refractive index of a known oil sample through each module, then transmitting data to a single chip microcomputer through a serial port, then packaging the data by the single chip microcomputer, and transmitting the data to an upper computer, wherein the upper computer trains a model by an SVM algorithm. After the device is trained, an unknown oil sample is sent into the device to be tested, the upper computer feeds the result back to the lower computer, and finally the judgment result is displayed through a display screen on the single chip microcomputer.
As shown in fig. 3, the specific process of the upper computer portion is as follows: firstly, the lower computer sends data to the upper computer, and the upper computer judges whether the data of the lower computer is received or not. If the upper computer receives the data of the lower computer, the upper computer judges whether the data sent by the lower computer is enough for training or not, and the data is enough for training, then the upper computer establishes an SVM model and begins to train the device. And if the data are not enough for training, the upper computer waits until the data are enough for training, and then the model is established. After the model is established, the randomly generated test set is sent to the SVM model for verifying the correctness of the model establishment. When the model is correctly established, a device capable of distinguishing unknown oil samples is completed. At the moment, an unknown oil sample is sent to an instrument for detection, the upper computer judges the type of the oil sample through a series of processing, and the oil sample is sent to the lower computer.
As shown in fig. 4, the structure of the support vector machine is similar to that of the neural network, the output is a linear combination of intermediate nodes, each intermediate node corresponds to a support vector, and the final obtained optimal classification function is:
as shown in fig. 5, the upper computer receives data sent by the lower computer, establishes a training set and a test set, stores input attribute matrix variables, namely, the conductivity, acid value, and refractive index of various oil samples and corresponding label variables, namely, non-gutter oil (category 1) and gutter oil (category 2), in a data file, and randomly generates the training set and the test set by using a randderm function for the randomness of the data.
Because the three input attributes of the conductivity, the acid value and the refractive index do not belong to the same order of magnitude, the difference of input variables is large, and the data processing is not facilitated, the data normalization of an input matrix is required before the model is established. Namely, the data are mapped to the range of 0-1, and the accuracy and the convergence speed of the model are improved. Here, the mapminmax normalization function is used, and in order to meet the requirement of the mapminmax normalization function, the input matrix needs to be transposed, and in order to meet the requirement of the SVM toolbox, the input matrix needs to be transposed again.
Before creating/training the SVM, the influence of the kernel function and related parameters on the model is considered, that is, part of abnormal points in the sample training set can influence the model precision, so a relaxation variable is introduced. The compromise between the proportion of misclassified samples and the algorithm complexity is realized by a parameter C (penalty factor). Meanwhile, the selection of the kernel function also has important influence on the classification performance of the SVM, and in order to optimize the performance of the model, the RBF kernel function is selected. In consideration of the influence of the sum kernel function and related parameters, the invention adopts a grid method to carry out value taking on C/g (variance in the kernel function). Each time a new set of C/g is generated, the model is trained using the svmtrain function, and if the newly generated set of C/g maximizes the performance of the model, the latest set of C/g is retained, and if the performance of the model is not maximized, the previous set of C/g is retained. Thus, the optimal parameters are determined to train the model, and then the model is trained. Meanwhile, when the performance of the models is the same, in order to reduce the calculation time, a parameter combination with a smaller C (penalty factor) is selected. After the model is trained, the randomly generated test set is sent to the model for testing, so that the correctness of the model is verified. By this time, the entire device has been trained. If the type of the unknown oil sample needs to be detected, the type of the unknown oil sample can be obtained only by sending the unknown oil sample into the device for detection.
Claims (8)
1. A novel illegal cooking oil detection method based on SVM is characterized in that: the method comprises the following steps:
s1; collecting the conductivity, refractive index and acid value of a known oil sample, wherein the oil sample comprises normal oil and abnormal oil;
s2: generating data files of the conductivity, the refractive index and the acid value data in the S1, and dividing the data files into a training set and a testing set;
s3: establishing an SVM model, importing a training set into the SVM model for training, testing the trained SVM model by using a test set, and judging the correctness of the establishment of the SVM model;
s4: and analyzing and processing the oil sample to be detected by using the trained SVM model, and outputting a processing result.
