CN116502117A - ResNet-based hazardous chemical identification method, device and equipment - Google Patents
ResNet-based hazardous chemical identification method, device and equipment Download PDFInfo
- Publication number
- CN116502117A CN116502117A CN202310391739.9A CN202310391739A CN116502117A CN 116502117 A CN116502117 A CN 116502117A CN 202310391739 A CN202310391739 A CN 202310391739A CN 116502117 A CN116502117 A CN 116502117A
- Authority
- CN
- China
- Prior art keywords
- module
- identification
- data
- dangerous
- resnet
- 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
- 238000000034 method Methods 0.000 title claims abstract description 49
- 239000000383 hazardous chemical Substances 0.000 title claims description 18
- 239000000126 substance Substances 0.000 claims abstract description 98
- 238000001228 spectrum Methods 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 29
- 230000003595 spectral effect Effects 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 230000006870 function Effects 0.000 claims description 20
- 230000004913 activation Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 9
- 238000013528 artificial neural network Methods 0.000 description 7
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000001069 Raman spectroscopy Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013529 biological neural network Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/08—Learning methods
-
- 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
Abstract
The invention discloses a dangerous chemical identification method, a device, equipment and a storage medium based on ResNet, which comprises the following steps: collecting spectrum data of dangerous chemicals, preprocessing, and taking the spectrum data obtained after preprocessing as training data; training the training data input based on a model constructed by ResNet to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module; and acquiring spectral data of the dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals. The identification result of dangerous chemicals can be obtained rapidly, and the identification accuracy is improved greatly.
Description
Technical Field
The invention relates to the technical field of detection, in particular to a dangerous chemical identification method, device and equipment based on ResNet.
Background
The classification and identification of dangerous chemicals is an important ring of chemical safety management, and how to accurately and rapidly classify and identify dangerous chemicals is always a difficult problem. The traditional chemical classification and identification method has the problems of insufficient feature extraction, insufficient classifier design, and the like, and the classification accuracy is difficult to guarantee along with the increase of a chemical database, so that the condition of reduced classification and identification accuracy along with the increase of the chemical database can be caused; the algorithm for comparing the similarity of the data in the database after the feature extraction by using the neural network is too time-consuming to perform rapid detection because of the need of traversing the database and calculating the cosine similarity of the data in the database and the input data, thereby resulting in too long calculation time and failing to meet the requirement of rapid detection. These problems may lead to misjudgment or missed judgment in the identification and management of dangerous chemicals in terms of speed, accuracy and safety.
Disclosure of Invention
In view of the above, the invention aims to provide a ResNet-based dangerous chemical identification method, a ResNet-based dangerous chemical identification device and ResNet-based dangerous chemical identification equipment, which aim to solve the problems that the existing dangerous chemical identification is low in identification precision, low in speed and the like.
To achieve the above object, the present invention provides a method for identifying dangerous chemicals based on ResNet, the method comprising:
collecting spectrum data of dangerous chemicals, preprocessing, and taking the spectrum data obtained after preprocessing as training data;
training the training data input based on a model constructed by ResNet to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module;
and acquiring spectral data of the dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals.
Preferably, the pre Net module performs feature extraction and downsampling operation on the input spectrum data; the PreNet module comprises a ConvReLU layer, wherein the ConvReLU layer comprises a convolution layer and an activation function layer.
Preferably, residual error learning is performed on the output data of the PreNet module through the ResBlock module; wherein, the ResBlock module includes a ConvBN layer including a convolutional layer and a BN layer.
Preferably, the output data of the ResBlock module is subjected to pooling and batch standardization operation through the MaxPool module; the MaxPool module comprises a convolution block and a Pool layer, wherein the convolution block comprises a convolution layer, a BN layer and an activation function layer.
Preferably, the FC_Net module is used for flattening the characteristic output data extracted by the main convolution network and converting the characteristic output data into vectors with the length equal to the number of dangerous chemicals; the backbone convolution network comprises the ResBlock module and the MaxPool module.
Preferably, the collecting and preprocessing of the spectrum data of the hazardous chemical includes:
and carrying out baseline removal and denoising treatment on the spectrum data, and interpolating the number of data points of each spectrum data into 2048 through a linear interpolation algorithm.
