CN115290596A - FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment - Google Patents
FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment Download PDFInfo
- Publication number
- CN115290596A CN115290596A CN202210928339.2A CN202210928339A CN115290596A CN 115290596 A CN115290596 A CN 115290596A CN 202210928339 A CN202210928339 A CN 202210928339A CN 115290596 A CN115290596 A CN 115290596A
- Authority
- CN
- China
- Prior art keywords
- sample
- data
- training
- fcn
- acgan
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 101
- 238000013145 classification model Methods 0.000 claims abstract description 57
- 238000004088 simulation Methods 0.000 claims abstract description 48
- 238000001228 spectrum Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 231100001261 hazardous Toxicity 0.000 claims description 18
- 230000003595 spectral effect Effects 0.000 claims description 16
- 238000001328 terahertz time-domain spectroscopy Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 7
- 230000008569 process Effects 0.000 description 16
- 238000013135 deep learning Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 239000013589 supplement Substances 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000013256 Gubra-Amylin NASH model Methods 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008033 biological extinction Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3581—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation
- G01N21/3586—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation by Terahertz time domain spectroscopy [THz-TDS]
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Toxicology (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a hidden dangerous goods identification method and equipment based on FCN-ACGAN data enhancement, relating to the technical field of goods detection and comprising the following steps: the method comprises the steps of preprocessing real terahertz spectrum data of an article acquired in advance to obtain a real sample, generating a simulation sample by using a pre-trained FCN-ACGAN network model, training a pre-constructed initial ResNet-LSTM classification model according to the real sample and the simulation sample to obtain an optimal classification model, classifying the terahertz time-domain spectrum data by using the optimal classification model, and determining the attribute and the type of the detected article according to a classification result.
Description
Technical Field
The invention relates to the technical field of article identification, in particular to a method and equipment for identifying hidden dangerous articles of articles by analyzing terahertz time-domain spectroscopy, and specifically relates to a method and equipment for identifying hidden dangerous articles based on FCN-ACGAN data enhancement.
Background
Terahertz waves are widely used in the field of material detection due to the fact that the wave bands of terahertz waves have fingerprint spectrum properties, nondestructive detection can be achieved by identifying various dangerous goods through terahertz waves, with the rapid development of artificial intelligence, terahertz time-domain spectral data are trained through deep learning, and nondestructive detection through a model obtained through training is increasingly common.
The method is mainly characterized in that an original spectrum data is expanded through a countermeasure network (GAN) with obvious effect in a time sequence data enhancement method of deep learning, a new generation artificial intelligence small sample data enhancement method based on WGAN (Wassertein GAN) firstly divides the original sample into a training set and a testing set sample, generates simulation sample data after the GAN is trained by the training set sample, and amplifies the scale of the training set sample; then, training by using a simulation sample to obtain a classifier; and finally, testing the classification effect of the classifier by using the test set samples.
However, the terahertz time-domain spectral data enhancement by using the generated countermeasure network (GAN) has certain limitations, the conventional GAN model is difficult to model time series data, and has the problems of unstable training, gradient disappearance, excessive free mode and easy collapse, so that finally generated data feature information is lost and the representativeness is poor, further, the classifier classification effect obtained by using the data with the feature information loss is difficult to meet the detection requirement, and the identification precision of article identification according to the classification result is low.
Disclosure of Invention
The invention provides a hidden dangerous goods identification method and equipment based on FCN-ACGAN data enhancement, which are used for solving the technical problems that in the existing data classification method, due to the fact that training data are insufficient, the classification effect of a classification model obtained through training is poor, and further the object identification precision is low.