2. The SVM-based novel illegal cooking oil detection method according to claim 1, characterized in that: in S2, the data file includes conductivity, acid value, refractive index, and corresponding tag variables for various oil samples, including normal and abnormal oils.
3. The SVM-based novel illegal cooking oil detection method according to claim 1, characterized in that: in S2, a training set and a test set are randomly generated by a randderm function.
4. The SVM-based novel illegal cooking oil detection method according to claim 1, characterized in that: in S3, the conductivity, acid value, and refractive index data input to the SVM model are subjected to data normalization.
5. The SVM-based novel illegal cooking oil detection method according to claim 1, characterized in that: in S3, when training the SVM model, assume a training set of samples of size m { (x)i,yi) I-1, 2, …, m } is composed of two classes, if xiBelongs to the first class, then, it is recorded as yi1 is ═ 1; if xiBelong to the second category, then, remember yi-1; searching a classification hyperplane:
wherein ω is an n-dimensional vector and b is an offset;
defining a sample point xiSet of samples with minimum geometrical spacing to the classification hyperplane:
δ＝minδi(5)
the size of the geometric interval is related to the number N of false scores of the sample, and the relation is as follows
Wherein R ═ max | | | xi1,2, …, m, which is the value of the longest vector length in the sample;
selecting an optimal hyperplane under the condition of satisfying the formula (2), so that the maximum delta obtains the minimum error fraction N;
the optimal classification hyperplane is:
in the formula,to support vector sample weights, yiRepresenting the attribute of the training sample, and b is a parameter to be optimized;
wherein x represents x1，x2，…xlσ is a spreading constant of the kernel function.
6. The utility model provides a novel gutter oil detection device based on SVM which characterized in that: the device comprises a conductivity detection module for measuring the conductivity of an oil sample, an acid value detection module for measuring the acid value of the oil sample, a refractive index detection module for measuring the refractive index of the oil sample, a single chip microcomputer and an upper computer; the single chip microcomputer is connected with the conductivity detection module, the acid value detection module and the refractive index detection module; the upper computer is connected with the single chip microcomputer and used for receiving data sent by the single chip microcomputer, building and training an SVM model and judging whether the test oil sample is the illegal cooking oil or not.
7. The novel gutter oil detection device based on SVM of claim 6, characterized in that: and the upper computer divides the received data into a training set and a testing set, and trains and tests the established SVM model respectively.
8. The novel gutter oil detection device based on SVM of claim 6, characterized in that: the optimal classification function in the SVM model is as follows:
in the formula,to support vector sample weights, yiRepresenting the attribute of the training sample, and b is a parameter to be optimized;
wherein x represents x1，x2，…xlσ is a spreading constant of the kernel function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910084360.7A CN109871887A (en) | 2019-01-29 | 2019-01-29 | A kind of novel trench oil detection method and detection device based on SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910084360.7A CN109871887A (en) | 2019-01-29 | 2019-01-29 | A kind of novel trench oil detection method and detection device based on SVM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109871887A true CN109871887A (en) | 2019-06-11 |
Family
ID=66918293
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910084360.7A Pending CN109871887A (en) | 2019-01-29 | 2019-01-29 | A kind of novel trench oil detection method and detection device based on SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109871887A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112924652A (en) * | 2021-02-02 | 2021-06-08 | 南京仁面集成电路技术有限公司 | FPGA (field programmable Gate array) acceleration-based portable low-power-consumption multi-feature illegal cooking oil detector |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103293118A (en) * | 2013-05-11 | 2013-09-11 | 江南大学 | Hogwash oil identification method based on near infrared reflectance spectroscopy |
