CN110533051B - Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network - Google Patents
Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network Download PDFInfo
- Publication number
- CN110533051B CN110533051B CN201910710483.7A CN201910710483A CN110533051B CN 110533051 B CN110533051 B CN 110533051B CN 201910710483 A CN201910710483 A CN 201910710483A CN 110533051 B CN110533051 B CN 110533051B
- Authority
- CN
- China
- Prior art keywords
- contraband
- network
- security inspection
- data set
- feature map
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
- G06V10/422—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation for representing the structure of the pattern or shape of an object therefor
- G06V10/424—Syntactic representation, e.g. by using alphabets or grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/05—Recognition of patterns representing particular kinds of hidden objects, e.g. weapons, explosives, drugs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
An automatic detection method for contraband in X-ray security inspection images based on a convolutional neural network. The method comprises the steps of constructing a security inspection image data set comprising a data set A and a data set B, wherein the data set A comprises a training set and a testing set; a semantic enrichment module and a residual error module are added on the basis of the FSSD network, so that an X-ray security inspection image contraband detection network is constructed; training an X-ray security inspection image contraband detection network by using a contraband image in the data set B, then loading network weight into the network, continuing training by using a training set in the data set A, and finally detecting network performance by using a test set in the data set A; inputting any X-ray security inspection image to be detected in the network, and automatically and correctly classifying and accurately positioning the contraband in the detection result graph output by the network (the classification and the accurate position of the contraband in the detection result graph output by the network can be automatically displayed). The method can realize automatic and correct classification and accurate positioning of contraband in the X-ray security inspection image, thereby reducing the working pressure of security inspection workers and improving the working efficiency of security inspection.
Description
Technical Field
The invention belongs to the technical field of detection of contraband in an X-ray security inspection image and computer vision, and particularly relates to an automatic detection method of contraband in an X-ray security inspection image based on a convolutional neural network.
Background
The security inspection is used as an important guarantee for the safety in the public field, and plays an important role in protecting the safety of people. In China, a large number of passengers are transported every year by transportation departments such as civil aviation, railways and the like. The large traffic demands bring huge working pressure to security personnel. Taking civil aviation security inspection as an example, most aviation accidents are caused by unsafe behaviors of human beings. Airport security check staff is as the work that a pressure is great, and the operational environment of long-term high pressure is inevitable can lead to its work mistake to influence the safety of aviation operation. Therefore, establishing a reliable automatic security inspection system is of great significance for improving the working efficiency of security inspection workers.
When a security check worker detects contraband, the working process of the security check worker is to firstly identify whether the baggage has the contraband and secondly give the correct position of the contraband. The target detection algorithm existing in the field of computer vision at present solves the problem of target location. Therefore, the computer is an effective solution for realizing automatic security inspection.
In recent years, with the rapid development of deep learning, especially convolutional neural networks, more and more target detection methods based on convolutional neural networks are proposed, and remarkable detection effects are achieved in the field of target detection. However, these target detection methods based on convolutional neural networks are mostly used for target detection in natural images. Different from natural images, security inspection images have the characteristics of complex background, large size difference of contraband, and the like, so that the existing target detection method based on the convolutional neural network is required to be improved according to the characteristics of the security inspection images, and the contraband in the security inspection images can be better detected.
Disclosure of Invention
In order to solve the above problems, an object of the present invention is to provide a method for automatically detecting contraband in an X-ray security image based on a convolutional neural network.
In order to achieve the purpose, the automatic detection method of contraband in the X-ray security inspection image based on the convolutional neural network comprises the following steps in sequence:
1) Constructing a security inspection image data set comprising a data set A and a data set B, wherein the data set A comprises a training set and a test set;
2) A semantic enrichment module and a residual error module are added on the basis of the FSSD network, so that an X-ray security inspection image contraband detection network is constructed;
3) Firstly, training the contraband detection network constructed in the step 2) by using the contraband image in the data set B obtained in the step 1), then loading the network weight trained by the data set B into the X-ray safety inspection image contraband detection network, then continuing training the X-ray safety inspection image contraband detection network by using the training set in the data set A, and finally testing the performance of the X-ray safety inspection image contraband detection network by using the test set in the data set A;
4) After the performance test of the X-ray security inspection image contraband detection network is qualified, any X-ray security inspection image to be detected is input into the network, and the contraband in the detection result image output by the network can be automatically and correctly classified and accurately positioned.
