CN114740091A - Watermelon maturity detection method and system based on acoustic analysis and machine learning - Google Patents
Watermelon maturity detection method and system based on acoustic analysis and machine learning Download PDFInfo
- Publication number
- CN114740091A CN114740091A CN202210663997.3A CN202210663997A CN114740091A CN 114740091 A CN114740091 A CN 114740091A CN 202210663997 A CN202210663997 A CN 202210663997A CN 114740091 A CN114740091 A CN 114740091A
- Authority
- CN
- China
- Prior art keywords
- watermelon
- sample
- time
- maturity
- samples
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/04—Analysing solids
- G01N29/045—Analysing solids by imparting shocks to the workpiece and detecting the vibrations or the acoustic waves caused by the shocks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
- G01N29/4418—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a model, e.g. best-fit, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biochemistry (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Immunology (AREA)
- Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Acoustics & Sound (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention discloses a watermelon maturity detection method and system based on acoustic analysis and machine learning, wherein the method comprises the following steps: acquiring a knocking sound signal and weight of a watermelon sample to form a data set, classifying the watermelon sample according to maturity, and dividing the data set into a training set and a testing set; constructing a maturity detection model to calculate the maturity of the target watermelon sample; training a maturity detection model by using a training set, testing the maturity detection model by using a testing set, finally calculating and recording the accuracy, and continuing to execute the step of training the maturity detection model by using the training set until the accuracy is highest to obtain a trained maturity detection model; and acquiring the knocking sound signal and the weight of the watermelon test sample, and inputting the knocking sound signal and the weight into the trained maturity detection model to obtain the maturity of the watermelon test sample. The invention realizes the requirement of detecting the watermelon maturity, improves the detection accuracy, can finish the training and testing of the model in the small sample data set, and improves the applicability.
Description
Technical Field
The invention relates to the field of audio analysis, in particular to a watermelon maturity detection method and system based on acoustic analysis and machine learning.
Background
The watermelon yield in China is high, but the problems of insufficient competitiveness, small mouth quantity, low export unit price and the like exist in the world market. The reason is that the detection system of the watermelon ripeness degree in China falls behind, so that the watermelons with various ripeness degrees are mixed together, and the overall quality is reduced.
At present, a detection method based on acoustic characteristics is mainly adopted for detecting the watermelon ripeness, and because a sound signal generated by knocking the watermelon is easier to obtain, the information such as pulp hardness, absorption and reflection characteristics, size, water content and the like of the watermelon can be reflected, so that the ripeness of the watermelon can be judged.
Most of the existing detection methods based on acoustic characteristics adopt an acoustic analysis method with a single frequency point, firstly, Fourier transform or fast Fourier transform is adopted, frequency domain analysis is carried out on a sound signal, then, the single frequency point is extracted, then, mathematical operation is carried out on the single frequency point and quality characteristics, an evaluation index is obtained, the maturity is determined by comparing the evaluation index with the set range, and a large number of data samples are needed to determine the evaluation index range. Since a single frequency point is difficult to express the distribution of the acoustic features of the sample, because it is only one maximum frequency component in the frequency domain, other frequency domain features are ignored, and the variation of the sound signal in the time domain is completely lost. Meanwhile, the size of the evaluation index is easily affected by the error of the calculation result of a single frequency point, because the evaluation index is obtained by a simple mathematical operation relation. Therefore, the existing detection method based on the acoustic characteristics is poor in applicability, is easily influenced by the error of a single frequency point of a sample, and is low in accuracy.
Other watermelon ripeness detection methods, such as infrared spectrum detection and machine vision detection, have the defects of complex detection equipment, high cost, difficulty in obtaining evaluation indexes and the like.
Machine learning is one of the most intelligent features of artificial intelligence, the frontmost research field, and can solve the problem that the traditional method is too complicated or no known algorithm exists. The application field is wide and the application prospect is good. How to apply the machine learning method to the detection of the watermelon maturity is also a problem worthy of research.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems in the prior art, the invention provides the watermelon maturity detection method and system based on acoustic analysis and machine learning, so that the requirement of watermelon maturity detection is met, the detection accuracy is improved, the training and testing of the model can be completed in a small sample data set, and the applicability is improved.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a watermelon maturity detection method based on acoustic analysis and machine learning comprises the following steps:
acquiring a knocking sound signal and weight of a watermelon sample, forming a data set, classifying the watermelon sample in the data set according to maturity, and dividing the data set into a training set and a testing set;
constructing a maturity detection model, wherein the maturity detection model is used for extracting time-frequency characteristics of a knocking sound signal of a watermelon sample, calculating the distance between the target watermelon sample and other watermelon samples in a training set according to the time-frequency characteristics and corresponding weight, and finally taking the class with the highest occurrence frequency as the maturity of the target watermelon sample in the classes corresponding to K watermelon samples closest to the target watermelon sample;
training a maturity detection model by using the training set, testing the maturity detection model by using the testing set, finally calculating and recording the accuracy, and repeating the steps until the accuracy is highest to obtain the trained maturity detection model;
and acquiring the knocking sound signal and the weight of the watermelon test sample, and inputting the knocking sound signal and the weight into the trained maturity detection model to obtain the maturity of the watermelon test sample.
Further, the specific steps of extracting the time-frequency characteristics of the watermelon sample knocking sound signal comprise:
calculating an FFT frequency spectrum and extracting frequency spectrum characteristics according to the knocking sound signals of all watermelon samples;
determining a spectral resolution parameter of the STFT window function from the extracted spectral featuresfAccording to the time resolution parameterτAnd spectral resolution parameterfValue of (a) construct an STFT window function
；
According to the STFT window function
And the knocking sound signal of each watermelon sample, and calculating the time-frequency matrix of each watermelon sample.
Further, the function expression of the time-frequency matrix is as follows:
in the above formula, the first and second carbon atoms are,x(t) Is the knock sound signal of the watermelon sample,
is a constructed STFT window function, whereinτIn order to be able to time-resolve the parameters,fis a spectral resolution parameterf i , f i+1]The spectral interval in between represents the bandwidth of the spectral resolution.
Further, the specific step of calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency characteristics and the weight comprises the following steps:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set, and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set;
fusing weight parameters according to informationλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the normalized time-frequency matrix distance and the normalized weight characteristic distance.