The invention provides a terahertz time-domain spectrum concealed hazardous article identification method based on FCN-ACGAN data enhancement, which comprises the following steps:
preprocessing pre-acquired real spectral data to obtain a real sample;
generating a simulation sample by utilizing a pre-trained FCN-ACGAN network model; the FCN-ACGAN network model comprises a generator and a discriminator, wherein the generator and the discriminator respectively comprise a plurality of full connection layers; the steps of training the FCN-ACGAN network model are as follows:
s1, pre-training the discriminator by using the real sample to obtain a primary discriminator;
s2, generating a primary simulation sample by using the generator;
s3, mixing the primary simulation sample with the real sample to obtain a training sample;
s4, actually training the primary arbiter and the generator by using the training samples, respectively updating the network parameters of the primary arbiter and the network parameters of the generator based on a RMSProp optimizer, judging whether the FCN-ACGAN network model reaches Nash equilibrium, if so, stopping training to obtain a trained FCN-ACGAN network model, and if not, returning to the S2;
training a pre-constructed initial ResNet-LSTM classification model according to the real sample and the simulation sample to obtain an optimal classification model;
and classifying the terahertz time-domain spectral data by using the optimal classification model, and determining hidden dangerous goods according to a classification result.
Preferably, the training of the pre-constructed initial ResNet-LSTM classification model according to the real sample and the simulated sample to obtain the optimal classification model specifically comprises:
mixing the real sample and the simulation sample to obtain an extended sample, and dividing the extended sample into a pre-training sample and an actual training sample;
constructing an initial ResNet-LSTM classification model, pre-training the initial ResNet-LSTM classification model by using the pre-training sample to obtain a super-parameter of the initial ResNet-LSTM classification model, actually training the initial ResNet-LSTM classification model based on the super-parameter and the actual training sample, and finishing actual training when a model error of the initial ResNet-LSTM classification model meets a preset error threshold to obtain an optimal classification model.
Preferably, the generating a primary simulation sample by using the generator specifically includes:
establishing a mapping relationship between the real data and random noise of a potential space, and generating a simulation sample based on the mapping relationship, wherein the simulation sample contains a class label.
Preferably, the updating the network parameters of the primary discriminator and the generator based on the RMSProp optimizer is specifically that:
keeping the network parameters of the generator unchanged, acquiring a network loss value of the primary discriminator, updating the primary discriminator based on the RMSProp optimizer and the network loss value of the primary discriminator, keeping the network parameters of the primary discriminator unchanged when the updating times of the primary discriminator meet a preset first updating threshold value, acquiring the network loss value of the generator, and updating the generator based on the RMSProp optimizer and the network loss value of the generator until the updating times of the generator meet a preset second updating threshold value.
Preferably, the generator comprises: the device comprises an input module, a Dense full-connection layer, a Tanh activation function layer and an output module.
Preferably, the discriminator includes: the system comprises an input module, a Dense full connection layer, a ReLU activation function layer and a softmax classifier.
Preferably, before step S2, the method further comprises: and pre-training the initial generator by using the real sample to obtain a trained generator.
Preferably, the preprocessing the real spectrum data acquired in advance to obtain the real sample specifically comprises:
performing data cleaning on pre-acquired real spectral data to obtain a first initial sample;
supplementing the missing value of the first sample to obtain a second initial sample;
and carrying out normalization processing on the second initial sample to obtain a real sample.
Preferably, the classification result includes a first classification result and a second classification result, and the determining of the concealed dangerous goods according to the classification result specifically includes:
and judging the type of the classification result, judging that the detected object is a non-concealed dangerous article when the classification result is a first classification result, judging that the detected object is a concealed dangerous article when the classification result is a second classification result, comparing the second classification result with a pre-established spectral feature database of the concealed dangerous article, and judging the type of the concealed dangerous article according to the comparison result.
The present invention also provides an electronic device, which is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data classification method.