In step 1), the method for constructing the security inspection image data set including the data set a and the data set B includes:
the X-ray security inspection machine is used for collecting X-ray security inspection images of a plurality of pieces of luggage containing contraband, six types of contraband are selected, and the X-ray security inspection images are respectively as follows: the system comprises a mobile power supply, a lighter, a fork, a knife, a pistol and scissors, wherein a data set A is formed by X-ray security inspection images; the X-ray security inspection images in the data set A are divided into simple images and complex images, and the simple images are characterized in that the background is complex and only contain one contraband; the complex image is characterized in that the background is complex and contains two to three contraband articles; all X-ray security inspection images are 300X 300 pixels in size; then respectively randomly extracting 75% of simple images and 75% of complex images in the data set A as training sets, and taking the rest 25% of complex images as test sets;
the data set B comprises a plurality of contraband images used for training an X-ray security inspection image contraband detection network, each contraband image only comprises contraband and has no complex background, all the contraband are extracted from the plurality of collected X-ray security inspection images by using an image preprocessing method, and the size of the contraband image is 300X 300 pixels; and finally, marking the contraband in the data set A and the data set B by using a labelImg marking tool to obtain the prior information of the contraband.
In step 2), the method for constructing the X-ray security inspection image contraband detection network by adding the semantic enrichment module and the residual error module on the basis of the FSSD network is as follows:
taking the FSSD network as a basic detection network of an X-ray security inspection image contraband detection network; a semantic enrichment module is added on the low-level feature map generated by the FSSD network; firstly, inputting a low-level feature map into a semantic enrichment module, then utilizing a void convolution layer to perform feature extraction on the input low-level feature map to obtain a feature map with abundant semantic information, and finally performing dot multiplication operation on the generated feature map with abundant semantic information and the input low-level feature map to obtain a final output feature map; the output feature map and the low-level feature map have the same dimension, and the output feature map replaces the input low-level feature map to carry out feature map splicing;
suppose X ∈ R C×H×W For the input low-level feature map, Y is equal to R C×H×W The feature map with rich semantic information is generated after the input low-level feature map is subjected to hole convolution:
Y＝H(X)∈R C×H×W (1)
performing dot product operation on the feature map Y rich in semantic information and the input low-level feature map X to obtain a final output feature map Z, namely:
Z＝X⊙Y (2)
after the characteristic diagram splicing is completed, a residual module is formed by adding several additional convolution layers and shortcut connections on the basis of the FSSD network.
The automatic detection method for the contraband in the X-ray security inspection image based on the convolutional neural network can realize automatic correct classification and accurate positioning of the contraband in the X-ray security inspection image, so that the working pressure of security inspection workers can be reduced, and the working efficiency of security inspection is improved.
Drawings
FIG. 1 is a schematic view of a security image dataset in accordance with the present invention.
Fig. 2 is a schematic diagram of a basic detection network structure in the present invention.
Fig. 3 is a schematic diagram of a contraband detection network of an X-ray security inspection image according to the present invention.
FIG. 4 is a schematic diagram of a semantic rich module according to the present invention.
Fig. 5 is a schematic diagram of a residual module structure according to the present invention.
Fig. 6 is a diagram illustrating the detection result of contraband in the X-ray security inspection image.
Detailed Description
The following describes in detail the automatic detection method of contraband in X-ray security inspection images based on convolutional neural network according to the present invention with reference to the accompanying drawings and specific embodiments.
The automatic detection method of contraband in the X-ray security inspection image based on the convolutional neural network comprises the following steps in sequence:
1) Constructing a security inspection image data set comprising a data set A and a data set B, wherein the data set A comprises a training set and a test set;
the security image data set is an important part for solving the security problem by using a convolutional neural network, and in order to make the used data more meaningful, the inventor utilizes an X-ray security inspection machine to acquire X-ray security inspection images of a plurality of pieces of luggage containing contraband. Contraband in the X-ray security inspection image of collection is various in size, and the background is complicated, and various article are placed at will and can overlap. Considering the diversity characteristics of the sizes of contraband in the X-ray security inspection image, six classes of contraband are finally selected, which are respectively as follows: portable power source, lighter, fork, sword, pistol and scissors. The X-ray security inspection images form a data set a, as shown in fig. 1, the X-ray security inspection images in the data set a are divided into two types, namely simple images and complex images, and the simple images are characterized in that the background is complex and only contain one contraband; the complex image is characterized by a complex background and two to three contraband articles. All X-ray security images were 300X 300 pixels in size. Then respectively randomly extracting 75% of simple images and 75% of complex images in the data set A as a training set, and taking the rest 25% of complex images as a test set; in the present invention, a total of 2074 simple images and 2178 complex images are acquired.