Further, the other watermelon samples in the training set include the second watermelon sample of the targetkThe nearest watermelon sample, the target watermelon sample andkthe functional expression for the distance between the nearest watermelon samples is:
in the above formula, the first and second carbon atoms are,λin order to fuse the weight parameters to the information,m k is the target watermelon sample andkthe time-frequency matrix distance between the nearest watermelon samples,n k is the target watermelon sample andkthe weight characteristic distance between the nearest watermelon samples.
Further, before training the maturity detection model with the training set, the method further comprises optimizing a time resolution parameterτSpectral resolution parameterfHexinInformation fusion weight parameterλThe method specifically comprises the following steps:
if the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe value of (d) does not reach the preset maximum value, and the time resolution parameter is increased by the preset step lengthτSpectral resolution parameterfAnd information fusion weight parameterλPerforming a step of training a maturity detection model with the training set;
if the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe value of (d) reaches a preset maximum value, the highest accuracy is found from all recorded accuracies, and the time resolution parameter is usedτSpectral resolution parameterfAnd information fusion weight parameterλIs adjusted to the value corresponding to the highest accuracy, the step of training the maturity detection model with the training set is performed.
Further, after obtaining the tapping sound signal of the watermelon sample, the method further comprises a step of extracting an effective signal, and specifically comprises the following steps:
calculating the energy of each frame of the tapping sound signal, and reserving the frames with the energy larger than a preset threshold value to obtain the tapping sound signal after primary filtering;
and filtering the primarily filtered tapping sound signal to obtain a secondarily filtered tapping sound signal which is used as an effective signal of the tapping sound signal.
The invention also provides a watermelon maturity detection system based on acoustic analysis and machine learning, which comprises:
the system comprises a data acquisition unit, a training unit and a testing unit, wherein the data acquisition unit is used for acquiring knocking sound signals and weight of watermelon samples, forming a data set, classifying the watermelon samples in the data set according to maturity and dividing the data set into a training set and a testing set; the device is also used for acquiring a knocking sound signal and weight of the watermelon test sample, and inputting the knocking sound signal and weight into the trained maturity detection model to obtain the maturity of the watermelon test sample;
the model construction unit is used for constructing a maturity detection model, the maturity detection model is used for extracting time-frequency characteristics of a watermelon sample knocking sound signal, then calculating the distance between a target watermelon sample and other watermelon samples in a training set according to the time-frequency characteristics and corresponding weight, and finally taking the class with the highest occurrence frequency as the maturity of the target watermelon sample in the classes corresponding to K watermelon samples closest to the target watermelon sample;
and the model training and testing unit is used for training the maturity detection model by using the training set, testing the maturity detection model by using the testing set, calculating and recording the accuracy, and continuing training the maturity detection model by using the training set until the accuracy is highest to obtain the trained maturity detection model.
Further, when the maturity detection model extracts time-frequency characteristics of a watermelon sample tapping sound signal, the model construction unit is configured to perform the following steps:
calculating an FFT frequency spectrum and extracting frequency spectrum characteristics according to the knocking sound signals of all watermelon samples;
determining a spectral resolution parameter of the STFT window function from the extracted spectral featuresfAccording to the time resolution parameterτAnd spectral resolution parameterfConstructing an STFT window function;
and calculating the time-frequency matrix of each watermelon sample according to the STFT window function and the knocking sound signal of each watermelon sample.
Further, when the maturity detection model calculates the distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency characteristics and the weight, the model construction unit is configured to execute the following steps:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set, and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set;
according to information fusionWeight parameterλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the normalized time-frequency matrix distance and the normalized weight characteristic distance.
Compared with the prior art, the invention has the advantages that:
the method and the device realize the watermelon maturity detection requirement based on acoustic analysis, extract the time-frequency characteristic matrix containing time information and frequency information as the acoustic characteristics of the sample, and solve the problem that a single frequency point is easy to interfere as the acoustic characteristics. And the calculation formula of the time-frequency matrix is optimized by improving the STFT window function, so that the dimensionality of the time-frequency characteristic matrix is reduced, the subsequent calculation amount is reduced, and the characteristic capable of representing the sample is provided for the subsequent classification algorithm.
The invention integrates the time-frequency characteristic matrix and the quality characteristics to carry out machine learning classification, establishes the relationship between various information characteristics and the watermelon maturity, and solves the problems of non-uniform dimensionality and large order difference among various information characteristics. Machine learning classification fused with time-frequency characteristic matrix and quality characteristics does not need a large number of samples to set evaluation index range any more, training and testing can be carried out on a small sample data set, higher accuracy can be achieved, and applicability is improved.
Drawings
Fig. 1 is a technical architecture diagram of a first embodiment of the present invention.
Fig. 2 is a flowchart of a first embodiment of the present invention.
Fig. 3 is a flowchart of the maturity detection model according to the first embodiment of the present invention.
Fig. 4 is a comparison graph of the time-frequency features extracted in the first embodiment of the present invention and the time-frequency features extracted by the conventional method.
Fig. 5 is a comparison graph of an original signal and a valid signal of a tapping sound signal according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the drawings and specific preferred embodiments of the description, without thereby limiting the scope of protection of the invention.
Example one
There are many Time-frequency analysis methods based on acoustic analysis, such as Short-Time Fourier Transform (STFT), wavelet Transform, S-Transform, etc., where STFT is generally applied because of its fast calculation speed and good sound feature extraction effect. The STFT is a two-dimensional matrix that can cut a sound signal into frame signals of a specified duration by a window function, perform Fast Fourier Transform (FFT) on each frame signal to obtain frequency domain features, and combine the frequency domain features of the frames to obtain a time-frequency feature. However, since the window function of the STFT is fixed, the time resolution is not changed, and the spectral resolution of the FFT is also determined, if the signal features are mainly concentrated in the low frequency region, data redundancy in the high frequency region is caused, so that the subsequent calculation speed is slowed down. Meanwhile, the time resolution of the window function is also an important factor influencing the quality of the evaluation index, and the determination of the window function needs to be changed in a self-adaptive manner according to the sample.