Preferably, before the collecting and preprocessing of the spectrum data of the hazardous chemical, the method further comprises:
the collected spectrum data of the dangerous chemicals are stored according to a preset mode to construct a database, and a substance index table is created according to the unique index number allocated to the name of each dangerous chemical.
To achieve the above object, the present invention also provides a dangerous chemical identification device based on ResNet, the device comprising:
the pretreatment unit is used for collecting spectrum data of dangerous chemicals and carrying out pretreatment, and the spectrum data obtained after the pretreatment is used as training data;
the model training unit is used for inputting the training data into a model constructed based on ResNet to train so as to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module;
the identification unit is used for acquiring spectral data of dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals.
In order to achieve the above object, the present invention also proposes a dangerous chemical identification device based on ResNet, comprising a processor, a memory and a computer program stored in said memory, said computer program being executed by said processor to implement the steps of a dangerous chemical identification method based on ResNet as described in the above embodiments.
To achieve the above object, the present invention also proposes a computer-readable storage medium having stored thereon a computer program to be executed by a processor to implement the steps of a method for identifying hazardous chemicals based on res net according to the above embodiments.
The beneficial effects are that:
according to the scheme, after the pretreatment of the optical data, the identification model is obtained after the model constructed based on the ResNet is input and trained, the ResNet network is a deep convolutional neural network, and the direct communication channels are added in the network, so that gradient disappearance and gradient explosion can be effectively solved, the performance of the model can be improved, then the dangerous chemicals to be identified can be quickly identified by classifying and identifying the dangerous chemicals through the obtained identification model, and the identification accuracy is greatly improved.
According to the scheme, the pre-Net is introduced into the network to pre-process the input spectrum data, so that better characteristics can be extracted; and the output of the convolution layer is normalized through MaxPool, so that overfitting is avoided, and the calculation complexity of the model is reduced.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a dangerous chemical identification method based on res net according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of an overall network framework of an identification model according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of a network structure of a pre net module according to an embodiment of the present invention.
Fig. 4 is a schematic network structure diagram of a ResBlock module according to an embodiment of the present invention.
Fig. 5 is a network structure schematic diagram of a MaxPool module according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of a network structure of an fcnet module according to an embodiment of the present invention.
Fig. 7 is a schematic structural diagram of a dangerous chemical identification device based on res net according to an embodiment of the present invention.
The realization of the object, the functional characteristics and the advantages of the invention will be further described with reference to the accompanying drawings in connection with the embodiments.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention.
The following describes the invention in detail with reference to examples.
The algorithms traditionally used for chemical classification and identification mainly include the following:
principal Component Analysis (PCA): principal component analysis is a commonly used method of data dimension reduction that can convert high-dimensional data into a low-dimensional space to reduce data dimensions and computational complexity. In chemical classification identification, PCA can dimension down process the raw spectral data to improve classification accuracy. But has the disadvantage of poor performance in processing non-linear data, and important information may be ignored, resulting in poor classification.
Support Vector Machine (SVM): the support vector machine is a commonly used classification algorithm that can achieve classification of different chemicals by constructing an optimal separation hyperplane. In chemical classification and identification, the SVM can perform feature extraction and classification modeling on the spectrum data so as to realize high-precision classification and identification. But has the disadvantages of higher computational complexity in processing large-scale data and sensitivity to data noise, requiring efficient data processing and optimization.
Artificial Neural Network (ANN): an artificial neural network is a computational model that mimics a biological neural network and that allows for the classification of different chemicals through learning and training. In chemical classification, ANN can perform feature extraction and classification modeling on spectral data to achieve high-precision classification. The method has the defects that the probability value of the input data cannot be directly output after the characteristics are extracted, and only the cosine similarity of the input data and the data in the database can be calculated to be used as the probability value, so that the calculated amount is large, and time is wasted relatively.
Based on the method, the ResNet-based dangerous chemical identification method provided by the invention has higher accuracy and robustness in classification identification of thousands of dangerous chemicals, and the identification accuracy can reach more than 95%.