| CN104764837A (en) * | 2014-01-08 | 2015-07-08 | 中国科学院沈阳应用生态研究所 | Hogwash oil identification method |
| WO2017079997A1 (en) * | 2015-11-10 | 2017-05-18 | 广东省生物资源应用研究所 | Method for fast detection of gutter oil by hyperspectral transmittance analysis |
-
2019
- 2019-01-29 CN CN201910084360.7A patent/CN109871887A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103293118A (en) * | 2013-05-11 | 2013-09-11 | 江南大学 | Hogwash oil identification method based on near infrared reflectance spectroscopy |
| CN104764837A (en) * | 2014-01-08 | 2015-07-08 | 中国科学院沈阳应用生态研究所 | Hogwash oil identification method |
| WO2017079997A1 (en) * | 2015-11-10 | 2017-05-18 | 广东省生物资源应用研究所 | Method for fast detection of gutter oil by hyperspectral transmittance analysis |
Non-Patent Citations (1)
| Title |
|---|
| 胡晓慧等: "基于气体传感器阵列的地沟油识别方法", 《测控技术》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112924652A (en) * | 2021-02-02 | 2021-06-08 | 南京仁面集成电路技术有限公司 | FPGA (field programmable Gate array) acceleration-based portable low-power-consumption multi-feature illegal cooking oil detector |
| CN112924652B (en) * | 2021-02-02 | 2022-04-15 | 南京仁面集成电路技术有限公司 | FPGA (field programmable Gate array) acceleration-based portable low-power-consumption multi-feature illegal cooking oil detector |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103033362A (en) | Gear fault diagnosis method based on improving multivariable predictive models | |
| CN110400231B (en) | Failure rate estimation method for electric energy metering equipment based on weighted nonlinear Bayes | |
| CN109993236A (en) | Few sample language of the Manchus matching process based on one-shot Siamese convolutional neural networks | |
| CN101738998B (en) | Industrial process monitoring system and method based on local discriminant analysis | |
| CN102721716A (en) | Edible oil quality inspection method based on 1H-nuclear magnetic resonance (NMR) fingerprint spectra and multivariate analysis | |
| CN111343147A (en) | Network attack detection device and method based on deep learning | |
| Saleh et al. | Palm oil classification using deep learning | |
| Mavi et al. | Mango ripeness classification system using hybrid technique | |
| CN109002810A (en) | Model evaluation method, Radar Signal Recognition method and corresponding intrument | |
| CN108875118A (en) | A kind of blast furnace molten iron silicon content prediction model accuracy estimating method and apparatus | |
| CN111931827B (en) | Hydraulic pump health status detection system based on multi-sensor information fusion | |
| CN113526282A (en) | Method, device, medium and equipment for diagnosing medium and long-term aging faults of elevator | |
| CN109886314B (en) | Kitchen waste oil detection method and device based on PNN neural network | |
| CN112016618A (en) | Measurement method for generalization capability of image semantic segmentation model | |
| CN116204825A (en) | Production line equipment fault detection method based on data driving | |
| CN109871887A (en) | A kind of novel trench oil detection method and detection device based on SVM | |
| CN112069621B (en) | Method for predicting residual service life of rolling bearing based on linear reliability index | |
| CN102809635B (en) | Methods for detecting and evaluating quality of peanuts suitable for soluble protein processing | |
| CN109146003A (en) | A kind of classifying identification method based on high light spectrum image-forming, apparatus and system | |
| CN110096708A (en) | A kind of determining method and device of calibration collection | |
| CN109886288A (en) | A kind of method for evaluating state and device for power transformer | |
| CN115795341A (en) | Two-dimensional piston pump health state assessment method based on variable rotating speed | |
| CN101710392B (en) | Important information acquiring method based on variable boundary support vector machine | |
| CN112070761B (en) | Prawn freshness nondestructive testing method based on deep learning | |
| CN110619167B (en) | Oil-gas-water three-phase flow phase state critical point identification method based on cluster analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190611 |
|
| RJ01 | Rejection of invention patent application after publication |