There are many machine learning classification algorithms, such as K-Nearest Neighbor (KNN), support vector machine, logistic regression, random forest, etc., where KNN has the advantages of simple principle, easy calculation, etc. and is suitable for detecting watermelon maturity. KNN is judged by calculating the distance between the test sample characteristic and the training sample characteristic. In the distance calculation process, each feature of the sample is given the same weight, which results in some important feature points being ignored, and thus the classification accuracy is reduced. In the watermelon maturity detection, the time-frequency characteristic two-dimensional matrix and the weight characteristic obtained by acoustic analysis have too much difference in data dimension and magnitude, but have important influence on the final classification result, and the influence weight of the two-dimensional matrix and the weight characteristic needs to be balanced, which cannot be realized by the traditional KNN classification algorithm.
In order to solve the above problems, as shown in fig. 1, this embodiment provides a watermelon maturity detection method based on acoustic analysis and machine learning, which establishes an optimized STFT time-frequency feature extraction algorithm and an information fusion KNN classification algorithm, and combines them to meet the requirement of watermelon maturity detection, improve detection accuracy, complete a training test of a system in a small sample data set, and improve applicability.
As shown in fig. 2, the method for detecting watermelon maturity based on acoustic analysis and machine learning of the present embodiment includes the following steps:
s1) acquiring knocking sound signals and weight of watermelon samples, forming a data set, and classifying the watermelon samples in the data set according to maturity, wherein the watermelon samples are generally classified into categories of [ immature, mature ] or [ immature, mature and over mature ];
s2) dividing the data set into a training set and a testing set;
s3), constructing a maturity detection model, wherein as shown in FIG. 3, the maturity detection model utilizes an optimized STFT time-frequency feature extraction algorithm to extract time-frequency features of a watermelon sample knocking sound signal, further utilizes an information fusion KNN classification algorithm to take a target watermelon sample as a test sample, calculates the distance between the test sample and other watermelon samples in a training set according to the time-frequency features and corresponding weight, and finally takes the category with the highest occurrence frequency as the maturity of the test sample in the categories corresponding to K watermelon samples closest to the test sample;
s4) training the maturity detection model by using a training set, testing the maturity detection model by using a testing set, finally calculating and recording the accuracy, and repeating the steps until the accuracy is highest to obtain the trained maturity detection model;
s5) acquiring the knocking sound signal and the weight of the watermelon test sample, and inputting the maturity detection model trained in the steps to obtain the maturity of the watermelon test sample.
As shown in fig. 1, the optimized STFT time-frequency feature extraction algorithm specifically includes the following steps:
the method comprises the following steps: calculating an FFT spectrum of a data set, wherein the data set comprises knocking sound signals of all watermelon samples, the spectrum of the signals can directly reflect frequency components of the signals, the frequency characteristic distribution condition of the sound signals can be observed through the spectrum, and the expression of the FFT spectrum is as follows:
in the above formulax(t) Is the knock sound signal of the watermelon sample,mis the serial number of the watermelon sample,fis the frequency;
step two: calculating the frequency points of all watermelon samplesfThe variance of the watermelon tapping sound signal is used for representing the difference between the frequency point samples, the frequency resolution is divided according to the variance value because the variance can show the difference of the watermelon tapping sound signal on each frequency point, the bigger the variance of the frequency point is, the bigger the difference is, the bigger the differentiation degree of the watermelon maturity is, and the expression of the variance is as follows:
in the above formulaMIs the total number of samples and is,mis the serial number of the watermelon sample,X m (f) Is the actual value of the frequency point or points,X’ m (f) Is the average of the frequency points;
step three: constructing STFT window functions
Time-resolved parameters of STFT window functionτDetermining the window length; spectral resolution parameterfHas a value range off=[f 1 ,f 2 …f N ]Determining the value rangef=[f 1 ,f 2 …f N ]The expression is as follows
In the above formula, the first and second carbon atoms are,DX(f) For all watermelon samples at each frequency pointfThe variance of (a) is determined,Nis the number of STFT spectrum bandwidthsf i , f i+1]The spectral interval between them represents the bandwidth of the spectral resolution, which can be obtained from step one and step two, the window function
Spectral resolution parameter offThe distribution situation of the spectral characteristics of the adaptive data set is changed;
step four: according to STFT window function
The derivation formula for optimizing STFT transform is as follows
In the above formula, the first and second carbon atoms are,S(f i ,τ) Is a time-frequency two-dimensional matrix after extracting time-frequency characteristics of a knocking sound signal of a watermelon sample,
for the window function constructed in the previous step,x(t) Is the knock sound signal of the watermelon sample.
The optimized STFT time-frequency feature extraction algorithm solves the problem that the conventional STFT can only change the time resolution but can not change the spectrum resolution, and the spectrum resolution is changed in a manner of being adaptive to the distribution of sample spectrum features. Therefore, in step S2), according to the optimized STFT time-frequency feature extraction algorithm, the specific step of extracting the time-frequency feature of the watermelon sample tapping sound signal includes:
according to the knocking sound signals of all watermelon samples, calculating FFT frequency spectrums by the formulas (1) and (2) and extracting frequency spectrum characteristics;
determining the spectral resolution parameter of the STFT window function from the extracted spectral features by equation (3)fAccording to the time resolution parameterτAnd spectral resolution parameterfValue of (a) construct an STFT window function
；
According to the STFT window function
And the knocking sound signal of each watermelon sample, and calculating the time-frequency matrix of each watermelon sample by the formula (4).