Referring to fig. 1, a schematic flow chart of a dangerous chemical identification method based on res net according to an embodiment of the present invention is shown.
In this embodiment, the method includes:
s11, collecting spectrum data of dangerous chemicals, preprocessing, and taking the spectrum data obtained after preprocessing as training data.
Wherein, the collection of spectral data of dangerous chemicals and pretreatment comprises:
and carrying out baseline removal and denoising treatment on the spectrum data, and interpolating the number of data points of each spectrum data into 2048 through a linear interpolation algorithm.
Further, before the collecting and preprocessing the spectrum data of the dangerous chemical, the method further comprises:
the collected spectrum data of the dangerous chemicals are stored according to a preset mode to construct a database, and a substance index table is created according to the unique index number allocated to the name of each dangerous chemical.
In this embodiment, the spectrum data of the hazardous chemical collected by the instrument (e.g., spectrometer) is stored as a csv file type, where the first column in the csv file indicates raman shift, the second column indicates intensity, and both columns of data have 1501 data points. The csv file is named in different modes, for example, the naming method of the csv file is named as a number, different dangerous chemical substances have different numbers, the number is named as RM20220812-012_1.csv, wherein RM20220812-012 is named as a number, and 1_1 represents 1 st data collected by the dangerous chemical substances, and 20 or more data are preferably collected by each dangerous chemical substance, so that the model can learn sufficient characteristics to obtain better results. The purpose of constructing the database is to store data and create an index table, the process includes placing csv files of the same dangerous chemical substance under the corresponding folders, and naming the folders in a preset manner, for example, the name of the folders starts from the number 1, and then sequentially adding one, and the name of the folders represents the id of the dangerous chemical substance. In the subsequent model training phase, the labels of the corresponding hazardous chemical substances are represented by the names of the folders. After the library is built, a dangerous chemical substance id list is further built, wherein the dangerous chemical substance id list consists of a folder name list, corresponding substance numbers and a substance name list, and the dangerous chemical substance id list is used for providing the numbers (index numbers) of dangerous chemical substances for the output of a follow-up model so as to find the corresponding dangerous chemical substance names.
S12, inputting the training data into a model constructed based on ResNet to train so as to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module.
As shown in fig. 2, the network structure of the recognition model generally includes pre Net (preprocessing module), resBlock (residual block), maxPool (pooling block), and fc_net (full connection layer). In this embodiment, the network architecture designs a ResBlock module as a residual block, improves the convolutional network, and introduces some new modules and components including pre net and MaxPool, where the pre net module can be used to pre-process the input spectral data so as to extract better features; the MaxPool module may be used to normalize the output of the convolutional layer to avoid overfitting. When training, using nn. Cross EntopyLoss () as a loss function, the loss function is suitable for multi-classification problem, and can process unbalanced classification data, and can compare probability vector output by model with real label, calculate cross entropy loss between model predicted value and real value, so as to implement model optimization.
Refer to the schematic diagram of the PreNet structure shown in FIG. 3. In the pre-net (preprocessing module), the pre-net module functions to perform feature extraction and downsampling operations on the input preprocessed spectrum data. In the figure, input represents Input, output represents Output, input of PreNet is a 1D signal of 1 channel, and the shape size is (BatchSize, 1, input_length), wherein BatchSize represents that Batch of data of Batch is Input, 1 represents channel number, input_length represents the length of the Input data, and input_length is equal to 2048. The combination of the convolution layer (Convolutional layer) and the activation function layer (ReLU layer) may be referred to simply as ConvReLU layer for extracting features and implementing nonlinear mapping. After three ConvReLU layers, a 16-channel 1D feature map is output, the shape is (BatchSize, 16,1024), the number of output channels is increased, and the length is halved. Shortcut is a jump connection, and by using the idea of a residual network, the neural network can be prevented from degradation.
