CN106485268A - A kind of image-recognizing method and device - Google Patents
A kind of image-recognizing method and device Download PDFInfo
- Publication number
- CN106485268A CN106485268A CN201610854506.8A CN201610854506A CN106485268A CN 106485268 A CN106485268 A CN 106485268A CN 201610854506 A CN201610854506 A CN 201610854506A CN 106485268 A CN106485268 A CN 106485268A
- Authority
- CN
- China
- Prior art keywords
- image
- scanogram
- obtains
- neural network
- depth convolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- 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
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Analysing Materials By The Use Of Radiation (AREA)
Abstract
The application is related to a kind of image-recognizing method and device, and methods described includes：Obtain scanogram；Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Carry out target detection using the described image feature extracted, based on depth convolution multilayer neural network target detection model, obtain candidate target；Using the described image feature extracted, based on depth convolution multilayer neural network object-class model, described candidate target is identified, obtains image recognition result；Wherein, described to described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain the characteristics of image of each level of depth convolution multilayer neural network, fusion treatment is carried out to the characteristics of image of described each level, obtain the characteristics of image after merging as the characteristics of image extracting.The application can improve accuracy rate and the efficiency of target detection.
Description
Technical field
The application is related to technical field of image processing, more particularly, to a kind of image-recognizing method and device.
Background technology
In the customs supervision such as trade port, railway station, airport region it is often necessary to be carried to passenger using rays safety detection apparatus
Article detected, to determine described article whether as dangerous materials or smuggled goods.How quick during safety check, accurate
Really detect that dangerous materials or smuggled goods become a problem demanding prompt solution.
In prior art, when passenger's belongings pass through rays safety detection apparatus, arise that through the image that X-ray scanning goes out
On the display screen being connected with rays safety detection apparatus, whether the image on security staff's manual observation screen is to be danger to article therein
Dangerous product or die Konterbande are identified., there is workload in prior art this artificial cognition dangerous materials or the method for die Konterbande
Greatly, the problem that efficiency is low, accuracy rate is not high.
Content of the invention
For solving existing technical problem, the application expectation provides a kind of image-recognizing method and device, Ke Yitong
Cross and automatically image is identified, improve accuracy rate and the efficiency of detection.
A kind of first aspect according to the embodiment of the present application, there is provided image-recognizing method, methods described includes：Acquisition is swept
Tracing picture；Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Using extract described image feature,
Target detection is carried out based on depth convolution multilayer neural network target detection model, obtains candidate target；Described in extracting
Characteristics of image, based on depth convolution multilayer neural network object-class model, described candidate target is identified, obtains image
Recognition result；Wherein, described to described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain
The characteristics of image of each level of depth convolution multilayer neural network, carries out fusion treatment to the characteristics of image of described each level, obtains
Characteristics of image after fusion is as the characteristics of image extracting.
Alternatively, before carrying out feature extraction to described scanogram, methods described also includes：Described scanogram is entered
Row pretreatment, is that different classes of scanned item image arranges different colors based on scanned item classification results.
Alternatively, described pretreatment is carried out to described scanogram, be different classes of based on scanned item classification results
The different color of scanned item image setting includes：Obtain the atomic number of scanned item, obtained based on described atomic number and sweep
Retouch the density of article；Determine the classification of scanned item according to the density of described scanned item, obtain scanned item classification results；Base
It is that different classes of scanned item image arranges different colors in described scanned item classification results.
Alternatively, described described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and include：Based on scanning
The color characteristic of article determines object candidate area；Carry out feature extraction process in described object candidate area, obtain and extract
Characteristics of image.
Alternatively, described using extract described image feature, be based on depth convolution multilayer neural network target detection mould
Type carries out target detection, obtains candidate target and includes：Using the described image feature extracted, based on multiple depth convolution multilamellar god
Carry out target detection through network objectives detection model, obtain multiple testing results；Merge the plurality of testing result, obtain final
Testing result as candidate target.
Alternatively, the plurality of testing result of described fusion, obtains final testing result and includes as candidate target：Base
In confidence calculations result, merge multiple testing results to obtain final testing result.
Alternatively, methods described also includes：Based on described image recognition result, judge whether dangerous materials or smuggling
Article；If judging dangerous product or smuggled goods, export information.
Alternatively, methods described also includes：Described image recognition result and list of articles are compared, obtains and compare knot
Really；Export described comparison result.