As shown in FIG. 4, FIG. 4 (a) shows the FFT spectra of all watermelon samples calculated by equation (1), wherein the horizontal axis represents frequency and the vertical axis represents watermelon sample number. Fig. 4 (b) is a variance value calculated by formula (2), and it can be seen from fig. 4 (a) that the part of the spectrum with large feature difference is mainly concentrated in low frequency, and the difference between frequency points is not the same. Therefore, the optimized STFT time-frequency feature extraction algorithm of the present embodiment divides the frequency resolution according to the variance value. Fig. 4 (c) is a time-frequency matrix obtained by performing STFT transformation on the tapping sound signal of the watermelon sample according to a conventional STFT, and fig. 4 (d) is a time-frequency matrix obtained by performing STFT transformation on the tapping sound signal of the watermelon sample by using the optimized STFT time-frequency feature extraction algorithm of the embodiment, wherein the horizontal axis is time and the vertical axis is frequency, so that the feature distribution of the time-frequency matrix extracted by the optimized STFT time-frequency feature extraction algorithm is more obvious, and the data dimension is from [10,256 ]]Reduced to [10,32 ]]Wherein 10 is determined by a time parameter, which is the total length of time/time resolution of the signalτAnd 32 is the number of STFT spectrum bandwidths in the formula (3)N。
The KNN is used for calculating the distance between the time-frequency matrix and the weight of the watermelon samplew d (S d ,y d ) And its firstkA nearest neighbor samplew k (S k ,y k ) The distance between the two adjacent channels adopts a mode of calculating Euclidean distance, and the calculation formula is as follows:
in the above formula, the first and second carbon atoms are,S d (f i ,τ) AndS k (f i ,τ) Respectively represent watermelon samplesw d (S d ,y d ) And withw k (S k ,y k ) Adopting a time-frequency matrix obtained by optimizing an STFT time-frequency characteristic extraction algorithm,y d andy k are respectively watermelon samplesw d (S d ,y d ) Andw k (S k ,y k ) The weight of (c).
In the traditional KNN Euclidean distance calculation process, the difference between characteristic dimensions is not considered, but a time-frequency characteristic matrixS(f i ,τ) Is a two-dimensional matrix, while the weight data is a data point and the magnitude is not uniform. In order to solve these problems, the present embodiment proposes an information fusion KNN-based classification algorithm, as shown in fig. 1, including the following steps:
the method comprises the following steps: calculating watermelon samplesw d (S d ,y d ) The time-frequency matrix distance and the weight characteristic distance between the watermelon sample and other watermelon samples in the training set are used for calculating the weight characteristic distance of the watermelon samplew d (S d ,y d ) With it at the firstkThe nearest watermelon samplew k (S k ,y k ) For example, calculate watermelon samplew d (S d ,y d ) Andw k (S k ,y k ) Time-frequency matrix distance ofm k Characteristic distance from weightn k The expression is as follows:
in the formulae (6) and (7),S d (f i ,τ) AndS k (f i ,τ) Respectively represent watermelon samplesw d (S d ,y d ) Andw k (S k ,y k ) The time frequency matrix obtained by optimizing the STFT time frequency characteristic extraction algorithm is adopted,y d andy k are respectively watermelon samplesw d (S d ,y d ) Andw k (S k ,y k ) The weight of (c);
step two: sampling watermelonw d (S d ,y d ) Normalizing the time-frequency matrix distance and the weight characteristic distance of the watermelon sample with other watermelon samples in the training set to solve the problem of non-uniform characteristic magnitude orderw d (S d ,y d ) With it at the firstkThe nearest watermelon samplew k (S k ,y k ) Example, watermelon samplew d (S d ,y d ) And withw k (S k ,y k ) Time-frequency matrix distance ofm k And characteristic distance of weightn k The normalized expression is as follows:
in the formulae (8) and (9),m max ，m min respectively represent watermelon samplesw d (S d ,y d ) And training setThe maximum value and the minimum value in the time-frequency matrix distance of other watermelon samples,n max ，n min respectively represent watermelon samplesw d (S d ,y d ) The maximum value and the minimum value of the weight characteristic distances between the watermelon samples and other watermelon samples in the training set;
step three: importing information fusion weight parametersλCalculating watermelon sample according to normalized time-frequency matrix distance and weight characteristic distancew d (S d ,y d ) The Euclidean distance from other watermelon samples in the training set is calculated according to the watermelon samplesw d (S d ,y d ) And it's the firstkThe nearest watermelon samplew k (S k ,y k ) Example, watermelon samplew d (S d ,y d ) Andw k (S k ,y k ) The expression of Euclidean distance is as follows:
in the above formula, the first and second carbon atoms are,λin order to obtain the information fusion weight parameter,m k is a watermelon samplew d (S d ,y d ) Andw k (S k ,y k ) The distance of the time-frequency matrix between the two,n k is a watermelon samplew d (S d ,y d ) And withw k (S k ,y k ) The weight characteristic distance therebetween.
Obtaining the watermelon sample by the calculation of the third stepw d (S d ,y d ) After Euclidean distance from other watermelon samples in the training set, according to the following formulaSequencing in the order from small to large to find the watermelon samplew d (S d ,y d ) Counting the top K watermelon samples, namely counting the category with the highest frequency of occurrence in the categories of the K watermelon samples, namely the watermelon samplew d (S d ,y d ) The classification of (1) to obtain a watermelon samplew d (S d ,y d ) Maturity of (c).
The accuracy of the detection system is further improved by fusing the audio features and the quality features based on the information fusion KNN classification algorithm, and the problems of non-uniform dimensionality and large order difference among various information features are solved. Therefore, in step S2), the specific step of calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency characteristics and the corresponding weight based on the information fusion KNN includes:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set by the formula (6), and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set by the formula (7) according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set by the formulas (8) and (9);
fusing weight parameters according to informationλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set by the formula (10) according to the normalized time-frequency matrix distance and weight characteristic distance.
Step S4) in this embodiment is to test the accuracy and applicability of the model after the model is built, so as to optimize the performance of the model, wherein the maturity detection model is trained by the training set, so as to optimize and store the relevant parameters of the model, and the test set is used to test the trained maturity detection model, i.e., according to the stored relevant parameters of the model, the maturity detection model performs the feature extraction and classification described in step S2) on the watermelon samples in the test set, so as to obtain the maturity of the watermelon samples, and the specific process is shown in the dashed box of fig. 3, and the time-frequency feature of each watermelon sample is calculated and the weight feature is recorded according to the foregoing optimized STFT time-frequency feature extraction algorithm; taking the watermelon samples in the test set as test samples, and calculating Euclidean distances between the test samples and other watermelon samples in the training set according to the information fusion KNN classification algorithm; selecting the first K watermelon samples closest to each other according to increasing sorting; and finally, counting the category with the highest occurrence frequency in the K watermelon samples, namely the maturity of the test sample.
Because the watermelon samples in the test set are also classified in the step S1), the accuracy of the maturity detection model can be obtained by counting the number of watermelon samples with the maturity of the maturity detection model consistent with the actual maturity for the watermelon samples in the test set, and dividing the number by the total number of the watermelon samples in the test set.