According to the technical scheme, the invention has the following advantages:
the invention provides a hidden dangerous article identification method and equipment based on FCN-ACGAN data enhancement. The full connection layer can help the FCN-ACGAN model to learn dynamic characteristics among real data and improve quality of data generated by the generator, and the RMSProp optimizer is used for updating network parameters of the discriminator and the generator, so that jitter caused by gradient difference in the updating process can be effectively eliminated, and the convergence process of the FCN-ACGAN model is accelerated. Further, a generator in the trained FCN-ACGAN network model is used for creating a simulation sample, the simulation sample and the real sample are mixed to train the initial ResNet-LSTM classification model, an optimal classification model meeting data classification standards is obtained, the optimal classification model is used for classifying the terahertz time-domain spectral data of the detected object, and finally the attribute and the type of the detected object are identified according to the classification result. According to the data classification method provided by the invention, the FCN-ACGAN model is adopted to create the simulation sample, so that the extension of the training sample is realized, and the technical problems that the classification effect of the classification model obtained by training is poor and the article identification precision is further low due to insufficient training data in the existing data classification method are solved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of a method for concealed hazardous article identification based on FCN-ACGAN data enhancement according to an embodiment of the present invention;
FIG. 2 is a network structure diagram of the FCN-ACGAN model provided in the embodiment of the present invention;
FIG. 3 is a network architecture diagram of a generator provided by an embodiment of the present invention;
fig. 4 is a network structure diagram of an arbiter according to an embodiment of the present invention.
Detailed Description
The embodiment of the invention provides a hidden dangerous goods identification method and equipment based on FCN-ACGAN data enhancement, which are used for training and updating a primary discriminator and a generator comprising a plurality of full connection layers to obtain an FCN-ACGAN network model reaching Nash equilibrium, establishing a simulation sample by using the generator in the FCN-ACGAN network model to realize the extension of a training sample, and further training an initial ResNet-LSTM classification model by using the extension sample to obtain an optimal classification model meeting data classification standards.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in 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 obvious that the embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the deep learning network model, general features of a data set are generally extracted, the features are used as characteristics for predicting a certain result, and the constructed model is trained by using the data set, so that the result output by the model is as close to the characteristics as possible. When the training data is less, the model obtained by the training mode is limited: when the training set data is used for result prediction on the model, the prediction result is better, and when the test set or verification set data is used for result prediction on the model, the error rate of the model is higher and the generalization capability is low. In order to avoid the above problems, the deep learning algorithm usually needs to acquire a large amount of training data, and the overfitting phenomenon of the model is avoided by increasing the amount of the training data.
The terahertz time-domain spectroscopy is a technology that terahertz pulses are reflected or transmitted on the surface of a sample to respectively measure a reference signal and a measurement signal before and after passing through the sample, then acquired time-domain signals are converted into frequency domains through Fast Fourier Transform (FFT) to obtain two frequency-domain spectrums, finally, optical parameters such as refractive index, extinction coefficient and absorption coefficient of the sample to be measured can be extracted through processing frequency-domain data, and article detection (article classification) can be achieved through analyzing related data. In the prior art, a deep learning network model is generally established by taking terahertz time-domain spectral data as a training sample, then a model is utilized to detect articles, however, in the field of detecting dangerous articles concealed by terahertz time-domain spectral data, manual marking is generally adopted for obtaining terahertz time-domain spectral data of different dangerous articles, a large amount of labor and time cost is often consumed, the obtained data size cannot cover an actual article detection scene, the recognition accuracy and universality of a deep learning classification model are low due to insufficient training data, and the recognition rate of article attributes according to a classification result is further low.
In view of the above, the present application provides a concealed hazardous article identification method based on FCN-ACGAN data enhancement, please refer to fig. 1, the method includes:
100. and preprocessing the pre-acquired real spectrum data to obtain a real sample.
It can be understood that noise inevitably exists in the process of spectrum data acquisition, and the original terahertz time-domain spectrum data (hereinafter referred to as data, which is not described any more) of the detected object is preprocessed, so that the data expression capability can be improved, and the data characteristics are more obvious.
200. And generating a simulation sample by utilizing a pre-trained FCN-ACGAN network model.
The constructed model is trained by sample data, so that the precision of the model is as high as possible, which is an ideal result of model training. However, when the training data is less, the model obtained by the training method is often found in the case where the result prediction is performed on the model by using the training set data, the prediction result is better, and when the result prediction is performed on the model by using the test set or the verification set data, the error rate of the model is higher. In order to avoid the above problems, the establishment of the model usually needs to acquire a larger amount of training data, and the overfitting phenomenon of the model is avoided by increasing the amount of the training data. When the amount of original data is small, the amount of data can be expanded by creating samples.