A kind of second aspect according to the embodiment of the present application, there is provided pattern recognition device, described device includes：Image obtains
Delivery block, for obtaining scanogram；Characteristic extracting module, for described scanogram is carried out with feature extraction, obtains and extracts
Characteristics of image；Wherein, described described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain
The characteristics of image of each level of depth convolution multilayer neural network, carries out fusion treatment to the characteristics of image of described each level, obtains
Characteristics of image after fusion is as the characteristics of image extracting；Module of target detection, for using the described image feature extracted, base
Carry out target detection in depth convolution multilayer neural network target detection model, obtain candidate target；Target classification module, is used for
Using the described image feature extracted, based on depth convolution multilayer neural network object-class model, described candidate target is carried out
Identification, obtains image recognition result.
Alternatively, described device also includes：Pretreatment module, for carrying out pretreatment to described scanogram, based on sweeping
Retouch the different color of the scanned item image setting that taxonomy of goods result is different classes of.
Alternatively, described pretreatment module specifically includes：Density acquiring unit, for obtaining the atomic number of scanned item
Number, obtains the density of scanned item based on described atomic number；Taxon, determines for the density according to described scanned item
The classification of scanned item, obtains scanned item classification results；Color arranging unit, for based on described scanned item classification results
For the different color of different classes of scanned item image setting.
Alternatively, described characteristic extracting module is specifically for determining target candidate area based on the color characteristic of scanned item
Domain；Carry out feature extraction process in described object candidate area, obtain the characteristics of image extracting.
Alternatively, described module of target detection specifically includes：Multi-model detector unit, for using the described image extracted
Feature, the target detection that carried out based on multiple depth convolution multilayer neural network target detection models, obtain multiple testing results；Knot
Fruit integrated unit, for merging the plurality of testing result, obtains final testing result as candidate target.
Alternatively, described result integrated unit specifically for based on confidence calculations result, merge multiple testing results with
Obtain final testing result.
Alternatively, described device also includes：Judge module, for based on described image recognition result, judging whether
Dangerous materials or smuggled goods；First output module, if for judging dangerous product or smuggled goods, output prompting letter
Breath.
Alternatively, described device also includes：Comparing module, for being compared described image recognition result and list of articles
Right, obtain comparison result；Second output module, for exporting described comparison result.
The third aspect according to the embodiment of the present application, there is provided for the device of image recognition, include memorizer, and
One or more than one program, one of or more than one program storage is in memorizer, and is configured to by one
Individual or more than one computing device is one or more than one program bag contains the instruction for carrying out following operation：
Obtain scanogram；Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Using extract
Described image feature, the target detection that carried out based on depth convolution multilayer neural network target detection model, obtain candidate target；Profit
With the described image feature of extraction, based on depth convolution multilayer neural network object-class model, described candidate target is known
Not, obtain image recognition result；Wherein, described to described scanogram is carried out with feature extraction, obtain the characteristics of image extracting
It is specially：Obtain the characteristics of image of each level of depth convolution multilayer neural network, the characteristics of image of described each level is melted
Conjunction is processed, and obtains the characteristics of image after merging as the characteristics of image extracting.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Pretreatment is carried out to described scanogram, is different classes of scanning thing based on scanned item classification results
The different color of product image setting.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Obtain the atomic number of scanned item, obtain the density of scanned item based on described atomic number；According to institute
The density stating scanned item determines the classification of scanned item, obtains scanned item classification results；Based on the classification of described scanned item
Result is that different classes of scanned item image arranges different colors.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Object candidate area is determined based on the color characteristic of scanned item；Carry out in described object candidate area
Feature extraction is processed, and obtains the characteristics of image extracting.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Using the described image feature extracted, it is based on multiple depth convolution multilayer neural network target detection models
Carry out target detection, obtain multiple testing results；Merge the plurality of testing result, obtain final testing result as candidate
Target.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Based on confidence calculations result, merge multiple testing results to obtain final testing result.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Based on described image recognition result, judge whether dangerous materials or smuggled goods；If judging there is danger
Dangerous product or smuggled goods, export information.
Alternatively, described process implement body be additionally operable to execute one or more than one program bag contain be used for carrying out with
The instruction of lower operation：Described image recognition result and list of articles are compared, obtains comparison result；Output is described to compare knot
Really.
Image-recognizing method and device that the embodiment of the present application provides, can extract to scanogram feature based, utilize
The characteristics of image extracting is based on depth convolution multilayer neural network target detection model and depth convolution multilayer neural network mesh
Mark disaggregated model carries out target detection and classification, automatically obtains image recognition result, improves the efficiency of detection.Further, since
When carrying out feature extraction, obtain the characteristics of image of each level of depth convolution multilayer neural network respectively, to described each level
Characteristics of image carries out fusion treatment, obtains the characteristics of image after merging as the characteristics of image extracting, is derived from characteristics of image
More accurate, effectively increase image detection and the accuracy and efficiency of classification.