While training and testing the maturity detection model, the problem of parameter optimization needs to be considered, and the parameters needing to be optimized comprise time resolution parametersτSpectral resolution parameterfAnd information fusion weight parameterλTherefore, the optimization time resolution parameter is included before training the maturity detection model with the training setτSpectral resolution parameterfAnd information fusion weight parameterλThe step (2) can adopt a grid search algorithm, a particle swarm search algorithm, a bird swarm search algorithm and the like, and takes the grid search algorithm as an example to optimize the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλComprises the following steps:
if the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe value of (a) does not reach the preset maximum value, and the time resolution parameter is increased by the preset step lengthτSpectral resolution parameterfAnd information fusion weight parameterλPerforming a step of training a maturity detection model with the training set;
if the time resolution parameterτSpectral resolutionParameter(s)fAnd information fusion weight parameterλThe value of (d) reaches a preset maximum value, the highest accuracy is found from all recorded accuracies, and the time resolution parameter is usedτSpectral resolution parameterfAnd information fusion weight parameterλIs adjusted to the value corresponding to the highest accuracy, the step of training the maturity detection model with the training set is performed.
In this embodiment, the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλCorresponding intervals and step sizes are preset, and in step S2), time resolution parameters are setτSpectral resolution parameterfAnd information fusion weight parameterλIs the minimum value of the corresponding preset interval, step S4) increases the time resolution parameter by the corresponding step each timeτSpectral resolution parameterfAnd information fusion weight parameterλUntil reaching the maximum value of the corresponding preset interval, at the moment, the time resolution parameter corresponding to the highest accuracyτSpectral resolution parameterfAnd information fusion weight parameterλIs the optimum value, so the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe well-trained maturity detection model has the optimal performance under the optimal value of (2).
Finally, the method for detecting the maturity of the watermelon based on acoustic analysis and machine learning of the embodiment considers the influence of the filtering algorithm and the preprocessing step on the noise immunity, and after the tapping sound signal of the watermelon sample is obtained, the method further comprises the step of extracting an effective signal, and specifically comprises the following steps:
calculating the energy of each frame of the tapping sound signal, and reserving the frames with the energy larger than a preset threshold value to obtain the tapping sound signal after primary filtering;
and filtering the primarily filtered tapping sound signal to obtain a secondarily filtered tapping sound signal which is used as an effective signal of the tapping sound signal.
For obtaining the tapping sound signal after one-time filtering, the recording time of the general watermelon tapping sound signal is longer, and an endpoint detection method is neededThe effect signal is extracted from the entire sound signal. The method can adopt methods of short-time energy, short-time zero-crossing rate, frequency domain characteristics and the like, and the method for detecting the end point of the sound signal based on the short-time energy is adopted in the embodiment and can judge the starting time point and the ending time point of the effective signal by calculating the energy of each frame of the sound signal. The sound signal isnThe short-time energy calculation formula of the frame is as follows:
in the above formula, the first and second carbon atoms are,Lwhich represents the length of the frame or frames,x(m) Indicates the watermelon samplemThe tapping sound signal of a frame.
FIG. 5 (a) is a sound signal of a watermelon sample when the watermelon sample is knockedE n If the value is larger than the preset threshold value, the knocking sound signal of the corresponding frame is reserved, and if the value is larger than the preset threshold value, the knocking sound signal of the corresponding frame is reservedE n If the value is smaller than the preset threshold value, the tapping sound signal of the corresponding frame is discarded so as to intercept the sound signal, and the finally once filtered tapping sound signal is as shown in fig. 5 (b), so that the accuracy of the time-frequency features extracted in the subsequent steps can be improved.
For the tapping sound signal after the secondary filtering is obtained, because the tapping sound signal of the watermelon sample may have interference of environmental noise, the detection accuracy is reduced, and all the tapping sound signals of the watermelon sample need to be filtered by a filtering algorithm. The adaptive filter can be a low-pass filter, a Butterworth filter, an adaptive filter and the like, the least mean square adaptive filter is adopted in the embodiment, and the adaptive filter can automatically correct weight coefficients by means of an effective adaptive algorithm so as to adapt to changes of an external environment, so that the filter is always kept in an optimal state, the adaptive filter has good adaptability and filtering performance, environmental noise can be effectively filtered, and the signal-to-noise ratio of effective signals is improved.
Example two
The embodiment provides a watermelon maturity detection system based on acoustic analysis and machine learning according to an embodiment, which comprises:
the system comprises a data acquisition unit, a training unit and a testing unit, wherein the data acquisition unit is used for acquiring knocking sound signals and weight of watermelon samples, forming a data set, classifying the watermelon samples in the data set according to maturity and dividing the data set into a training set and a testing set; the device is also used for acquiring the knocking sound signal and the weight of the watermelon test sample, and inputting the knocking sound signal and the weight into the trained maturity detection model to obtain the maturity of the watermelon test sample;
the model construction unit is used for constructing a maturity detection model, the maturity detection model utilizes an optimized STFT time-frequency feature extraction algorithm to extract time-frequency features of a watermelon sample knocking sound signal, also utilizes an information fusion KNN classification algorithm to take a target watermelon sample as a test sample, calculates the distance between the test sample and other watermelon samples in a training set according to the time-frequency features and corresponding weight, and finally takes the category with the highest occurrence frequency as the maturity of the test sample in the category corresponding to the K watermelon samples closest to the test sample;
and the model training and testing unit is used for training the maturity detection model by using the training set, testing the maturity detection model by using the testing set, calculating and recording the accuracy, and continuing training the maturity detection model by using the training set until the accuracy is highest to obtain the trained maturity detection model.
In this embodiment, when the maturity detection model extracts the time-frequency feature of the knocking sound signal of the watermelon sample, the model construction unit is configured to execute the following steps:
calculating FFT frequency spectrums according to the knocking sound signals of all watermelon samples by the formulas (1) and (2) and extracting frequency spectrum characteristics;
determining the STFT window function spectrum resolution parameter by the formula (3) according to the extracted spectrum characteristicsfAccording to the time resolution parameterτAnd spectral resolution parameterfValue of (a) constructs an STFT window function
According to the STFT window function
And the knocking sound signal of each watermelon sample, and calculating the time-frequency matrix of each watermelon sample by the formula (4).