In order to ensure that the following X-ray security inspection image contraband detection network is not interfered by a complex background in an X-ray security inspection image during training, so as to more fully learn the characteristics of contraband, the inventor also constructs a data set B, wherein the data set B comprises a plurality of contraband images used for training the X-ray security inspection image contraband detection network, each contraband image only comprises the contraband and has no complex background, all the contrabands are extracted from the plurality of X-ray security inspection images by using an image preprocessing method, and the sizes of the contraband images are both 300 pixels by 300 pixels. And finally, marking the contraband in the data set A and the data set B by using a labelImg marking tool to obtain the prior information of the contraband. In the present invention, 1645 images of contraband are included in the data set B.
2) A semantic enrichment module and a residual error module are added on the basis of the FSSD network, so that an X-ray security inspection image contraband detection network is constructed;
the invention takes an FSSD (feature fusion single step target detector) network belonging to a convolutional neural network as a basic detection network of an X-ray security inspection image contraband detection network. The structure of the FSSD network is shown in fig. 2. The FSSD network utilizes a feature pyramid to detect contraband of different sizes. The low-level feature map in the feature pyramid contains more position information and detail information, which is beneficial to the positioning of contraband and the identification of small-size contraband, and the high-level feature map contains more abstract semantic information, which is beneficial to the identification of the contraband with larger size. In addition, since the detection of small-size contraband also depends on context information, the FSSD network performs dimensional stitching on the high-level feature map and the low-level feature map in the feature pyramid, so that the detection accuracy of small-size contraband can be improved.
Unlike natural scene images and other X-ray images, the size difference of contraband in X-ray security images is large, which increases the difficulty in automatically detecting contraband in X-ray security images. In order to better detect small-size contraband in the X-ray security inspection image and improve the detection precision, the invention adds a semantic enrichment module and a residual error module on the basis of the FSSD network, thereby constructing the X-ray security inspection image contraband detection network, as shown in FIG. 3.
In order to enrich semantic information of low-level feature maps in the feature pyramid and improve detection accuracy of small-size contraband, the invention adds a semantic enrichment module on the low-level feature maps (first-level feature maps for splicing) generated by the FSSD network. The semantic enrichment module is mainly realized by a hole convolution layer, and the hole convolution enriches the semantic information of the characteristic diagram by increasing the receptive field. In the semantic rich module, the convolution kernel size of the hole convolution layer is 3 × 3, the void ratio of the first three hole convolution layers is 2, and the void ratio of the last hole convolution layer is 4.
The work flow of the semantic enrichment module is as shown in fig. 4, firstly, the low-level feature map is input into the semantic enrichment module, then the cavity convolution layer is used for carrying out feature extraction on the input low-level feature map to obtain a feature map with rich semantic information, and finally, the generated feature map with rich semantic information and the input low-level feature map are subjected to point multiplication operation to obtain a final output feature map. The output feature map has the same dimensions as the low-level feature map and the output feature map will be used for feature map stitching instead of the input low-level feature map.
Suppose X ∈ R C×H×W For the input low-level feature map, Y is equal to R C×H×W The feature map with rich semantic information generated by the input low-level feature map after the hole convolution is obtained from fig. 4:
Y＝H(X)∈R C×H×W (1)
performing dot product operation on the feature map Y rich in semantic information and the input low-level feature map X to obtain a final output feature map Z, namely:
Z＝X⊙Y (2)
in order to better extract the characteristics of contraband, after the characteristic diagram is spliced, a plurality of additional convolutional layers are added on the basis of the FSSD network, and in order to prevent the problem of network performance degradation caused by the fact that the network is too deep, a residual error structure is added in the spliced convolutional layers.
As shown in fig. 5, the residual module is mainly implemented by convolution operations and shortcut connections (shortcuts connections) operations. In the residual module, the size of the convolution kernel used is 3 × 3, when the feature maps generated by the convolution layers are the same, the convolution layers use the same number of convolution kernels, and two feature maps with the same size can be directly subjected to shortcut connection operation. When the size of the feature map generated by the convolutional layers is half of the original size, the number of convolution kernels used by the convolutional layers is twice of the original number of convolution kernels, and when two feature maps with different sizes are subjected to shortcut connection operation, the size of the feature map with a larger size needs to be adjusted through downsampling operation, so that the two feature maps subjected to the shortcut connection operation have the same size.