Brief description
For the technical scheme being illustrated more clearly that in the embodiment of the present application, will make to required in embodiment description below
Accompanying drawing be briefly described it should be apparent that, drawings in the following description are only some embodiments of the present application, for
For those of ordinary skill in the art, without having to pay creative labor, it can also be obtained according to these accompanying drawings
His accompanying drawing.
The image-recognizing method flow chart that Fig. 1 provides for the application one embodiment；
Fig. 2 processes schematic diagram for the image co-registration that the embodiment of the present application provides；
The multi-model fusion treatment schematic diagram that Fig. 3 provides for the embodiment of the present application；
The image-recognizing method flow chart that Fig. 4 provides for another embodiment of the application
A kind of pattern recognition device schematic diagram that Fig. 5 provides for the embodiment of the present application；
Fig. 6 is the block diagram of the pattern recognition device that another embodiment of the application provides.
Specific embodiment
The purpose of the application is to provide a kind of image-recognizing method and device, can by being automatically identified to image,
Improve accuracy rate and the efficiency of detection.
For enabling present invention purpose, feature, advantage more obvious and understandable, below in conjunction with the application
Accompanying drawing in embodiment, is described to the technical scheme in the embodiment of the present application it is clear that described embodiment is only this
Apply for a part of embodiment, and not all embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not having
The every other embodiment being obtained under the premise of making creative work, broadly falls into the scope of the application protection.
As shown in figure 1, being the flow chart of the image-recognizing method according to the application one embodiment, specifically for example can wrap
Include：
S101, obtains scanogram.
Wherein, described scanogram is specifically as follows the radioscopic image of X-ray rays safety detection apparatus collection.
S102, carries out feature extraction to described scanogram, obtains the characteristics of image extracting.
When implementing, before feature extraction is carried out to described scanogram, described scanogram can also be carried out pre-
Process, described pretreatment can include：The scanned item image setting being different classes of based on scanned item classification results is different
Color.Specifically, the atomic number of scanned item can be obtained, obtain the density of scanned item based on described atomic number；
Determine the classification of scanned item according to the density of described scanned item, obtain scanned item classification results；Based on described scanning thing
Product classification results are that different classes of scanned item image arranges different colors.Illustrate, Image semantic classification can be passed through,
Scanned item is divided into Organic substance and inorganic matters two class, to detect offer prior information for image object, improves target detection accurate
Exactness.It is for instance possible to use mild steel (corresponding inorganic matters) and two kinds of materials of lucite (corresponding Organic substance) configure effectively
Atomic is the different densities of (7,25) material, and sets up look-up table by linear interpolation.When obtaining scanogram, pass through
High and low power X ray irradiating item with electron radiation, to obtain the atomic number of different article, is tabled look-up according to described atomic number and can be obtained thing
The density of body.Then, the density according to object can determine that object is inorganic matters or Organic substance, is inorganic matters and Organic substance
Different colors are set.For example, inorganic matters can be represented with blueness, and Organic substance can be represented with orange.So, just for scanning figure
As imparting colouring information.It is, of course, also possible to be that different article arrange different colors according to different density values, here is not
It is defined.
In some embodiments, described scanogram is being carried out with feature extraction, when obtaining the characteristics of image extracting, can
Object candidate area is determined with the color characteristic based on scanned item；Carry out at feature extraction in described object candidate area
Reason, obtains the characteristics of image extracting.Illustrate it is assumed that dangerous materials to be detected are the inorganic matters such as knife, rifle, in advance for inorganic matters
The blueness of the color of this classification setting, then in feature extraction, using blue region as object candidate area, can be just for
The image of described object candidate area carries out feature extraction, which thereby enhances efficiency and the accuracy rate of image procossing.