In this embodiment, when the maturity detection model calculates the distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency characteristics and the weight, the model construction unit is configured to execute the following steps:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set by the formula (6), and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set by the formula (7) according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set by the formula (8) and the formula (9);
fusing weight parameters according to informationλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set by the formula (10) according to the normalized time-frequency matrix distance and weight characteristic distance.
The foregoing is considered as illustrative of the preferred embodiments of the invention and is not to be construed as limiting the invention in any way. The core idea of the invention is to adopt a combined model of a time-frequency analysis feature extraction algorithm and a machine learning classification algorithm to solve the problem of watermelon maturity detection, and to use sound signals generated by knocking watermelon and the quality of watermelon as a basis. Other schemes based on machine vision or infrared spectrum detection and the like are greatly different from the scheme of the invention in information sources and technical schemes, but the technical scheme of watermelon maturity detection based on acoustic analysis can be replaced.
Specifically, taking a time-frequency analysis method of the sound signal as an example, the method adopts an improved method based on STFT, and also adopts other basic time-frequency analysis methods and improved methods such as wavelet transformation, S transformation, Mel cepstrum analysis and the like to replace the method of the invention, but the essence of the methods is also the time-frequency analysis method, namely, the method can simultaneously extract the characteristics of the sound signal in the time domain and the characteristics of the sound signal in the frequency domain; taking a mechanical classification algorithm as an example, the invention adopts an information fusion algorithm based on KNN improvement, can replace classification algorithms of KNN, such as a support vector machine, a logistic regression, a random forest, a decision tree, a neural network and the like, but essentially needs to combine the time-frequency characteristics and the quality characteristics, and completes the watermelon maturity classification through the training and the testing of a model. The characteristic extraction algorithm of time-frequency analysis and the machine learning classification algorithm can be replaced and combined at will, the combination scheme adopted by the invention is a combination scheme with better performance after model training and testing are carried out on an actual data set, but other combination schemes can also realize the purpose of detecting the watermelon maturity.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention shall fall within the protection scope of the technical solution of the present invention, unless the technical essence of the present invention departs from the content of the technical solution of the present invention.
Claims (10)
1. A watermelon maturity detection method based on acoustic analysis and machine learning is characterized by comprising the following steps:
acquiring a knocking sound signal and weight of a watermelon sample, forming a data set, classifying the watermelon sample in the data set according to maturity, and dividing the data set into a training set and a testing set;
constructing a maturity detection model, wherein the maturity detection model is used for extracting time-frequency characteristics of a knocking sound signal of a watermelon sample, calculating the distance between the target watermelon sample and other watermelon samples in a training set according to the time-frequency characteristics and corresponding weight, and finally taking the class with the highest occurrence frequency as the maturity of the target watermelon sample in the classes corresponding to K watermelon samples closest to the target watermelon sample;
training a maturity detection model by using the training set, testing the maturity detection model by using the testing set, finally calculating and recording the accuracy, and repeating the steps until the accuracy is highest to obtain the trained maturity detection model;
and acquiring the knocking sound signal and the weight of the watermelon test sample, and inputting the knocking sound signal and the weight into the trained maturity detection model to obtain the maturity of the watermelon test sample.
2. The method for detecting the maturity of the watermelon according to claim 1, wherein the step of extracting the time-frequency characteristics of the knocking sound signal of the watermelon sample comprises the following steps:
calculating an FFT frequency spectrum and extracting frequency spectrum characteristics according to the knocking sound signals of all watermelon samples;
determining a spectral resolution parameter of the STFT window function from the extracted spectral featuresfAccording to the time resolution parameterτAnd spectral resolution parameterfValue of (a) construct an STFT window function
；
According to the STFT window function
And the knocking sound signal of each watermelon sample, and calculating the time-frequency matrix of each watermelon sample.
3. The method for detecting the maturity of watermelon according to claim 2, wherein the function expression of the time-frequency matrix is as follows:
in the above-mentioned formula, the compound has the following structure,x(t) Is the knock sound signal of the watermelon sample,
is a constructed STFT window function, whereinτIs time of dayThe resolution parameters are used to determine the resolution parameters,fis a spectral resolution parameter [ alpha ], [ alpha ]f i , f i+1]The spectral interval in between represents the bandwidth of the spectral resolution.
4. The method for detecting watermelon ripeness based on acoustic analysis and machine learning of claim 2, wherein the specific step of calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency characteristics and the weight comprises:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set, and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set;
fusing weight parameters according to informationλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the normalized time-frequency matrix distance and the normalized weight characteristic distance.
5. The method of claim 4, wherein the other watermelon samples in the training set comprise the first watermelon sample of the target watermelonkThe nearest watermelon sample, the target watermelon sample andkthe functional expression for the distance between the nearest watermelon samples is:
in the above formula, the first and second carbon atoms are,λin order to obtain the information fusion weight parameter,m k is the target watermelon sample andkthe time-frequency matrix distance between the nearest watermelon samples,n k is the target watermelon sample andkat the mostThe weight characteristic distance between adjacent watermelon samples.
6. The method of claim 4, wherein training the maturity detection model with a training set further comprises optimizing a temporal resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe method specifically comprises the following steps:
if the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe value of (a) does not reach the preset maximum value, and the time resolution parameter is increased by the preset step lengthτSpectral resolution parameterfAnd information fusion weight parameterλPerforming a step of training a maturity detection model with the training set;
if the time resolution parameterτSpectral resolution parameterfAnd information fusion weight parameterλThe value of (d) reaches a preset maximum value, the highest accuracy is found from all recorded accuracies, and the time resolution parameter is usedτSpectral resolution parameterfAnd information fusion weight parameterλIs adjusted to the value corresponding to the highest accuracy, the step of training the maturity detection model with the training set is performed.
7. The method for detecting the maturity of watermelon based on acoustic analysis and machine learning of claim 1, wherein after acquiring the tapping sound signal of the watermelon sample, the method further comprises a step of extracting effective signals, and the method specifically comprises the following steps:
calculating the energy of each frame of the knocking sound signal, and reserving the frames with the energy larger than a preset threshold value to obtain the knocking sound signal after primary filtering;
and filtering the primarily filtered tapping sound signal to obtain a secondarily filtered tapping sound signal which is used as an effective signal of the tapping sound signal.