Reference is made to the schematic diagram of the ResBlock architecture shown in FIG. 4. The role of ResBlock (residual block) is to improve the performance of the network through residual learning. Each residual block receives a data tensor in the shape of (BatchSize, input_channel, input_length) and outputs a data tensor in the shape of (BatchSize, output_channel, input_length), and the learning capability of the network is enhanced by using convolution kernels of different sizes and designing the input_channel and the output_channel to different values. Increasing the convolution kernel can make the receptive field of the model larger, the better the obtained global feature, but the increase of the calculated amount is brought, and increasing the output_channel can make the network learn more parameters, but the network size is increased, so that the network response time is slowed down. The module BN (Batch Normalization) layer is mainly used for normalizing the characteristics of each Batch (a Batch of training data output after passing through the previous convolution layer) so that the network model is more stable and rapid in the training process. The modules Convolutional layer (convolution layer) and Batch Normalization (BN layer) are combined together to be called ConvBN layer for short, which can improve the performance of the network, accelerate training and enhance the robustness and reliability of the network.
Reference is made to the MaxPool structure schematic shown in fig. 5. Each MaxPool module receives a data tensor in the shape of (batch size, input_channel, input_length) and outputs a data tensor in the shape of (batch size,2 x input_channel, input_length/2). The number of output channels is multiplied by 2, but the output data length becomes half of the original. The MaxPool has the function of carrying out pooling and batch standardization operation on the output of the ResBlock module, and reducing the size of the feature map, thereby reducing the computational complexity of the model. The combination of the module convolution layer (Convolutional layer), the BN layer and the activation function layer (ReLU layer) can be simply called a convolution block (Convolutional block), and the main functions are to increase nonlinearity in the deep learning neural network, speed up training and improve classification accuracy. The module Pool (Pooling Layer) layer is used for reducing the space size of the feature map and reducing the parameter number of the model, so that the condition of over fitting is relieved.
Referring to the fc_net structure diagram shown in fig. 6. The function of fc_net (fully connected layer) is to implement fully connected layer operation, in order to convert the characteristic Output (Input) extracted by the backbone convolution network composed of a plurality of structures ResBlock (residual block) and MaxPool (pooled block) after flattening into a vector (Output) with a length equal to the number of chemical types in the database, so that the probability value corresponding to each chemical can be Output. Then, the vector (Output) is input into an activation function sigmoid, each value in the vector Output can be mapped between 0 and 1, the probability of identifying dangerous chemical substances can be obtained only by obtaining the maximum value in the Output and the position index thereof, and the identified corresponding dangerous chemical substances can be found from the id table according to the index.
In another embodiment, the robustness and generalization capability of the network model can be increased by adding some additional network layers, such as adding a Pooling layer (Pooling layer), adding a Dropout layer (random inactivation layer), etc. in the convolution block. In addition, in other embodiments, the training speed can be improved by reducing some network layers, such as removing the Batchnormalization layer, removing the activation function layer, and the like, so as to simplify the network structure.
S13, acquiring spectral data of dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals.
In this embodiment, after the spectrum data sequence collected by the spectrometer is preprocessed (including baseline removal and noise removal, and data points are interpolated into 2048 data points by a linear interpolation algorithm), the spectrum data is input into the recognition model by loading the recognition model stored in the format of a. Pt file, so that the recognition model can output an index value and a probability value, and the corresponding recognized dangerous chemical substances can be found in the id table by the index value. Based on the method, the identification result of the dangerous chemicals can be obtained rapidly, meanwhile, the precision is improved greatly compared with the traditional method, the identification rate of thousands of dangerous chemicals can reach more than 95%, and the problems of low precision, low speed and the like can be effectively solved.
Referring to fig. 7, a schematic structural diagram of a dangerous chemical identification device based on res net according to an embodiment of the present invention is shown.
In this embodiment, the apparatus 70 includes:
a preprocessing unit 71, configured to collect and preprocess spectral data of dangerous chemicals, and take the spectral data obtained after preprocessing as training data;
the model training unit 72 is configured to input the training data into a model constructed based on res Net to perform training, so as to obtain an identification model, where a network structure of the identification model includes a pre Net module, a res block module, a MaxPool module, and an fc_net module;
the identifying unit 73 is configured to acquire spectral data of the dangerous chemical to be identified, input the spectral data into the identifying model, and identify the dangerous chemical, so as to obtain an identifying result including a probability value and an index number of the corresponding dangerous chemical.