In some embodiments, the described characteristics of image tool described scanogram being carried out with feature extraction, obtaining extraction
Body is：Obtain the characteristics of image of each level of depth convolution multilayer neural network, the characteristics of image of described each level is merged
Process, obtain the characteristics of image after merging as the characteristics of image extracting.It should be noted that for the standard improving image recognition
Really rate, when extracting characteristics of image, the mode employing multi-level features fusion obtains characteristics of image to the application.Specifically
Neutral net shallow-layer characteristics of image and deep layer characteristics of image are carried out fusion treatment as final characteristics of image.Figure after fusion
As feature can preferably improve the accuracy rate of detection, and the identification classification to wisp has a clear superiority.As Fig. 2 institute
Show, process schematic diagram for the image co-registration that the application provides.Wherein, conv1 represents neutral net ground floor, and conv2 represents god
Through the network second layer, conv3 represents neutral net third layer, and conv4 represents the 4th layer of neutral net, and conv5 represents nerve net
Network layer 5.It is assumed that the number of plies of depth convolution multilayer neural network is 5 layers during concrete process, extract this 5 layers image respectively special
Levy, then fusion treatment is carried out to the 5 tomographic image features extracted, obtain the characteristics of image after merging as the final image extracting
Feature.
S103, is carried out using the described image feature extracted, based on depth convolution multilayer neural network target detection model
Target detection, obtains candidate target.
When implementing, the application has pre-build depth convolution multilayer neural network target detection model and target classification
Network model, for example, can obtain depth convolution multilayer neural network target detection model using samples pictures training and target is divided
Class network model.Illustrate, the training initial stage can be initialized from the good depth convolutional neural networks model of pre-training,
Then carry out small parameter perturbations (fine-tuning) using the X-ray sample image gathering in advance, generate target detection network respectively
Model and target classification network model.The good initial network model of described pre-training can be a kind of ZF network model (depth
Practise neural network model) or VGG (English full name is visual geometry group, a kind of deep learning nerve
Network model) network model.Using the X-ray sample image gathering in advance, initial training model is trained, you can with
To depth convolution multilayer neural network target detection model and target classification network model.In some embodiments, in order to carry
High algorithm performance, depth convolution multilayer neural network target detection network model and depth convolution multilayer neural network target classification
Network model carries out convolution feature and shares.That is, in this application, a feature extraction can only be carried out, extract
Feature is separately in target detection model and object-class model, as such, it is possible to improve the treatment effeciency of algorithm.
In some embodiments, described using extract described image feature, be based on depth convolution multilayer neural network
Target detection model carries out target detection, obtains candidate target and includes：Using the described image feature extracted, it is based on multiple depth
Convolution multilayer neural network target detection model carries out target detection, obtains multiple testing results；Merge the plurality of detection knot
Really, obtain final testing result as candidate target.Illustrate, when being trained to target detection model, can be with base
In different training samples, obtain multiple different depth convolution multilayer neural network target detection models, such as 3.Then,
Using the 3 depth convolution multilayer neural network target detection models training, target is detected, obtain 3 detection knots
Really.As shown in figure 3, the multi-model fusion treatment schematic diagram providing for the embodiment of the present application.For same width scanogram, permissible
Obtain the first testing result using target detection model 1 (Model 1), it is possible to use target detection model 2 (Model 2) obtains
Second testing result, it is possible to use target detection model 3 (Model 3) obtains the 3rd testing result.Then to 3 testing results
Carry out fusion treatment, obtain the result after merging as final output result.Wherein, described fusion is the plurality of detects knot
Really, obtain final testing result to include as candidate target：Based on confidence calculations result, merge multiple testing results to obtain
Obtain testing result finally.Illustrate, all corresponding confidence calculations result of each testing result is it is assumed that first is detected
Confidence level in result is 0.9, and the confidence level of second testing result is 0.8, and the confidence level of the 3rd testing result is 0.7,
It is then final testing result with confidence level highest testing result.It is of course also possible to be obtained after fusion using other modes
Testing result, here is not defined.
S104, using the described image feature extracted, is based on depth convolution multilayer neural network object-class model to institute
State candidate target to be identified, obtain image recognition result.
It is previously noted, it is possible to use samples pictures pre-build depth convolution multilayer neural network object-class model.So
Afterwards, using the described image feature extracted, it is based on depth convolution multilayer neural network object-class model to described candidate target
It is identified, obtain image recognition result.When implementing, the described image feature extracted is inputted described depth convolution multilamellar
Neutral net object-class model, you can to obtain to recognition result, described recognition result is used for the classification of marking articles, for example
Whether it is knife or rifle etc..
Referring to Fig. 4, the image-recognizing method flow chart providing for another embodiment of the application, methods described for example can be wrapped
Include：
S401, obtains scanogram.
S402, carries out pretreatment to described scanogram.
When implementing, can be that different classes of scanned item image setting is different based on scanned item classification results
Color.So, by Image semantic classification, Organic substance, inorganic matters are made a distinction, detecting for image object provides prior information,
Improve target detection accuracy.
S403, carries out feature extraction to described scanogram, obtains the characteristics of image extracting.