8. A watermelon maturity detection system based on acoustic analysis and machine learning, comprising:
the system comprises a data acquisition unit, a training unit and a testing unit, wherein the data acquisition unit is used for acquiring knocking sound signals and weight of watermelon samples, forming a data set, classifying the watermelon samples in the data set according to maturity and dividing the data set into a training set and a testing set; the device is also used for acquiring a knocking sound signal and weight of the watermelon test sample, and inputting the knocking sound signal and weight into the trained maturity detection model to obtain the maturity of the watermelon test sample;
the model construction unit is used for constructing a maturity detection model, the maturity detection model is used for extracting time-frequency characteristics of a watermelon sample knocking sound signal, then calculating the distance between a target watermelon sample and other watermelon samples in a training set according to the time-frequency characteristics and corresponding weight, and finally taking the class with the highest occurrence frequency as the maturity of the target watermelon sample in the classes corresponding to K watermelon samples closest to the target watermelon sample;
and the model training and testing unit is used for training the maturity detection model by using the training set, testing the maturity detection model by using the testing set, calculating and recording the accuracy, and continuing training the maturity detection model by using the training set until the accuracy is highest to obtain the trained maturity detection model.
9. The system of claim 8, wherein when the maturity detection model extracts time-frequency features of a watermelon sample tapping sound signal, the model construction unit is configured to perform the following steps:
calculating an FFT frequency spectrum and extracting frequency spectrum characteristics according to the knocking sound signals of all watermelon samples;
determining a spectral resolution parameter of the STFT window function from the extracted spectral featuresfAccording to the time resolution parameterτAnd spectral resolution parameterfConstructing an STFT window function;
and calculating a time-frequency matrix of each watermelon sample according to the STFT window function and the knocking sound signal of each watermelon sample.
10. The acoustic analysis and machine learning based watermelon maturity detection system of claim 9 wherein said maturity detection model, when calculating the distance between the target watermelon sample and other watermelon samples in the training set from time-frequency features and weight, said model building unit is configured to perform the steps of:
calculating the time-frequency matrix distance between the target watermelon sample and other watermelon samples in the training set according to the time-frequency matrix of the target watermelon sample and other watermelon samples in the training set, and calculating the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set according to the weight of the target watermelon sample and other watermelon samples in the training set;
normalizing the time-frequency matrix distance and the weight characteristic distance between the target watermelon sample and other watermelon samples in the training set;
fusing weight parameters according to informationλAnd calculating the distance between the target watermelon sample and other watermelon samples in the training set according to the normalized time-frequency matrix distance and the normalized weight characteristic distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210663997.3A CN114740091B (en) | 2022-06-14 | 2022-06-14 | Watermelon maturity detection method and system based on acoustic analysis and machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210663997.3A CN114740091B (en) | 2022-06-14 | 2022-06-14 | Watermelon maturity detection method and system based on acoustic analysis and machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114740091A true CN114740091A (en) | 2022-07-12 |
| CN114740091B CN114740091B (en) | 2022-09-06 |
Family
ID=82287013
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210663997.3A Active CN114740091B (en) | 2022-06-14 | 2022-06-14 | Watermelon maturity detection method and system based on acoustic analysis and machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114740091B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115272617A (en) * | 2022-08-29 | 2022-11-01 | 北京京航计算通讯研究所 | Virtual simulation display method and system for object acoustics |
| CN117007552A (en) * | 2023-10-07 | 2023-11-07 | 北京市农林科学院智能装备技术研究中心 | Watermelon maturity detection method, device, system, electronic equipment and storage medium |
| CN117969670A (en) * | 2024-04-02 | 2024-05-03 | 湖南大学 | Watermelon maturity rapid nondestructive detection method and system based on acoustic characteristics |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0450716A2 (en) * | 1990-04-03 | 1991-10-09 | Kalman Peleg | Method and equipment for measuring firmness of fruits and vegetables |
| JPH10318992A (en) * | 1997-05-20 | 1998-12-04 | Toshiba Eng Co Ltd | Fruits and vegetables judging system |
| JP2982802B1 (en) * | 1998-09-29 | 1999-11-29 | 松下電器産業株式会社 | Fruit texture measuring device |
| US20110040504A1 (en) * | 2009-08-11 | 2011-02-17 | Rong Zhi Xin Science and Technology Development (Beijing) Co., Ltd. | Fruit maturity determination method and system |
| CN104359978A (en) * | 2014-11-09 | 2015-02-18 | 兰州交通大学 | Ripeness-degree measuring device for watermelons |
| CN105403625A (en) * | 2015-12-05 | 2016-03-16 | 浙江大学 | Portable fruit maturity detection device |
| CN106525981A (en) * | 2016-12-19 | 2017-03-22 | 浙江大学 | Acoustic nondestructive detection experimental device for ripeness of thick-peel melons and fruits |
| CN206378427U (en) * | 2016-11-25 | 2017-08-04 | 江西老庄主绿色农业发展有限公司 | A kind of Watermelon Maturity detection means |
| CN111640451A (en) * | 2020-05-07 | 2020-09-08 | Oppo广东移动通信有限公司 | Maturity evaluation method and device, and storage medium |
| CN111652326A (en) * | 2020-06-30 | 2020-09-11 | 华南农业大学 | Improved fruit maturity identification method and identification system based on MobileNet v2 network |
| CN112529152A (en) * | 2020-12-03 | 2021-03-19 | 开放智能机器（上海）有限公司 | System and method for detecting watermelon maturity based on artificial intelligence |
| CN113295690A (en) * | 2021-05-17 | 2021-08-24 | 福州大学 | Strawberry maturity rapid discrimination method based on machine learning |