The respective unit modules of the apparatus 70 may perform the corresponding steps in the above method embodiments, so that the detailed description of the respective unit modules is omitted herein.
The embodiment of the invention also provides a device, which comprises the dangerous chemical identification device based on ResNet, wherein the dangerous chemical identification device based on ResNet can adopt the structure of the embodiment of fig. 7, correspondingly, the technical scheme of the method embodiment shown in fig. 1 can be executed, the implementation principle and the technical effect are similar, and detailed description can be referred to relevant records in the embodiment and is not repeated here.
The apparatus comprises: a device with a photographing function such as a mobile phone, a digital camera or a tablet computer, or a device with an image processing function, or a device with an image display function. The device may include a memory, a processor, an input unit, a display unit, a power source, and the like.
The memory may be used to store software programs and modules, and the processor executes the software programs and modules stored in the memory to perform various functional applications and data processing. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program (e.g., an image playing function, etc.) required for at least one function, and the like; the storage data area may store data created according to the use of the device, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory may also include a memory controller to provide access to the memory by the processor and the input unit.
The input unit may be used to receive input digital or character or image information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control. Specifically, the input unit of the present embodiment may include a touch-sensitive surface (e.g., a touch display screen) and other input devices in addition to the camera.
The display unit may be used to display information entered by a user or provided to a user as well as various graphical user interfaces of the device, which may be composed of graphics, text, icons, video and any combination thereof. The display unit may include a display panel, and alternatively, the display panel may be configured in the form of an LCD (Liquid Crystal Display ), an OLED (Organic Light-Emitting Diode), or the like. Further, the touch-sensitive surface may overlay the display panel, and upon detection of a touch operation thereon or thereabout, the touch-sensitive surface is communicated to the processor to determine the type of touch event, and the processor then provides a corresponding visual output on the display panel based on the type of touch event.
The embodiment of the present invention also provides a computer readable storage medium, which may be a computer readable storage medium contained in the memory in the above embodiment; or may be a computer-readable storage medium, alone, that is not assembled into a device. The computer readable storage medium has stored therein at least one instruction that is loaded and executed by a processor to implement the ResNet-based hazardous chemical identification method shown in FIG. 1. The computer readable storage medium may be a read-only memory, a magnetic disk or optical disk, etc.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described as different from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other. For the device embodiments, the apparatus embodiments and the storage medium embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and reference is made to the description of the method embodiments for relevant points.
Also, herein, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
While the foregoing description illustrates and describes the preferred embodiments of the present invention, it is to be understood that the invention is not limited to the forms disclosed herein, but is not to be construed as limited to other embodiments, but is capable of use in various other combinations, modifications and environments and is capable of changes or modifications within the scope of the inventive concept, either as described above or as a matter of skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.
Claims (10)
1. A method for identifying hazardous chemicals based on res net, the method comprising:
collecting spectrum data of dangerous chemicals, preprocessing, and taking the spectrum data obtained after preprocessing as training data;
training the training data input based on a model constructed by ResNet to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module;
and acquiring spectral data of the dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals.
2. The method for identifying dangerous chemicals based on ResNet according to claim 1, wherein the PreNet module is used for carrying out feature extraction and downsampling operation on the input spectrum data; the PreNet module comprises a ConvReLU layer, wherein the ConvReLU layer comprises a convolution layer and an activation function layer.
3. The method for identifying dangerous chemicals based on ResNet according to claim 1, wherein residual learning is performed on output data of the PreNet module through the ResBlock module; wherein, the ResBlock module includes a ConvBN layer including a convolutional layer and a BN layer.
4. The ResNet-based hazardous chemical identification method according to claim 1, wherein output data of the ResBlock module is subjected to pooling and batch standardization operation through the MaxPool module; the MaxPool module comprises a convolution block and a Pool layer, wherein the convolution block comprises a convolution layer, a BN layer and an activation function layer.
5. The ResNet-based hazardous chemical identification method according to claim 1, wherein characteristic output data extracted from a backbone convolutional network are flattened through the FC_Net module and then converted into vectors with the length equal to the number of hazardous chemicals; the backbone convolution network comprises the ResBlock module and the MaxPool module.