It can be understood that when a sufficient amount of training data is available, a model with higher precision can be trained, but if the quality of the training data is poor and the training data is insufficient, the model obtained by training is usually poor in precision or cannot meet the use requirement, so that the sufficient amount of training data is ensured, and meanwhile, the sufficient amount of training data has better quality.
When the data volume is small, the total amount of the added samples can be obtained through data enhancement, but if the total amount of the samples is obtained through the data enhancement preferentially and then the total data is processed to improve the data quality, the processing workload can be increased, so that the embodiment preferentially preprocesses the original data and then generates the simulation samples by using a pre-trained FCN-ACGAN network model based on the original data to realize the expansion of the number of samples.
300. And training a pre-constructed initial ResNet-LSTM classification model according to the real sample and the generated sample to obtain an optimal classification model.
According to the step 200, the number of simulation samples meeting the training requirement can be obtained, the simulation samples and the original samples are mixed to realize training sample extension, the extended samples are used for training the ResNet-LSTM classification model, and when the model error meets a preset error threshold value, the training is stopped, so that the optimal classification model is obtained.
400. And classifying the terahertz time-domain spectral data by using the optimal classification model, and determining hidden dangerous goods according to a classification result.
Preferably, in the present embodiment, the classification results are divided into two types: a first classification result and a second classification result, wherein the first classification result shows that the detected object has no danger, and the second classification result shows that the dangerous object is hidden in the detected object
The specific steps for determining the hidden dangerous goods according to the classification result are as follows:
firstly, judging the type of the classification result, judging that the detected object is a non-hidden dangerous article when the classification result is a first classification result, judging that the detected object is a hidden dangerous article when the classification result is a second classification result, comparing the second classification result with a pre-established spectral feature database of the hidden dangerous article, and judging the type of the hidden dangerous article according to the comparison result.
It can be understood that, when the terahertz spectrum data of different articles are different, the terahertz spectrum characteristics of the articles are compared with the established spectrum characteristic database of various dangerous articles, so that whether the detected articles are dangerous articles or not and which dangerous articles can be judged.
According to the data classification method provided by the invention, the expansion of a digital sample is realized by utilizing a pre-trained FCN-ACGAN network model, the initial ResNet-LSTM classification model is further trained by utilizing an expanded sample to obtain an optimal classification model meeting the data classification standard, then the terahertz time domain spectral data of the detected object is classified by utilizing the optimal classification model, and the hidden dangerous object is determined according to the classification result, so that the technical problems that the classification effect of the classification model obtained by training is poor due to insufficient training data, and the recognition rate of the object attribute is low according to the classification result in the existing data classification method are solved.
On the basis of the foregoing embodiment, the present application provides another preferred embodiment, and step 100 may specifically be implemented by:
the method comprises the steps of carrying out data cleaning on real spectrum data acquired in advance to obtain a first initial sample, then carrying out missing value supplement on the first sample to obtain a second initial sample, and finally carrying out normalization processing on the second initial sample to obtain a real sample.
It can be understood that before the missing value supplement is performed on the data, the data is cleaned, so that the noise data and irrelevant data in the data set can be removed, the data quality is improved, the increase of workload due to the processing of useless data is avoided, further, the data after being cleaned is subjected to the missing value supplement, the data restoration can be realized, the bias risk is reduced, the sample representativeness is improved, the data after being restored is subjected to standardization and normalization operation, the data characteristics are relatively consistent, and the influence of the salient characteristics on the model training is eliminated.
On the basis of the foregoing embodiment, the present application provides another preferred embodiment, in step 200, the FCN-ACGAN network model includes a generator and a discriminator, where the generator and the discriminator respectively include several fully connected layers, and the FCN-ACGAN network model is constructed through the following steps:
s1, pre-training the discriminator by using the real sample to obtain a primary discriminator;
s2, creating a primary generation sample by using the generator;
s3, mixing the primary generated sample with the real sample to obtain a mixed sample;
s4, actually training the primary discriminator and the generator by using the mixed sample, respectively updating network parameters of the primary discriminator and the generator based on a RMSProp optimizer, judging whether the FCN-ACGAN network model reaches Nash equilibrium, if so, stopping training to obtain a trained FCN-ACGAN network model, and if not, returning to the step S2;
wherein, also include before step S2: and pre-training the initial generator by using the real sample to obtain a trained generator.