3) Firstly, training the contraband detection network constructed in the step 2) by using the contraband image in the data set B obtained in the step 1), then loading the network weight trained by the data set B into the X-ray safety inspection image contraband detection network, then continuing training the X-ray safety inspection image contraband detection network by using the training set in the data set A, and finally testing the performance of the X-ray safety inspection image contraband detection network by using the test set in the data set A;
4) After the performance test of the X-ray security inspection image contraband detection network is qualified, any X-ray security inspection image to be detected is input into the network, and the contraband in the detection result image output by the network can be automatically and correctly classified and accurately positioned.
In order to verify the effect of the method of the present invention, the present inventors performed an experiment on an X-ray security inspection image by using the method of the present invention, and compared the experiment results using an SSD (single step target detector) network and an FSSD network, and the detection result graph is shown in fig. 6. The experiment was performed in the ubuntu16.04 system with the programming language python3.5 and the deep learning framework pytorch0.3. The algorithm runs on the GPU, and the model of the video card is NVIDIA 1080Ti 11GB. The average accuracy mean (mAP) is selected as the evaluation index of the detection accuracy, the experimental result is shown in Table 1,
as can be seen from Table 1, the method of the present invention improves the mean average accuracy of the SSD network and FSSD network detection by 6.9% and 2.3%, respectively. The SSD network and the FSSD network can well detect contraband articles with larger sizes, for example, the average accuracy mean values of detection for the mobile power source are 90.6% and 90.8%, respectively, and the average accuracy mean values of detection for the handgun are 97.0% and 98.8%, respectively. Although the detection accuracy of the FSSD network is improved compared with that of the SSD network when detecting the contraband with smaller size, the method of the invention obtains the best detection result when detecting the contraband with smaller size, for example, compared with the SSD network and the FSSD network, the method of the invention respectively improves the detection accuracy of the lighter from 72.1% to 89.2% and from 87.1% to 89.2%. Therefore, the method can well detect the contraband with smaller size in the X-ray security inspection image.
TABLE 1
Claims (2)
1. A method for automatically detecting contraband in an X-ray security inspection image based on a convolutional neural network is characterized by comprising the following steps: the automatic detection method for contraband in the X-ray security inspection image based on the convolutional neural network comprises the following steps in sequence:
1) Constructing a security inspection image data set comprising a data set A and a data set B, wherein the data set A comprises a training set and a test set;
2) A semantic enrichment module and a residual error module are added on the basis of the FSSD network, so that an X-ray security inspection image contraband detection network is constructed;
3) Firstly, training the contraband detection network constructed in the step 2) by using the contraband image in the data set B obtained in the step 1), then loading the network weight trained by the data set B into the X-ray safety inspection image contraband detection network, then continuing training the X-ray safety inspection image contraband detection network by using the training set in the data set A, and finally testing the performance of the X-ray safety inspection image contraband detection network by using the test set in the data set A;
4) After the performance test of the X-ray security inspection image contraband detection network is qualified, inputting any X-ray security inspection image to be detected in the network, and automatically and correctly classifying and accurately positioning contraband in a detection result graph output by the network;
in step 2), the method for constructing the X-ray security inspection image contraband detection network by adding the semantic enrichment module and the residual error module on the basis of the FSSD network is as follows:
taking the FSSD network as a basic detection network of an X-ray security inspection image contraband detection network; a semantic enrichment module is added on the low-level feature map generated by the FSSD network; the low-level feature map is a first-level feature map used for splicing; firstly, inputting a low-level feature map into a semantic enrichment module, then utilizing a void convolution layer to perform feature extraction on the input low-level feature map to obtain a feature map with rich semantic information, and finally performing dot multiplication operation on the generated feature map with rich semantic information and the input low-level feature map to obtain a final output feature map; the output feature map and the low-level feature map have the same dimension, and the output feature map replaces the input low-level feature map to carry out feature map splicing;
suppose X ∈ R C×H×W For the input low-level feature map, Y is equal to R C×H×W The feature map with rich semantic information is generated after the input low-level feature map is subjected to hole convolution:
Y＝H(X)∈R C×H×W (1)
performing dot product operation on the feature map Y rich in semantic information and the input low-level feature map X to obtain a final output feature map Z, namely:
Z＝X⊙Y (2)
after the characteristic diagram splicing is completed, the convolution layer and the shortcut connection are added on the basis of the FSSD network to form a residual module.
2. The automatic detection method of contraband in X-ray security inspection image based on convolutional neural network as claimed in claim 1, characterized in that: in step 1), the method for constructing the security inspection image data set including the data set a and the data set B includes:
the X-ray security inspection machine is used for collecting X-ray security inspection images of a plurality of pieces of luggage containing contraband, six types of contraband are selected, and the X-ray security inspection images are respectively as follows: the system comprises a mobile power supply, a lighter, a fork, a knife, a pistol and scissors, wherein a data set A is formed by X-ray security inspection images; the X-ray security inspection images in the data set A are divided into simple images and complex images, and the simple images are characterized in that the background is complex and only contain one contraband; the complex image is characterized in that the background is complex and contains two to three contraband articles; all X-ray security inspection images are 300X 300 pixels in size; then respectively randomly extracting 75% of simple images and 75% of complex images in the data set A as training sets, and taking the rest 25% of complex images as test sets;
the data set B comprises a plurality of contraband images used for training an X-ray security inspection image contraband detection network, each contraband image only comprises contraband and has no complex background, all the contraband are extracted from the plurality of collected X-ray security inspection images by using an image preprocessing method, and the size of the contraband image is 300X 300 pixels; and finally, marking the contraband in the data set A and the data set B by using a labelImg marking tool to obtain the prior information of the contraband.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910710483.7A CN110533051B (en) | 2019-08-02 | 2019-08-02 | Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910710483.7A CN110533051B (en) | 2019-08-02 | 2019-08-02 | Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110533051A CN110533051A (en) | 2019-12-03 |
| CN110533051B true CN110533051B (en) | 2023-01-17 |
Family
ID=68661364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910710483.7A Active CN110533051B (en) | 2019-08-02 | 2019-08-02 | Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110533051B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160434A (en) * | 2019-12-19 | 2020-05-15 | 中国平安人寿保险股份有限公司 | Training method and device of target detection model and computer readable storage medium |
| CN111474186A (en) * | 2020-03-31 | 2020-07-31 | 安徽理工大学 | X-ray imaging and CNN express package contraband detection method |
| CN111582367A (en) * | 2020-05-07 | 2020-08-25 | 电子科技大学 | Small metal threat detection method |
| CN111899220A (en) * | 2020-06-29 | 2020-11-06 | 北京航空航天大学 | Contraband detection method |
| CN112215266B (en) * | 2020-09-25 | 2022-07-19 | 华东师范大学 | X-ray image contraband detection method based on small sample learning |
| CN112562255B (en) * | 2020-12-03 | 2022-06-28 | 国家电网有限公司 | Intelligent image detection method for cable channel smoke and fire conditions in low-light-level environment |
| CN112861924B (en) * | 2021-01-17 | 2023-04-07 | 西北工业大学 | Visible light/infrared image multi-platform distributed fusion multi-target detection method |
| CN113706497B (en) * | 2021-08-24 | 2024-04-09 | 南京金盾公共安全技术研究院有限公司 | Intelligent contraband identification device and system |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086807A (en) * | 2018-07-16 | 2018-12-25 | 哈尔滨工程大学 | A kind of semi-supervised light stream learning method stacking network based on empty convolution |
| CN109145920A (en) * | 2018-08-21 | 2019-01-04 | 电子科技大学 | A kind of image, semantic dividing method based on deep neural network |
| CN109271856A (en) * | 2018-08-03 | 2019-01-25 | 西安电子科技大学 | Remote sensing image object detection method based on expansion residual error convolution |
| CN109389556A (en) * | 2018-09-21 | 2019-02-26 | 五邑大学 | The multiple dimensioned empty convolutional neural networks ultra-resolution ratio reconstructing method of one kind and device |
| CN109522966A (en) * | 2018-11-28 | 2019-03-26 | 中山大学 | A kind of object detection method based on intensive connection convolutional neural networks |
| CN109584339A (en) * | 2018-12-04 | 2019-04-05 | 中国民航大学 | A kind of safety check contraband x-ray image generation method based on production confrontation network |
| CN109636905A (en) * | 2018-12-07 | 2019-04-16 | 东北大学 | Environment semanteme based on depth convolutional neural networks builds drawing method |