6. The method for identifying dangerous chemicals based on ResNet according to claim 1, wherein the steps of collecting and preprocessing spectrum data of dangerous chemicals comprise:
and carrying out baseline removal and denoising treatment on the spectrum data, and interpolating the number of data points of each spectrum data into 2048 through a linear interpolation algorithm.
7. The method for identifying hazardous chemicals based on ResNet according to claim 1, further comprising, before said collecting and preprocessing the spectral data of the hazardous chemicals:
the collected spectrum data of the dangerous chemicals are stored according to a preset mode to construct a database, and a substance index table is created according to the unique index number allocated to the name of each dangerous chemical.
8. A res net based hazardous chemical identification device, the device comprising:
the pretreatment unit is used for collecting spectrum data of dangerous chemicals and carrying out pretreatment, and the spectrum data obtained after the pretreatment is used as training data;
the model training unit is used for inputting the training data into a model constructed based on ResNet to train so as to obtain an identification model, wherein the network structure of the identification model comprises a PreNet module, a ResBlock module, a MaxPool module and an FC_Net module;
the identification unit is used for acquiring spectral data of dangerous chemicals to be identified, inputting the spectral data into the identification model for dangerous chemical identification, and obtaining an identification result containing probability values and index numbers of the corresponding dangerous chemicals.
9. A res net based hazardous chemical identification device comprising a processor, a memory and a computer program stored in said memory, said computer program being executed by said processor to perform the steps of a res net based hazardous chemical identification method as claimed in any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program for execution by a processor to perform the steps of a res net based hazardous chemical identification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310391739.9A CN116502117B (en) | 2023-04-13 | 2023-04-13 | ResNet-based hazardous chemical identification method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310391739.9A CN116502117B (en) | 2023-04-13 | 2023-04-13 | ResNet-based hazardous chemical identification method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116502117A true CN116502117A (en) | 2023-07-28 |
| CN116502117B CN116502117B (en) | 2023-12-15 |
Family
ID=87329552
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310391739.9A Active CN116502117B (en) | 2023-04-13 | 2023-04-13 | ResNet-based hazardous chemical identification method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116502117B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110068544A (en) * | 2019-05-08 | 2019-07-30 | 广东工业大学 | Material identification network model training method and tera-hertz spectra substance identification |
| CN111881987A (en) * | 2020-07-31 | 2020-11-03 | 西安工业大学 | Apple virus identification method based on deep learning |
| CN112767443A (en) * | 2021-01-18 | 2021-05-07 | 深圳市华尊科技股份有限公司 | Target tracking method, electronic equipment and related product |
| US20210390338A1 (en) * | 2020-06-15 | 2021-12-16 | Dalian University Of Technology | Deep network lung texture recogniton method combined with multi-scale attention |
| US20220101276A1 (en) * | 2020-09-30 | 2022-03-31 | X Development Llc | Techniques for predicting the spectra of materials using molecular metadata |
| US20220198326A1 (en) * | 2020-12-17 | 2022-06-23 | Virtual Control Limited | Spectral data processing for chemical analysis |
| CN115078327A (en) * | 2022-06-14 | 2022-09-20 | 北京信息科技大学 | Rapid detection method for dangerous chemicals based on neural network architecture search |
| US11483330B1 (en) * | 2021-09-18 | 2022-10-25 | University Of Tabuk | Systems and methods for detecting cyberattacks targeting internet of things environments |
| CN115290596A (en) * | 2022-08-03 | 2022-11-04 | 广东工业大学 | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment |
| CN115880654A (en) * | 2022-11-07 | 2023-03-31 | 武汉路特斯科技有限公司 | Vehicle lane change risk assessment method and device, computer equipment and storage medium |
| CN115906624A (en) * | 2022-11-11 | 2023-04-04 | 湖南航天远望科技有限公司 | Hazardous chemical substance gas spectrum generation method, terminal equipment and storage medium |