Referring to fig. 2, the initial FCN-ACGAN model includes a discriminator (network) and a generator (network), and unlike the conventional GAN network, in the FCN-ACGAN model of this embodiment, a full connection layer is added in the generator and the discriminator, respectively, and the full connection layer can help the FCN-ACGAN model to learn the dynamic characteristics between real data, improve the quality of data generated by the generator, and enhance the capability of the discriminator to identify the authenticity of data. Referring to fig. 3 and 4, fig. 3 is a network structure diagram of the generator according to the present embodiment, and fig. 4 is a network structure diagram of the arbiter according to the present embodiment.
The generator consists of 1 input module, 5 sense full-connection layers, 4 Tanh activation function layers and 1 output module. The input of the generator is random noise and label data, and the output is an analog sample.
The discriminator consists of 1 input module, 5 sense full connection layers, 3 ReLU activation function layers and 1 softmax classifier. The input of the discriminator contains the 'simulated' sample and the real sample of the output of the generator, and the output is the discrimination result with the label type.
It can be understood that the untrained initial generator is not clear the "pattern" of the real data at the beginning, and the untrained initial generator is directly used to create data, so that the difference between the distribution of the created data and the distribution of the real data is large, and if the data generated by the initial generator is directly mixed with the real data, and then the mixed data is sent to the discriminator for model training, the training process of the model is increased. Therefore, in order to accelerate the training process, the initial generator needs to perform simulation learning first to make the initial generator have a certain simulation capability, and then the data created by the trained generator is used to train the discriminator.
Similarly, if the data generated by the generator is directly mixed with the real data, and then the mixed data is sent to the discriminator for model training, because the discriminator does not know the 'pattern' of the real data at first, after the mixed data enters the discriminator, the discriminator cannot make accurate judgment, the capability of the discriminator for distinguishing the true data from the false data can be improved only by updating for many times, and in the updating process, the generator is continuously updated, the discrimination capability of the discriminator is slowly increased in the game process, so that the training process of the model can be increased by directly mixing the true data with the false data. In order to accelerate the training process of the FCN-ACGAN model, the embodiment preferentially uses real data to pre-train the discriminator before performing model training by using mixed data, so that the discriminator has the capability of distinguishing real data from fake data at an early stage.
In the actual training stage, the generator establishes a mapping relation with real data distribution by using random noise in a potential space, generates a plurality of simulation samples with labels, then mixes the simulation samples with the real samples to obtain training data, and inputs the training data into a pre-trained discriminator. The arbiter trains by using the training data to obtain the network loss value of the arbiter, and updates the network parameters of the arbiter according to the network loss value of the arbiter and the RMSProp optimizer, and the arbiter becomes more clever every time when the network parameters of the arbiter are updated. When the updating times of the discriminator meet a first updating threshold value, the parameters of the discriminator are kept unchanged, the network loss value of the generator is obtained, the network parameters of the generator are updated according to the network loss value of the generator and the RMSProp optimizer, and similarly, the generator becomes smart every time the generator is updated, until the updating times of the generator meet a preset second updating threshold value. The process is circulated continuously, the generator generates new simulation samples which are more similar to real samples continuously, the new simulation samples and original data are used for training the discriminator, the discriminator is updated continuously to improve discrimination capability, meanwhile, the generator is updated continuously, and the generator and the discriminator are updated alternately until the whole FCN-ACGAN model reaches Nash equilibrium. In the updating process, the learning rates of the generator and the discriminator are preset fixed values, and in the embodiment, the updating frequency of the generator and the discriminator is not specifically limited, and can be set by a person skilled in the art as required.
When the FCN-ACGAN model reaches nash equilibrium, it means that the simulation sample generated by the generator in the FCN-ACGAN model can "cheat" the arbiter, and it can be understood that the simulation sample generated by the generator already meets the requirement as training data, and the expression capability of the simulation sample is close to the real data.