S404, using the described image feature extracted, is based on multiple depth convolution multilayer neural network target detection models
Carry out target detection, obtain candidate target.
S405, using the described image feature extracted, is based on multiple depth convolution multilayer neural network object-class models
Described candidate target is identified, obtains image recognition result.
S406, based on described image recognition result, judges whether dangerous materials or smuggled goods.If judging to exist,
Enter S409, export information.
S407, if not existing, described image recognition result and list of articles is compared, and obtains comparison result.
S408, exports described comparison result.
If described comparison result display described image recognition result is mated with described list of articles, terminate program.If institute
State comparison result display described image recognition result to mismatch with described list of articles, export warning message, enter artificial reinspection
Program.
S409, exports prompt messages.
The image-recognizing method that the embodiment of the present application provides, can carry out detection identification automatically to scanogram, can be right
All kinds of dangerous materials, contraband or smuggled goods are used for quickly detecting, thus effectively alleviating the workload of security staff, improving and disobeying
Prohibit the accuracy of Articles detecting, the commodity simultaneously passing through will identify that are mated with declaration data, it is possible to achieve clearance automatically,
It is effectively improved clearance speed.Additionally, the application passes through Image semantic classification, Organic substance, inorganic matters are made a distinction, for follow-up
Image object detection provides prior information, improves target detection accuracy.In addition, in feature extraction, using shallow-layer and deep layer
Feature Fusion, as final characteristics of image, can preferably improve the accuracy rate of detection, to the little thing occurring in scanogram
The identification classification of body has more preferable effect.Finally, the application carries out detection classification using the parallel means of multi-model to target,
And final result is provided by suitably rule, effectively improve accuracy rate.
It is more than the detailed description that the image-recognizing method that the embodiment of the present application is provided is carried out, below the application is carried
For pattern recognition device be described in detail.
A kind of pattern recognition device schematic diagram that Fig. 5 provides for the embodiment of the present application.
A kind of pattern recognition device 500, described device 500 includes：
Image collection module 501, for obtaining scanogram.
Characteristic extracting module 502, for described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Its
In, described to described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain depth convolution multilamellar
The characteristics of image of each level of neutral net, carries out fusion treatment to the characteristics of image of described each level, obtains the image after merging
Feature is as the characteristics of image extracting.
Module of target detection 503, for using the described image feature extracted, based on depth convolution multilayer neural network mesh
Mark detection model carries out target detection, obtains candidate target.
Target classification module 504, for using the described image feature extracted, based on depth convolution multilayer neural network mesh
Mark disaggregated model is identified to described candidate target, obtains image recognition result.
In some embodiments, described device also includes：Pretreatment module, for carrying out pre- place to described scanogram
Reason, is that different classes of scanned item image arranges different colors based on scanned item classification results.
In some embodiments, described pretreatment module specifically includes：Density acquiring unit, for obtaining scanned item
Atomic number, based on described atomic number obtain scanned item density；Taxon, for according to described scanned item
Density determines the classification of scanned item, obtains scanned item classification results；Color arranging unit, for based on described scanned item
Classification results are that different classes of scanned item image arranges different colors.
In some embodiments, described characteristic extracting module determines mesh specifically for the color characteristic based on scanned item
Mark candidate region；Carry out feature extraction process in described object candidate area, obtain the characteristics of image extracting.
In some embodiments, described module of target detection specifically includes：Multi-model detector unit, for using extraction
Described image feature, carry out target detection based on multiple depth convolution multilayer neural network target detection models, obtain multiple
Testing result；Result integrated unit, for merging the plurality of testing result, obtains final testing result as candidate's mesh
Mark.
In some embodiments, described result integrated unit is specifically for based on confidence calculations result, merging multiple
Testing result is to obtain final testing result.
In some embodiments, described device also includes：Judge module, for based on described image recognition result, sentencing
Break and whether there is dangerous materials or smuggled goods；First output module, if for judging dangerous product or smuggled goods, defeated
Go out information.
In some embodiments, described device also includes：Comparing module, for by described image recognition result and article
Inventory is compared, and obtains comparison result；Second output module, for exporting described comparison result.
The function of above-mentioned each module may correspond to the process step of the above-mentioned image-recognizing method of Fig. 1, Fig. 4 detailed description,
Repeat no more in this.