| CN109741383A (en) * | 2018-12-26 | 2019-05-10 | 西安电子科技大学 | Picture depth estimating system and method based on empty convolution sum semi-supervised learning |
| CN109784203A (en) * | 2018-12-24 | 2019-05-21 | 中国科学院大学 | Weakly supervised x-ray image contraband inspection method based on delamination propagation and activation |
-
2019
- 2019-08-02 CN CN201910710483.7A patent/CN110533051B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086807A (en) * | 2018-07-16 | 2018-12-25 | 哈尔滨工程大学 | A kind of semi-supervised light stream learning method stacking network based on empty convolution |
| CN109271856A (en) * | 2018-08-03 | 2019-01-25 | 西安电子科技大学 | Remote sensing image object detection method based on expansion residual error convolution |
| CN109145920A (en) * | 2018-08-21 | 2019-01-04 | 电子科技大学 | A kind of image, semantic dividing method based on deep neural network |
| CN109389556A (en) * | 2018-09-21 | 2019-02-26 | 五邑大学 | The multiple dimensioned empty convolutional neural networks ultra-resolution ratio reconstructing method of one kind and device |
| CN109522966A (en) * | 2018-11-28 | 2019-03-26 | 中山大学 | A kind of object detection method based on intensive connection convolutional neural networks |
| CN109584339A (en) * | 2018-12-04 | 2019-04-05 | 中国民航大学 | A kind of safety check contraband x-ray image generation method based on production confrontation network |
| CN109636905A (en) * | 2018-12-07 | 2019-04-16 | 东北大学 | Environment semanteme based on depth convolutional neural networks builds drawing method |
| CN109784203A (en) * | 2018-12-24 | 2019-05-21 | 中国科学院大学 | Weakly supervised x-ray image contraband inspection method based on delamination propagation and activation |
| CN109741383A (en) * | 2018-12-26 | 2019-05-10 | 西安电子科技大学 | Picture depth estimating system and method based on empty convolution sum semi-supervised learning |
Non-Patent Citations (1)
| Title |
|---|
| 基于FD-SSD的遥感图像多目标检测方法;朱敏超;《计算机应用与软件》;20190131;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110533051A (en) | 2019-12-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110533051B (en) | Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network | |
| CN106611169B (en) | A kind of dangerous driving behavior real-time detection method based on deep learning | |
| CN112926405B (en) | Method, system, equipment and storage medium for detecting wearing of safety helmet | |
| CN110018524B (en) | X-ray security inspection contraband identification method based on vision-attribute | |
| CN110188209A (en) | Cross-module state Hash model building method, searching method and device based on level label | |
| CN106650740B (en) | A kind of licence plate recognition method and terminal | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN112801146B (en) | Target detection method and system | |
| CN110490238A (en) | A kind of image processing method, device and storage medium | |
| CN105205449A (en) | Sign language recognition method based on deep learning | |
| CN110059734A (en) | A kind of training method, object identification method, device, robot and the medium of target identification disaggregated model | |
| CN108875805A (en) | The method for improving detection accuracy using detection identification integration based on deep learning | |
| CN112712093A (en) | Security check image identification method and device, electronic equipment and storage medium | |
| CN104504412A (en) | Method and system for extracting and identifying handwriting stroke features | |
| CN109583357A (en) | A kind of improvement LBP and the cascade face identification method of light weight convolutional neural networks | |
| CN105260734A (en) | Commercial oil surface laser code recognition method with self modeling function | |
| CN109858414A (en) | A kind of invoice piecemeal detection method | |
| CN114495029B (en) | Traffic target detection method and system based on improved YOLOv4 | |
| CN109344740A (en) | Face identification system, method and computer readable storage medium | |
| CN110488368A (en) | A kind of contraband recognition methods and device based on dual intensity X-ray screening machine | |
| CN106056074A (en) | Single training sample face identification method based on area sparse | |
| CN116662839A (en) | Associated big data cluster analysis method and device based on multidimensional intelligent acquisition | |
| CN116012653A (en) | Method and system for classifying hyperspectral images of attention residual unit neural network | |
| Lu | Improved yolov8 detection algorithm in security inspection image | |
| CN110321867A (en) | Shelter target detection method based on part constraint network |
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 |