In the above embodiment, by adding the full connection layer in the generator network and the discriminator network, the FCN-ACGAN model can be helped to learn the dynamic characteristics between real data, improve the quality of the generated data of the generator, and enhance the capability of the discriminator to identify the authenticity of the data, so that the generator can generate a simulation sample more similar to the real data, and meanwhile, in the updating process of the generator network and the discriminator, the RMSProp optimizer is used to update the network parameters of the discriminator and the generator, thereby effectively eliminating the jitter caused by the gradient difference in the updating process, and accelerating the model convergence process.
On the basis of the foregoing embodiment, the present application provides another preferred embodiment, and step 300 may specifically be implemented by:
and after enough simulation samples are obtained through the generator, mixing the simulation samples with the real samples to obtain extended samples, training the ResNet-LSTM classification model by using the extended samples, and stopping training when the model error meets a preset threshold value to obtain an optimal classification model, wherein the mixed samples are divided into training samples and actual training samples.
Further, training the ResNet-LSTM classification model by using the mixed samples specifically comprises: constructing an initial ResNet-LSTM classification model, pre-training the initial ResNet-LSTM classification model by utilizing a pre-training sample to obtain a super-parameter of the initial ResNet-LSTM classification model, actually training the initial ResNet-LSTM classification model based on the super-parameter and the actual training sample, and finishing actual training when a model error of the initial ResNet-LSTM classification model meets an error threshold value to obtain an optimal classification model.
Further, the terahertz time-domain spectral data can be classified by using the optimal classification model.
In order to verify the feasibility of the FCN-ACGAN data enhancement-based terahertz time-domain spectrum concealed hazardous article identification method, the embodiment provides an example of hazardous article identification accuracy verification based on the optimal classification model.
Selecting a plurality of preprocessed real samples, simulation samples created by the FCN-ACGAN and extended samples obtained by mixing the simulation samples created by the FCN-ACGAN and the real samples. Wherein the number of the real samples, the simulation samples and the extended samples are consistent. And respectively inputting the original sample, the simulation sample and the extension sample into a ResNet-LSTM classification model for classification and identification, wherein the classification result is shown in Table 1. As can be seen from Table 1, the simulation samples generated by the FCN-ACGAN model are basically consistent with the real samples in the ResNet-LSTM classification model, and the side surface verifies that the simulation samples generated by the FCN-ACGAN model provided by the application can not only capture effective characteristics of original data, but also generate new samples adaptive to the real data characteristics. The recognition rate of the classification model to the extended sample obtained by mixing the simulation sample created by the FCN-ACGAN and the real sample is 99.42%, the recognition rate of the real sample is 98.33%, compared with the real sample, the recognition accuracy of the model obtained by training the extended sample is improved by 1.09%, the recognition precision of article recognition according to the classification result is improved, and the feasibility and the effectiveness of the terahertz time-domain spectrum concealed hazardous article recognition method based on the FCN-ACGAN data enhancement are verified.
TABLE 1 Performance Table combining FCN-ACGAN and ResNet-LSTM recognition algorithms
The invention relates to a terahertz time-domain spectrum concealed hazardous article identification method based on FCN-ACGAN data enhancement, which can effectively improve the over-fitting problem caused by insufficient data after a data set is expanded based on an FCN-ACGAN model, and improve the identification accuracy after the classification model is trained, so that the concealed hazardous article identification accuracy is higher.
The present application further provides an electronic device, which comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data classification method.
It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is substantially or partly contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A terahertz time-domain spectrum concealed hazardous article identification method based on FCN-ACGAN data enhancement is characterized by comprising the following steps:
preprocessing pre-acquired real terahertz spectrum data to obtain a real sample;
generating a simulation sample by using a pre-trained FCN-ACGAN network model; the FCN-ACGAN network model comprises a generator and a discriminator, wherein the generator and the discriminator respectively comprise a plurality of full connection layers; the steps of training the FCN-ACGAN network model are as follows:
s1, pre-training the discriminator by using the real sample to obtain a primary discriminator;
s2, generating a primary simulation sample by using the generator;
s3, mixing the primary simulation sample with the real sample to obtain a training sample;
s4, actually training the primary arbiter and the generator by using the training samples, respectively updating the network parameters of the primary arbiter and the network parameters of the generator based on a RMSProp optimizer, judging whether the FCN-ACGAN network model reaches Nash equilibrium, if so, stopping training to obtain a trained FCN-ACGAN network model, and if not, returning to the S2;
training a pre-constructed initial ResNet-LSTM classification model according to the real sample and the simulation sample to obtain an optimal classification model;
and classifying the terahertz time-domain spectral data by using the optimal classification model, and determining hidden dangerous goods according to a classification result.