Referring to Fig. 6, it is the block diagram of the device of image recognition that another embodiment of the application provides.Including：At at least one
Reason device 601 (such as CPU), memorizer 602 and at least one communication bus 603, lead to for realizing the connection between these devices
Letter.Processor 601 is used for executing the executable module of storage in memorizer 602, such as computer program.Memorizer 602 may
Comprise high-speed random access memory (RAM：Random Access Memory) it is also possible to also include non-labile memorizer
(non-volatile memory), for example, at least one disk memory.One or more than one program storage are in memorizer
In, and be configured to execute one or more than one program bag by one or more than one processor 601 contain and be used for
Carry out the instruction of following operation：
Obtain scanogram；Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Using extract
Described image feature, the target detection that carried out based on depth convolution multilayer neural network target detection model, obtain candidate target；Profit
With the described image feature of extraction, based on depth convolution multilayer neural network object-class model, described candidate target is known
Not, obtain image recognition result；Wherein, described to described scanogram is carried out with feature extraction, obtain the characteristics of image extracting
It is specially：Obtain the characteristics of image of each level of depth convolution multilayer neural network, the characteristics of image of described each level is melted
Conjunction is processed, and obtains the characteristics of image after merging as the characteristics of image extracting.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Pretreatment is carried out to described scanogram, is different classes of scanned item image based on scanned item classification results
Different colors are set.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Obtain the atomic number of scanned item, obtain the density of scanned item based on described atomic number；Swept according to described
The density retouching article determines the classification of scanned item, obtains scanned item classification results；Based on described scanned item classification results
For the different color of different classes of scanned item image setting.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Object candidate area is determined based on the color characteristic of scanned item；Carry out feature to carry in described object candidate area
Take process, obtain the characteristics of image extracting.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Carry out mesh using the described image feature extracted, based on multiple depth convolution multilayer neural network target detection models
Mark detection, obtains multiple testing results；Merge the plurality of testing result, obtain final testing result as candidate target.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Based on confidence calculations result, merge multiple testing results to obtain final testing result.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Based on described image recognition result, judge whether dangerous materials or smuggled goods；If judging dangerous product
Or smuggled goods, exports information.
In some embodiments, processor 601 is specifically for executing one or more than one program bag containing use
In the instruction carrying out following operation：
Described image recognition result and list of articles are compared, obtains comparison result；Export described comparison result.
Professional should further appreciate that, each example describing in conjunction with the embodiments described herein
Module and algorithm steps, can be hard in order to clearly demonstrate with electronic hardware, computer software or the two be implemented in combination in
Part and the interchangeability of software, generally describe composition and the step of each example in the above description according to function.
These functions to be executed with hardware or software mode actually, the application-specific depending on technical scheme and design constraint.
Professional and technical personnel can use different methods to each specific application realize described function, but this realization
It is not considered that exceeding scope of the present application.
The step of the method in conjunction with the embodiments described herein description or algorithm can be with hardware, computing device
Software module, or the combination of the two is implementing.Software module can be placed in random access memory (RAM), internal memory, read only memory
(ROM), electrically programmable ROM, electrically erasable ROM, depositor, hard disk, moveable magnetic disc, CD-ROM or technical field
In interior known any other form of storage medium.
Above-described specific embodiment, has been carried out to the purpose of the application, technical scheme and beneficial effect further
Describe in detail, be should be understood that the specific embodiment that the foregoing is only the application, be not used to limit the application
Protection domain, all any modification, equivalent substitution and improvement within spirit herein and principle, done etc., all should comprise
Within the protection domain of the application.
Claims (10)
1. a kind of image-recognizing method is it is characterised in that methods described includes：
Obtain scanogram；
Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；
Carry out target detection using the described image feature extracted, based on depth convolution multilayer neural network target detection model,
Obtain candidate target；
Using the described image feature extracted, it is based on depth convolution multilayer neural network object-class model to described candidate target
It is identified, obtain image recognition result；
Wherein, described described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain depth convolution
The characteristics of image of each level of multilayer neural network, carries out fusion treatment to the characteristics of image of described each level, obtains after merging
Characteristics of image is as the characteristics of image extracting.
2. method according to claim 1 is it is characterised in that before carrying out feature extraction to described scanogram, described
Method also includes：
Pretreatment is carried out to described scanogram, is different classes of scanned item image setting based on scanned item classification results
Different colors.
3. method according to claim 2 is it is characterised in that described carry out pretreatment to described scanogram, based on sweeping
Retouch the different color of the scanned item image setting that taxonomy of goods result is different classes of to include：
Obtain the atomic number of scanned item, obtain the density of scanned item based on described atomic number；
Determine the classification of scanned item according to the density of described scanned item, obtain scanned item classification results；
It is that different classes of scanned item image arranges different colors based on described scanned item classification results.