-
2023
- 2023-04-13 CN CN202310391739.9A patent/CN116502117B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110068544A (en) * | 2019-05-08 | 2019-07-30 | 广东工业大学 | Material identification network model training method and tera-hertz spectra substance identification |
| US20210390338A1 (en) * | 2020-06-15 | 2021-12-16 | Dalian University Of Technology | Deep network lung texture recogniton method combined with multi-scale attention |
| CN111881987A (en) * | 2020-07-31 | 2020-11-03 | 西安工业大学 | Apple virus identification method based on deep learning |
| US20220101276A1 (en) * | 2020-09-30 | 2022-03-31 | X Development Llc | Techniques for predicting the spectra of materials using molecular metadata |
| US20220198326A1 (en) * | 2020-12-17 | 2022-06-23 | Virtual Control Limited | Spectral data processing for chemical analysis |
| CN112767443A (en) * | 2021-01-18 | 2021-05-07 | 深圳市华尊科技股份有限公司 | Target tracking method, electronic equipment and related product |
| US11483330B1 (en) * | 2021-09-18 | 2022-10-25 | University Of Tabuk | Systems and methods for detecting cyberattacks targeting internet of things environments |
| CN115078327A (en) * | 2022-06-14 | 2022-09-20 | 北京信息科技大学 | Rapid detection method for dangerous chemicals based on neural network architecture search |
| CN115290596A (en) * | 2022-08-03 | 2022-11-04 | 广东工业大学 | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment |
| CN115880654A (en) * | 2022-11-07 | 2023-03-31 | 武汉路特斯科技有限公司 | Vehicle lane change risk assessment method and device, computer equipment and storage medium |
| CN115906624A (en) * | 2022-11-11 | 2023-04-04 | 湖南航天远望科技有限公司 | Hazardous chemical substance gas spectrum generation method, terminal equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116502117B (en) | 2023-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11776244B2 (en) | Systems and methods for generating and using semantic images in deep learning for classification and data extraction | |
| CN112434721B (en) | Image classification method, system, storage medium and terminal based on small sample learning | |
| CN110689036B (en) | Method and system for automatic chromosome classification | |
| Goodfellow et al. | Multi-digit number recognition from street view imagery using deep convolutional neural networks | |
| RU2661750C1 (en) | Symbols recognition with the use of artificial intelligence | |
| CA3088686C (en) | Automated document extraction and classification | |
| US10867169B2 (en) | Character recognition using hierarchical classification | |
| Desai et al. | Hybrid approach for content-based image retrieval using VGG16 layered architecture and SVM: an application of deep learning | |
| Toledo et al. | Handwritten word image categorization with convolutional neural networks and spatial pyramid pooling | |
| CN117391466A (en) | Novel early warning method and system for contradictory dispute cases | |
| CN116502117B (en) | ResNet-based hazardous chemical identification method, device and equipment | |
| CN111797396A (en) | Malicious code visualization and variety detection method, device, equipment and storage medium | |
| Saeed et al. | An application of deep learning in character recognition: an overview | |
| CN115937887A (en) | Method and device for extracting document structured information, electronic equipment and storage medium | |
| Dey et al. | A time efficient offline handwritten character recognition using convolutional extreme learning machine | |
| CN112801960A (en) | Image processing method and device, storage medium and electronic equipment | |
| CN111768214A (en) | Product attribute prediction method, system, device and storage medium | |
| CN116503674B (en) | Small sample image classification method, device and medium based on semantic guidance | |
| Kaur et al. | An Efficient Predictive Model for High Dimensional Data | |
| Rayeed et al. | BdSL47: A complete depth-based Bangla sign alphabet and digit dataset | |
| da Silva Amarante | Extracting medical information from personal child health records | |
| Sunitha et al. | NOVEL CBIR System Using Spark MAP-Reduce with a Firefly Macqueen’s K-Means Clustering Algorithm | |
| Riese | Classifying hand-drawn documents in mobile settings, using transfer learning and model compression | |
| Debard | Automatic learning of next generation human-computer interactions | |
| Srinivasa Kumar et al. | Image Retrieval System Based on Perceptual Browsing Component Using Interactive Genetic Algorithm |
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 |