2. The method for identifying the terahertz time-domain spectrum concealed hazardous articles based on FCN-ACGAN data enhancement as claimed in claim 1, wherein the training of the pre-constructed initial ResNet-LSTM classification model according to the real samples and the simulation samples to obtain the optimal classification model specifically comprises:
mixing the real sample and the simulation sample to obtain an extended sample, and dividing the extended sample into a pre-training sample and an actual training sample;
constructing an initial ResNet-LSTM classification model, utilizing the pre-training sample to pre-train the initial ResNet-LSTM classification model to obtain a super parameter of the initial ResNet-LSTM classification model, actually training the initial ResNet-LSTM classification model based on the super parameter and the actual training sample, and ending the actual training when the model error of the initial ResNet-LSTM classification model meets a preset error threshold to obtain an optimal classification model.
3. The method for identifying the terahertz time-domain spectroscopy concealed hazardous articles based on FCN-ACGAN data enhancement as claimed in claim 2, wherein the generating of the primary simulation sample by the generator specifically comprises:
establishing a mapping relationship between the real data and random noise of a potential space, and generating a simulation sample based on the mapping relationship, wherein the simulation sample contains a class label.
4. The method for identifying the terahertz time-domain spectroscopy concealed hazardous article based on FCN-ACGAN data enhancement as claimed in claim 3, wherein the updating of the network parameters of the primary discriminator and the generator respectively based on the RMSProp optimizer comprises:
keeping the network parameters of the generator unchanged, acquiring a network loss value of the primary discriminator, updating the primary discriminator based on the RMSProp optimizer and the network loss value of the primary discriminator, keeping the network parameters of the primary discriminator unchanged when the updating times of the primary discriminator meet a preset first updating threshold, acquiring the network loss value of the generator, and updating the generator based on the RMSProp optimizer and the network loss value of the generator until the updating times of the generator meet a preset second updating threshold.
5. The FCN-ACGAN data enhancement based terahertz time-domain spectroscopy concealed hazardous article identification method according to claim 4, wherein the generator comprises: the device comprises an input module, a Dense full-connection layer, a Tanh activation function layer and an output module.
6. The FCN-ACGAN data enhancement based terahertz time-domain spectroscopy concealed hazardous article identification method according to claim 5, wherein the discriminator comprises: the device comprises an input module, a Dense full connection layer, a ReLU activation function layer and a softmax classifier.
7. The method for identifying the concealed hazardous article based on the FCN-ACGAN data enhancement terahertz time-domain spectroscopy as claimed in claim 6, further comprising, before step S2: and pre-training the initial generator by using the real sample to obtain a trained generator.
8. The method for identifying the terahertz time-domain spectrum concealed hazardous article based on FCN-ACGAN data enhancement as claimed in claim 1, wherein the preprocessing is performed on the pre-acquired real spectrum data to obtain the real sample specifically as follows:
performing data cleaning on pre-acquired real spectrum data to obtain a first initial sample;
supplementing missing values to the first sample to obtain a second initial sample;
and carrying out normalization processing on the second initial sample to obtain a real sample.
9. The method for identifying the concealed hazardous article based on the FCN-ACGAN data enhancement terahertz time-domain spectroscopy as claimed in claim 1, wherein the classification result comprises a first classification result and a second classification result, and the determining the concealed hazardous article according to the classification result specifically comprises:
and judging the type of the classification result, judging that the detected object is a non-hidden dangerous article when the classification result is a first classification result, judging that the detected object is a hidden dangerous article when the classification result is a second classification result, comparing the second classification result with a pre-established spectral feature database of the hidden dangerous article, and judging the type of the hidden dangerous article according to the comparison result.
10. An electronic device, comprising a memory and a processor, wherein the memory has stored thereon a computer program, which, when executed by the processor, causes the processor to carry out the steps of the data classification method according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210928339.2A CN115290596A (en) | 2022-08-03 | 2022-08-03 | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210928339.2A CN115290596A (en) | 2022-08-03 | 2022-08-03 | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115290596A true CN115290596A (en) | 2022-11-04 |
Family
ID=83826962
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210928339.2A Pending CN115290596A (en) | 2022-08-03 | 2022-08-03 | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115290596A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116502117A (en) * | 2023-04-13 | 2023-07-28 | 厦门市帕兰提尔科技有限公司 | ResNet-based hazardous chemical identification method, device and equipment |
| CN116663619A (en) * | 2023-07-31 | 2023-08-29 | 山东科技大学 | Data enhancement method, device and medium based on GAN network |
-
2022
- 2022-08-03 CN CN202210928339.2A patent/CN115290596A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116502117A (en) * | 2023-04-13 | 2023-07-28 | 厦门市帕兰提尔科技有限公司 | ResNet-based hazardous chemical identification method, device and equipment |
| CN116502117B (en) * | 2023-04-13 | 2023-12-15 | 厦门市帕兰提尔科技有限公司 | ResNet-based hazardous chemical identification method, device and equipment |
| CN116663619A (en) * | 2023-07-31 | 2023-08-29 | 山东科技大学 | Data enhancement method, device and medium based on GAN network |
| CN116663619B (en) * | 2023-07-31 | 2023-10-13 | 山东科技大学 | Data enhancement method, device and medium based on GAN network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115290596A (en) | FCN-ACGAN data enhancement-based hidden dangerous goods identification method and equipment | |
| CN110766051A (en) | Lung nodule morphological classification method based on neural network | |
| Zhang et al. | Wood defect detection method with PCA feature fusion and compressed sensing | |
| JP6798614B2 (en) | Image recognition device, image recognition method and image recognition program | |
| CN111915437A (en) | RNN-based anti-money laundering model training method, device, equipment and medium | |
| Li et al. | A defect detection method for unpatterned fabric based on multidirectional binary patterns and the gray-level co-occurrence matrix | |
| US20200125836A1 (en) | Training Method for Descreening System, Descreening Method, Device, Apparatus and Medium | |
| Chingovska et al. | Evaluation methodologies for biometric presentation attack detection | |
| CN111582359B (en) | Image identification method and device, electronic equipment and medium | |
| CN110493262B (en) | Classification-improved network attack detection method and system | |
| CN108492298A (en) | Based on the multispectral image change detecting method for generating confrontation network | |
| CN108766464A (en) | Digital audio based on mains frequency fluctuation super vector distorts automatic testing method | |
| CN111932269A (en) | Equipment information processing method and device | |
| CN112163450A (en) | Based on S3High-frequency ground wave radar ship target detection method based on D learning algorithm | |
| CN113673312B (en) | Deep learning-based radar signal intra-pulse modulation identification method | |
| CN113052577A (en) | Method and system for estimating category of virtual address of block chain digital currency | |
| CN104573699A (en) | Trypetid identification method based on medium field intensity magnetic resonance dissection imaging | |
| Saxena et al. | An Examination on Implementation of Deep Fake in Images Through Deep Learning | |
| CN111353607A (en) | Method and device for obtaining quantum state discrimination model | |
| CN108764112A (en) | A kind of Remote Sensing Target object detecting method and equipment | |
| Yao et al. | Design of artistic creation style extraction model based on color feature data | |
| CN108766465A (en) | A kind of digital audio based on ENF universal background models distorts blind checking method | |
| Yang et al. | A semantic information decomposition network for accurate segmentation of texture defects | |
| CN112766320A (en) | Classification model training method and computer equipment | |
| CN112819486B (en) | Method and system for identity certification |
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 |