4., according to the method in claim 2 or 3 it is characterised in that described carry out feature extraction to described scanogram, obtain
The characteristics of image that must extract includes：
Object candidate area is determined based on the color characteristic of scanned item；
Carry out feature extraction process in described object candidate area, obtain the characteristics of image extracting.
5. method according to claim 1 is it is characterised in that described utilize the described image feature extracted, be based on depth
Convolution multilayer neural network target detection model carries out target detection, obtains candidate target and includes：
Carry out target inspection using the described image feature extracted, based on multiple depth convolution multilayer neural network target detection models
Survey, obtain multiple testing results；
Merge the plurality of testing result, obtain final testing result as candidate target.
6. method according to claim 5, it is characterised in that the plurality of testing result of described fusion, obtains final
Testing result includes as candidate target：
Based on confidence calculations result, merge multiple testing results to obtain final testing result.
7. method according to claim 1 is it is characterised in that methods described also includes：
Based on described image recognition result, judge whether dangerous materials or smuggled goods；
If judging dangerous product or smuggled goods, export information.
8. the method according to claim 1 or 6 is it is characterised in that methods described also includes：
Described image recognition result and list of articles are compared, obtains comparison result；
Export described comparison result.
9. a kind of pattern recognition device is it is characterised in that described device includes：
Image collection module, for obtaining scanogram；
Characteristic extracting module, for described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Wherein, described
To described scanogram is carried out with feature extraction, obtain the characteristics of image extracting and be specially：Obtain depth convolution multilamellar nerve net
The characteristics of image of each level of network, carries out fusion treatment to the characteristics of image of described each level, obtains the characteristics of image after merging and makees
For the characteristics of image extracting；
Module of target detection, for using the described image feature extracted, based on depth convolution multilayer neural network target detection
Model carries out target detection, obtains candidate target；
Target classification module, for using the described image feature extracted, based on depth convolution multilayer neural network target classification
Model is identified to described candidate target, obtains image recognition result.
10. a kind of device for image recognition, includes memorizer, and one or more than one program, and wherein one
Individual or more than one program storage is in memorizer, and is configured to one described in one or more than one computing device
Individual or more than one program bag contains the instruction for carrying out following operation：
Obtain scanogram；
Described scanogram is carried out with feature extraction, obtains the characteristics of image extracting；Wherein, described described scanogram is carried out
Feature extraction, obtains the characteristics of image extracting and is specially：Obtain the characteristics of image of each level of depth convolution multilayer neural network, right
The characteristics of image of described each level carries out fusion treatment, obtains the characteristics of image after merging as the characteristics of image extracting；
Carry out target detection using the described image feature extracted, based on depth convolution multilayer neural network target detection model,
Obtain candidate target；
Using the described image feature extracted, it is based on depth convolution multilayer neural network object-class model to described candidate target
It is identified, obtain image recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610854506.8A CN106485268B (en) | 2016-09-27 | 2016-09-27 | Image identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610854506.8A CN106485268B (en) | 2016-09-27 | 2016-09-27 | Image identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106485268A true CN106485268A (en) | 2017-03-08 |
| CN106485268B CN106485268B (en) | 2020-01-21 |
Family
ID=58268114
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610854506.8A Active CN106485268B (en) | 2016-09-27 | 2016-09-27 | Image identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106485268B (en) |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106960186A (en) * | 2017-03-17 | 2017-07-18 | 王宇宁 | Ammunition recognition methods based on depth convolutional neural networks |