| CN113588787A (en) * | 2021-07-28 | 2021-11-02 | 广东工业大学 | Watermelon maturity detector and detection method |
| CN113984901A (en) * | 2021-09-29 | 2022-01-28 | 钱思行 | Method for detecting watermelon maturity by using sound signal of smart phone |
-
2022
- 2022-06-14 CN CN202210663997.3A patent/CN114740091B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0450716A2 (en) * | 1990-04-03 | 1991-10-09 | Kalman Peleg | Method and equipment for measuring firmness of fruits and vegetables |
| JPH10318992A (en) * | 1997-05-20 | 1998-12-04 | Toshiba Eng Co Ltd | Fruits and vegetables judging system |
| JP2982802B1 (en) * | 1998-09-29 | 1999-11-29 | 松下電器産業株式会社 | Fruit texture measuring device |
| US20110040504A1 (en) * | 2009-08-11 | 2011-02-17 | Rong Zhi Xin Science and Technology Development (Beijing) Co., Ltd. | Fruit maturity determination method and system |
| CN104359978A (en) * | 2014-11-09 | 2015-02-18 | 兰州交通大学 | Ripeness-degree measuring device for watermelons |
| CN105403625A (en) * | 2015-12-05 | 2016-03-16 | 浙江大学 | Portable fruit maturity detection device |
| CN206378427U (en) * | 2016-11-25 | 2017-08-04 | 江西老庄主绿色农业发展有限公司 | A kind of Watermelon Maturity detection means |
| CN106525981A (en) * | 2016-12-19 | 2017-03-22 | 浙江大学 | Acoustic nondestructive detection experimental device for ripeness of thick-peel melons and fruits |
| CN111640451A (en) * | 2020-05-07 | 2020-09-08 | Oppo广东移动通信有限公司 | Maturity evaluation method and device, and storage medium |
| CN111652326A (en) * | 2020-06-30 | 2020-09-11 | 华南农业大学 | Improved fruit maturity identification method and identification system based on MobileNet v2 network |
| CN112529152A (en) * | 2020-12-03 | 2021-03-19 | 开放智能机器（上海）有限公司 | System and method for detecting watermelon maturity based on artificial intelligence |
| CN113295690A (en) * | 2021-05-17 | 2021-08-24 | 福州大学 | Strawberry maturity rapid discrimination method based on machine learning |
| CN113588787A (en) * | 2021-07-28 | 2021-11-02 | 广东工业大学 | Watermelon maturity detector and detection method |
| CN113984901A (en) * | 2021-09-29 | 2022-01-28 | 钱思行 | Method for detecting watermelon maturity by using sound signal of smart phone |
Non-Patent Citations (3)
| Title |
|---|
| GAL RIBAK ET AL.: "Free flight maneuvers of stalk-eyed flies:do eye-stalks affect aerial turing behavior?", 《J COMP PHYSIOL A》 * |
| PUNJAB: "Fruit quality identification using image processing, machine learning, and deep learning: a review", 《ADVANCES AND APPLICATION IN MATHEMATICAL SCIENCES》 * |
| ROBERT G. DE LUNA ET AL.: "Tomato growth stage monitoring for smart farm using deep transfer learning with machine learning-based maturity grading", 《JOURNAL OF AGRICULTURAL SCIENCE》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115272617A (en) * | 2022-08-29 | 2022-11-01 | 北京京航计算通讯研究所 | Virtual simulation display method and system for object acoustics |
| CN115272617B (en) * | 2022-08-29 | 2023-05-02 | 北京京航计算通讯研究所 | Virtual simulation display method and system for object acoustics |
| CN117007552A (en) * | 2023-10-07 | 2023-11-07 | 北京市农林科学院智能装备技术研究中心 | Watermelon maturity detection method, device, system, electronic equipment and storage medium |
| CN117007552B (en) * | 2023-10-07 | 2024-02-06 | 北京市农林科学院智能装备技术研究中心 | Watermelon maturity detection method, device, system, electronic equipment and storage medium |
| CN117969670A (en) * | 2024-04-02 | 2024-05-03 | 湖南大学 | Watermelon maturity rapid nondestructive detection method and system based on acoustic characteristics |
| CN117969670B (en) * | 2024-04-02 | 2024-06-25 | 湖南大学 | Watermelon maturity rapid nondestructive detection method and system based on acoustic characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114740091B (en) | 2022-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114740091B (en) | Watermelon maturity detection method and system based on acoustic analysis and machine learning | |
| CN107393542B (en) | Bird species identification method based on two-channel neural network | |
| CN109034046B (en) | Method for automatically identifying foreign matters in electric energy meter based on acoustic detection | |
| CN102982804A (en) | Method and system of voice frequency classification | |
| Huang et al. | Large-scale weakly-supervised content embeddings for music recommendation and tagging | |
| CN114897023B (en) | Underwater sound target identification method based on extraction of sensitive difference features of underwater sound targets | |
| CN109408660A (en) | A method of the music based on audio frequency characteristics is classified automatically | |
| CN117727307B (en) | Bird voice intelligent recognition method based on feature fusion | |
| CN109378014A (en) | A kind of mobile device source discrimination and system based on convolutional neural networks | |
| CN115457966B (en) | Pig cough sound identification method based on improved DS evidence theory multi-classifier fusion | |
| CN105306673A (en) | Mobile terminal and automatic scene mode adjustment method thereof | |
| CN111540368B (en) | Stable bird sound extraction method and device and computer readable storage medium | |
| CN113936667A (en) | Bird song recognition model training method, recognition method and storage medium | |
| CN117275129B (en) | Access control system based on voice of Internet of things and face recognition | |
| CN111968669B (en) | Multi-element mixed sound signal separation method and device | |
| CN117219088A (en) | Continuous cough voice recognition method for live pigs in complex environment | |
| CN114863286B (en) | Mixed waste plastic classification method based on multi-algorithm collaborative optimization | |
| CN116347723A (en) | Control system of sound control type wall switch with adjustable lamplight sample color | |
| CN113361610B (en) | Intelligent identification method and system for wine production place | |
| Zhao | Anomalous sound detection based on convolutional neural network and mixed features | |
| CN115862639A (en) | Artificial intelligence voice analysis method based on K-means clustering analysis | |
| CN115046956A (en) | Zheng panel material detection method based on optical information | |
| CN117969670B (en) | Watermelon maturity rapid nondestructive detection method and system based on acoustic characteristics | |
| CN111274989A (en) | Deep learning-based field vehicle identification method | |
| CN112949671B (en) | Signal classification method and system based on unsupervised feature optimization |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |