CN114527241B - Wavelet transformation-capsule neural network cascading type gas identification method and device - Google Patents

Wavelet transformation-capsule neural network cascading type gas identification method and device Download PDF

Info

Publication number
CN114527241B
CN114527241B CN202210157351.8A CN202210157351A CN114527241B CN 114527241 B CN114527241 B CN 114527241B CN 202210157351 A CN202210157351 A CN 202210157351A CN 114527241 B CN114527241 B CN 114527241B
Authority
CN
China
Prior art keywords
layer
gas
neural network
capsule
signal
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.)
Active
Application number
CN202210157351.8A
Other languages
Chinese (zh)
Other versions
CN114527241A (en
Inventor
蒋坤
王磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN202210157351.8A priority Critical patent/CN114527241B/en
Publication of CN114527241A publication Critical patent/CN114527241A/en
Application granted granted Critical
Publication of CN114527241B publication Critical patent/CN114527241B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
    • G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
    • G01N33/0031—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array
    • G01N33/0034—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array comprising neural networks or related mathematical techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
    • G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
    • G01N33/0036—Specially adapted to detect a particular component
    • G01N33/004—Specially adapted to detect a particular component for CO, CO2
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
    • G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
    • G01N33/0036—Specially adapted to detect a particular component
    • G01N33/005—Specially adapted to detect a particular component for H2
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
    • Y02A50/20—Air quality improvement or preservation, e.g. vehicle emission control or emission reduction by using catalytic converters

Abstract

The invention relates to a wavelet transformation-capsule neural network cascading type gas identification method and device, comprising the steps of inputting mixed gas into a multi-channel sensor array to obtain a gas time domain signal; five layers of wavelet transformation is carried out on the gas time domain signal to obtain a signal sequence; the method comprises the steps of inputting a signal sequence into a capsule neural network after dimension transformation, and combining a dynamic routing algorithm, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer, so that a gas identification result is obtained. Compared with the prior art, the method has the advantages of high identification accuracy and the like.

Description

Wavelet transformation-capsule neural network cascading type gas identification method and device
Technical Field
The invention relates to the field of mixed gas detection, in particular to a wavelet transformation-capsule neural network cascading type gas identification method and device.
Background
The gas sensor is widely applied to tail gas monitoring, smoke alarm, civil gas leakage detection, dangerous gas monitoring in chemical plants and the like. Common gas sensors are classified into infrared type, thermal conduction type, catalytic combustion type, solid electrolytic type, metal oxide semiconductor (metal oxide semiconductor, MOS) and the like according to the working principle. Among them, the MOS gas sensor has the advantages of high sensitivity, small volume, low energy consumption, low cost, etc., and is widely paid attention to both industry and academia.
The gas sensing mechanism of the MOS gas sensor is based on the difference of oxidation-reduction property between the gas to be detected and oxygen, taking n-type semiconductor as an example, and taking a carrier as electrons, when the gas sensing material is positioned in the air, the oxygen in the air can abstract electrons from a conduction band of the gas sensing material to generate a space charge layer, and the electron barrier on the surface of a crystal is raised, so that the gas sensing material is in a high-resistance state. If the gas to be detected has stronger oxidizing property, the surface of the gas sensitive material further loses electrons, and the resistance value is further increased; on the contrary, if the gas to be detected has stronger reducibility, electrons are recovered from the surface of the gas-sensitive material, the resistance is reduced, and the reaction process is completely opposite for the p-type semiconductor. The MOS gas sensor converts the components and concentration information of the gas to be detected into an electric signal, and the information of the gas to be detected can be analyzed by performing digital processing analysis on the electric signal.
The prominent problem with MOS gas sensors is that they have cross sensitivity, i.e. there is no MOS gas sensor with single sensitivity, and therefore single MOS gas sensors perform poorly for component detection of mixed gas. In order to solve the problem, researchers on one hand start from the physical and chemical properties of the sensitive material, and improve the gas selectivity and inhibit the cross sensitivity through the modes of element doping, surface modification, micro-morphology modification and the like; on the other hand, a series of MOS gas sensors with different sensitivity characteristics are combined into an array sensor, and the component identification of the mixed gas is realized by comparing and analyzing the response signal difference of each channel of MOS gas sensor.
The identification algorithms based on the MOS gas sensor commonly used at present comprise a principal component analysis algorithm (PCA), an independent component analysis algorithm (ICA), a partial retention projection algorithm (LPP), a kernel principal component analysis algorithm (KPCA), a K nearest neighbor algorithm (KNN) and a support vector machine algorithm (SVM). The PCA, ICA and LPP algorithms are mainly used for extracting linear characteristics, and nonlinear characteristics in a dynamic response stage cannot be effectively extracted; although the KPCA algorithm can extract nonlinear characteristics, when the KNN algorithm is combined for classification, the calculated amount of the KNN algorithm is rapidly increased along with the increase of the sample size, and the classification time is prolonged; although the SVM can extract nonlinear characteristics and has low requirements on sample size, the kernel function of the SVM needs to meet the Mercer condition, and adjustment is needed to be made frequently for different gases to be detected.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a wavelet transformation-capsule neural network cascade type gas identification method and device.
The aim of the invention can be achieved by the following technical scheme:
a wavelet transformation-capsule neural network cascade type gas identification method comprises the following steps:
s1, inputting mixed gas into a multi-channel sensor array to obtain a gas time domain signal;
s2, five-layer wavelet transformation is carried out on the gas time domain signal to obtain a signal sequence;
s3, inputting the signal sequence after dimension transformation into a capsule neural network, and combining a dynamic routing algorithm to obtain a gas identification result, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer.
Further, in the five-layer wavelet transform, each layer of wavelet transform decomposes an input signal into a signal high-frequency part and a signal low-frequency part, the signal low-frequency part obtained by each layer is used as an input of the wavelet transform of the next layer, and finally the signal high-frequency part obtained by the fifth layer and the signal low-frequency part obtained by each layer are overlapped to be used as an output signal sequence.
Further, the number of the digital capsules in the digital capsule layer is the same as the number of the gas types in the gas to be detected.
Further, the dimension transformation in step S3 specifically includes:
and performing point-separating sampling on the signal sequence to obtain sampling data, and arranging the sampling data of all channels according to the channel sequence to obtain a wavelet transformation coefficient image after dimension transformation as the input of the capsule neural network.
Further, the input layer includes 256 9*9 convolution kernels with a stride of 1, and the convolution layer includes 32 9*9 convolution layers with a stride of 2.
A wavelet transformation-capsule neural network cascading type gas identification device comprises a memory and a processor; the memory is used for storing a computer program; the processor is configured to implement the following method when executing the computer program:
s1, inputting mixed gas into a multi-channel sensor array to obtain a gas time domain signal;
s2, five-layer wavelet transformation is carried out on the gas time domain signal to obtain a signal sequence;
s3, the signal sequence is input into a capsule neural network after dimension transformation, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer, and a dynamic routing algorithm is combined to obtain a gas identification result.
Compared with the prior art, the invention has the following advantages:
1. after the time domain signals of the mixed gas are extracted, five layers of wavelet transformation are performed, and low-frequency data and high-frequency data are simultaneously obtained, namely, linear characteristics in a steady state and nonlinear characteristics in a dynamic response stage are simultaneously extracted, so that the subsequent discrimination based on the characteristics is more accurate, and the recognition accuracy is improved; and the capsule neural network is used for outputting a discrimination result after processing a large amount of data, so that the calculation is simplified, and the recognition efficiency is improved.
2. In the setting of the capsule neural network, the invention only needs to set the number of the capsules of the digital capsule layer according to the gas type number of the gas to be detected, does not need to set other parameters due to the change of the gas, has strong universality and can be suitable for gas detection in different application scenes.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2 is a flow chart of the method for acquiring the time domain signal of the mixed gas.
Fig. 3 is a schematic diagram of a five-layer wavelet transform used in the present invention.
Fig. 4 is a schematic diagram of a encapsulated neural network for use with the present invention.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.
The embodiment provides a wavelet transformation-capsule neural network cascading type gas identification method, which is characterized by comprising the following steps of:
s1, inputting mixed gas into a multichannel sensor array, and obtaining a gas time domain signal through calculation of an acquisition board and an upper computer;
s2, performing five-layer wavelet transformation on the gas time domain signal, and acquiring a signal sequence according to the extracted low-frequency data and high-frequency data;
and S3, inputting the wavelet transformation coefficient image obtained after the signal sequence is subjected to dimensional transformation into a capsule neural network, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer, and acquiring a gas identification result by combining a dynamic routing algorithm.
In this embodiment, step S1 specifically includes:
first, the present embodiment uses CO gas and H 2 The gas and the gas mixed by the gas and the gas are used as the gas for detection, and the acquisition method comprises the following steps:
with 100% CO and 20% H 2 Concentration of N80% 2 The mixed gas of the two gas sources is used as a gas source, the gas proportion of the two gas sources is controlled through a mass flowmeter, and mixed gas samples with different concentrations are obtained after the gas mixing cavity is fully mixed.
After the mixed gas is obtained, the mixed gas is transmitted to a multi-channel sensor array, the 8-channel sensor array is adopted in the embodiment, the 8-channel 12-bit A/D conversion signal acquisition board acquires the mixed gas at the frequency of 10Hz, the signals are transmitted to an upper computer carrying Labview-based signal acquisition software, and finally an 8-channel gas time domain signal curve is obtained. The whole signal acquisition flow diagram is shown in fig. 1.
In this embodiment, step S2 specifically includes:
five layers of wavelet transformation are carried out on the gas time domain signals of each channel, each layer of wavelet transformation is used for decomposing an input signal into high-frequency data of reaction signal details and low-frequency data of reaction signal profiles, and the obtained low-frequency data is used as input of the next layer of wavelet transformation for further decomposition, as shown in fig. 3. In conjunction with the Mallat algorithm, the original signal can be expressed as a superposition of the low frequency data of the last layer (fifth layer) and all the high frequency data as follows:
S＝A 5 +D 5 +D 4 +D 3 +D 2 +D 1
where S represents a gas time domain signal, A5 represents low frequency data after the fifth layer wavelet transform, and Di represents high frequency data after the i-th layer wavelet transform.
And finally outputting the superimposed data as a signal sequence.
In this embodiment, step S3 specifically includes:
firstThe method comprises the steps of sampling signal sequences of all channels at intervals to obtain 512 pieces of data in total, obtaining 4096 pieces of data in total by 8 channels, sequentially arranging the pieces of data according to channel sequence, converting dimensions into wavelet transformation coefficient images with the size of 64 multiplied by 64, inputting the wavelet transformation coefficient images into an input layer of a capsule neural network, firstly adopting 256 convolution kernels with the size of 9 multiplied by 9, and carrying out convolution with the stride of 1, and selecting a ReLU (ReLU) as an activation function to obtain tensors with the size of 256 multiplied by 056 multiplied by 156, wherein the purpose is to extract low-level features in the input images. Then, the convolution layer is passed through, and the convolution kernel with the size of 32 9×29 is adopted, the stride is 2, the tensor of 256×356×456 is convolved, so as to obtain the tensor of 32×524×624, then the dimension is converted into 32×1×24×24, the operation in the convolution layer is repeated 8 times due to the total data of 8 channels, and the result is combined, so as to finally obtain the tensor of 32×8×24×24. Then through the main capsule layer, comprising a total of 32 main capsules, wherein each main capsule in turn comprises 24 x 24 1 x 8 sub-capsules. Since the gas types of the gases to be measured selected in the embodiment share CO gas and H 2 The three kinds of gases and the mixed gases are arranged in the digital capsule layer, so that three digital capsules are arranged in the digital capsule layer, each capsule receives 32 multiplied by 8 multiplied by 24 input tensor, the sub-capsules are subjected to matrix transformation through an 8 multiplied by 16 gesture matrix, and a 3 multiplied by 16 matrix is output by combining a dynamic routing algorithm. By the convolution dimension conversion, the data characteristics of the gas time domain signal can be fed back more accurately. The L2 norm of each row vector of the 3×16 matrix is the probability of the corresponding category, the row number of the maximum value in the 3 corresponding probabilities is input to the classifier layer, and the classifier layer can acquire the corresponding specific gas category and transmit the specific gas category to the output layer. An architecture diagram of the encapsulated neural network is shown in fig. 4.
Finally, the result of gas identification is obtained through the output layer.
The embodiment also provides a wavelet transformation-capsule neural network cascading type gas identification device, which comprises a memory and a processor; a memory for storing a computer program; a processor for implementing the following method when executing a computer program:
s1, inputting mixed gas into a multichannel sensor array, and obtaining a gas time domain signal through calculation of an acquisition board and an upper computer;
s2, performing five-layer wavelet transformation on the gas time domain signal, and acquiring a signal sequence according to the extracted low-frequency data and high-frequency data;
and S3, inputting the wavelet transformation coefficient image obtained after the signal sequence is subjected to dimensional transformation into a capsule neural network, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer, and acquiring a gas identification result by combining a dynamic routing algorithm.
The present embodiment further proposes a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements a wavelet transform-capsule neural network cascade type gas identification method as mentioned in the above embodiment, any combination of one or more computer-readable media may be employed. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.

Claims (8)

1. The wavelet transformation-capsule neural network cascading type gas identification method is characterized by comprising the following steps of:
s1, inputting mixed gas into a multi-channel sensor array to obtain a gas time domain signal;
s2, five-layer wavelet transformation is carried out on the gas time domain signal to obtain a signal sequence; in five-layer wavelet transformation, each layer of wavelet transformation decomposes an input signal into a signal high-frequency part and a signal low-frequency part, the signal low-frequency part obtained by each layer is used as the input of the wavelet transformation of the next layer, and the signal high-frequency part obtained by the fifth layer and the signal low-frequency part obtained by each layer are finally overlapped to be used as an output signal sequence;
s3, inputting the signal sequence after dimension transformation into a capsule neural network, and combining a dynamic routing algorithm to obtain a gas identification result, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer.
2. The wavelet transform-capsule neural network cascade type gas identification method of claim 1, wherein the number of digital capsules in the digital capsule layer is the same as the number of gas types in the gas to be detected.
3. The wavelet transform-capsule neural network cascade type gas identification method according to claim 1, wherein the dimension transform in step S3 specifically comprises:
and performing point-separating sampling on the signal sequence to obtain sampling data, and arranging the sampling data of all channels according to the channel sequence to obtain a wavelet transformation coefficient image after dimension transformation as the input of the capsule neural network.
4. The wavelet transform-capsule neural network cascade type gas identification method of claim 1, wherein the input layer comprises 256 convolution kernels of 9*9 with a stride of 1, and the convolution layers comprise 32 convolution layers of 9*9 with a stride of 2.
5. The wavelet transformation-capsule neural network cascading type gas identification device is characterized by comprising a memory and a processor; the memory is used for storing a computer program; the processor is configured to implement the following method when executing the computer program:
s1, inputting mixed gas into a multi-channel sensor array to obtain a gas time domain signal;
s2, five-layer wavelet transformation is carried out on the gas time domain signal to obtain a signal sequence; in five-layer wavelet transformation, each layer of wavelet transformation decomposes an input signal into a signal high-frequency part and a signal low-frequency part, the signal low-frequency part obtained by each layer is used as the input of the wavelet transformation of the next layer, and the signal high-frequency part obtained by the fifth layer and the signal low-frequency part obtained by each layer are finally overlapped to be used as an output signal sequence;
s3, inputting the signal sequence after dimension transformation into a capsule neural network, and combining a dynamic routing algorithm to obtain a gas identification result, wherein the capsule neural network comprises an input layer, a convolution layer, a main capsule layer, a digital capsule layer, a classifier layer and an output layer.
6. The wavelet transform-capsule neural network cascade type gas recognition device according to claim 5, wherein the number of digital capsules in the digital capsule layer is the same as the number of gas types in the gas to be detected.
7. The wavelet transform-capsule neural network cascade type gas recognition device according to claim 5, wherein the dimension transform in step S3 specifically comprises:
and performing point-separating sampling on the signal sequence to obtain sampling data, and arranging the sampling data of all channels according to the channel sequence to obtain a wavelet transformation coefficient image after dimension transformation as the input of the capsule neural network.
8. The wavelet transform-capsule neural network cascade type gas recognition device of claim 5, wherein the input layer comprises 256 convolution kernels of 9*9 having a stride of 1, and the convolution layer comprises 32 convolution layers of 9*9 having a stride of 2.
CN202210157351.8A 2022-02-21 2022-02-21 Wavelet transformation-capsule neural network cascading type gas identification method and device Active CN114527241B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210157351.8A CN114527241B (en) 2022-02-21 2022-02-21 Wavelet transformation-capsule neural network cascading type gas identification method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210157351.8A CN114527241B (en) 2022-02-21 2022-02-21 Wavelet transformation-capsule neural network cascading type gas identification method and device

Publications (2)

Publication Number Publication Date
CN114527241A CN114527241A (en) 2022-05-24
CN114527241B true CN114527241B (en) 2023-08-04

Family

ID=81625391

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210157351.8A Active CN114527241B (en) 2022-02-21 2022-02-21 Wavelet transformation-capsule neural network cascading type gas identification method and device

Country Status (1)

Country Link
CN (1) CN114527241B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6105015A (en) * 1997-02-03 2000-08-15 The United States Of America As Represented By The Secretary Of The Navy Wavelet-based hybrid neurosystem for classifying a signal or an image represented by the signal in a data system
CN111443165A (en) * 2020-03-27 2020-07-24 华中科技大学 Odor identification method based on gas sensor and deep learning
CN113823403A (en) * 2021-08-05 2021-12-21 山东大学 Neural network-based early lung cancer exhaled breath component detection and recognition system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6105015A (en) * 1997-02-03 2000-08-15 The United States Of America As Represented By The Secretary Of The Navy Wavelet-based hybrid neurosystem for classifying a signal or an image represented by the signal in a data system
CN111443165A (en) * 2020-03-27 2020-07-24 华中科技大学 Odor identification method based on gas sensor and deep learning
CN113823403A (en) * 2021-08-05 2021-12-21 山东大学 Neural network-based early lung cancer exhaled breath component detection and recognition system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
气体传感器阵列混合气体检测算法研究;谭光韬;张文文;王磊;;电子测量与仪器学报(07);全文 *

Also Published As

Publication number Publication date
CN114527241A (en) 2022-05-24

Similar Documents

Publication Publication Date Title
Yan et al. Calibration transfer and drift compensation of e-noses via coupled task learning
CN111443165B (en) Odor identification method based on gas sensor and deep learning
KR101852074B1 (en) Electronic Nose System and Method for Gas Classification
CN109657582B (en) Face emotion recognition method and device, computer equipment and storage medium
Hanh et al. Enhancing the performance of transferred efficientnet models in leaf image-based plant disease classification
Zhang et al. A novel background interferences elimination method in electronic nose using pattern recognition
CN113705788B (en) Infrared image temperature estimation method and system based on full convolution neural network
Zhou et al. Early-stage gas identification using convolutional long short-term neural network with sensor array time series data
CN114707598A (en) Mixed gas identification method, system, terminal equipment and readable storage medium
Zhang et al. Sparsity divergence index based on locally linear embedding for hyperspectral anomaly detection
CN115496976B (en) Visual processing method, device, equipment and medium for multi-source heterogeneous data fusion
CN114527241B (en) Wavelet transformation-capsule neural network cascading type gas identification method and device
Wei et al. Artificial intelligence for defect detection in infrared images of solid oxide fuel cells
Eslami et al. Statistical monitoring of image data using multi-channel functional principal component analysis
Yamashita et al. Direct feature extraction from two-dimensional X-ray diffraction images of semiconductor thin films for fabrication analysis
CN115049917B (en) Neural network training method, object detection method and intelligent terminal device
CN115984624A (en) Electronic nose-based gas detection model training method, device, medium and equipment
CN114544868B (en) Gas detection method and system for eliminating influence of interference gas
US10176407B1 (en) Nonlinear sparse representation-based classification for foveated analysis of spectral data with distortions
Rochac et al. Reference clusters based feature extraction approach for mixed spectral signatures with dimensionality disparity
CN115512164A (en) Substance analysis method based on combination of spectral analysis and video classification algorithm
CN112330622A (en) Hyperspectral image band selection method based on ground feature maximum discrimination
Li et al. A novel DenseNet with warm restarts for gas recognition in complex airflow environments
Zhang et al. Robust approach to the MAD change detection method
Cho et al. A Mixture-Gas Edge-Computing Multi-Sensor Device with Generative Learning Framework

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