| CN107273936A (en) * | 2017-07-07 | 2017-10-20 | 广东工业大学 | A kind of GAN image processing methods and system |
| CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图（厦门）科技有限公司 | The method and device of Face datection |
| CN107463965A (en) * | 2017-08-16 | 2017-12-12 | 湖州易有科技有限公司 | Fabric attribute picture collection and recognition methods and identifying system based on deep learning |
| CN107563290A (en) * | 2017-08-01 | 2018-01-09 | 中国农业大学 | A kind of pedestrian detection method and device based on image |
| CN107909093A (en) * | 2017-10-27 | 2018-04-13 | 浙江大华技术股份有限公司 | A kind of method and apparatus of Articles detecting |
| CN108229523A (en) * | 2017-04-13 | 2018-06-29 | 深圳市商汤科技有限公司 | Image detection, neural network training method, device and electronic equipment |
| CN108510116A (en) * | 2018-03-29 | 2018-09-07 | 哈尔滨工业大学 | A kind of luggage space planning system based on mobile terminal |
| CN108647559A (en) * | 2018-03-21 | 2018-10-12 | 四川弘和通讯有限公司 | A kind of danger recognition methods based on deep learning |
| CN109001833A (en) * | 2018-06-22 | 2018-12-14 | 天和防务技术（北京）有限公司 | A kind of Terahertz hazardous material detection method based on deep learning |
| CN109034245A (en) * | 2018-07-27 | 2018-12-18 | 燕山大学 | A kind of object detection method merged using characteristic pattern |
| CN109557114A (en) * | 2017-09-25 | 2019-04-02 | 清华大学 | Inspection method and inspection equipment and computer-readable medium |
| CN109583266A (en) * | 2017-09-28 | 2019-04-05 | 杭州海康威视数字技术股份有限公司 | A kind of object detection method, device, computer equipment and storage medium |
| WO2019096181A1 (en) * | 2017-11-14 | 2019-05-23 | 深圳码隆科技有限公司 | Detection method, apparatus and system for security inspection, and electronic device |
| CN109816037A (en) * | 2019-01-31 | 2019-05-28 | 北京字节跳动网络技术有限公司 | The method and apparatus for extracting the characteristic pattern of image |
| WO2019154383A1 (en) * | 2018-02-06 | 2019-08-15 | 同方威视技术股份有限公司 | Tool detection method and device |
| CN110222641A (en) * | 2019-06-06 | 2019-09-10 | 北京百度网讯科技有限公司 | The method and apparatus of image for identification |
| CN110245564A (en) * | 2019-05-14 | 2019-09-17 | 平安科技（深圳）有限公司 | A kind of pedestrian detection method, system and terminal device |
| CN110459225A (en) * | 2019-08-14 | 2019-11-15 | 南京邮电大学 | A kind of speaker identification system based on CNN fusion feature |
| CN110781911A (en) * | 2019-08-15 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Image matching method, device, equipment and storage medium |
| CN110796127A (en) * | 2020-01-06 | 2020-02-14 | 四川通信科研规划设计有限责任公司 | Embryo prokaryotic detection system based on occlusion sensing, storage medium and terminal |
| CN110909604A (en) * | 2019-10-23 | 2020-03-24 | 深圳市华讯方舟太赫兹科技有限公司 | Security image detection method, terminal device and computer storage medium |
| CN110942453A (en) * | 2019-11-21 | 2020-03-31 | 山东众阳健康科技集团有限公司 | CT image lung lobe identification method based on neural network |
| CN111103629A (en) * | 2018-10-25 | 2020-05-05 | 杭州海康威视数字技术股份有限公司 | Target detection method and device, NVR (network video recorder) equipment and security check system |
| CN111241893A (en) * | 2018-11-29 | 2020-06-05 | 阿里巴巴集团控股有限公司 | Identification recognition method, device and system |
| CN111340775A (en) * | 2020-02-25 | 2020-06-26 | 湖南大学 | Parallel method and device for acquiring ultrasonic standard tangent plane and computer equipment |
| WO2020134848A1 (en) * | 2018-12-28 | 2020-07-02 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detection method and device applied to millimeter wave security check instrument, and storage device |
| WO2020173021A1 (en) * | 2019-02-25 | 2020-09-03 | 平安科技（深圳）有限公司 | Artificial intelligence-based forbidden object identification method, apparatus and device, and storage medium |
| CN111856445A (en) * | 2019-04-11 | 2020-10-30 | 杭州海康威视数字技术股份有限公司 | Target detection method, device, equipment and system |
| CN112185077A (en) * | 2019-07-01 | 2021-01-05 | 云丁网络技术(北京)有限公司 | Intelligent reminding method, device and system and camera equipment |
| CN112215095A (en) * | 2020-09-24 | 2021-01-12 | 西北工业大学 | Contraband detection method, device, processor and security inspection system |
| CN112730468A (en) * | 2019-10-28 | 2021-04-30 | 同方威视技术股份有限公司 | Article detection device and method for detecting article |
| CN114549900A (en) * | 2022-02-23 | 2022-05-27 | 智慧航安（北京）科技有限公司 | Article classification method, device and system |
| CN114693612A (en) * | 2022-03-16 | 2022-07-01 | 深圳大学 | Knee joint bone tumor detection method based on deep learning and related device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102013098A (en) * | 2010-10-11 | 2011-04-13 | 公安部第一研究所 | Method for removing organic and inorganic matters from security inspection images |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105320945A (en) * | 2015-10-30 | 2016-02-10 | 小米科技有限责任公司 | Image classification method and apparatus |
| CN105631482A (en) * | 2016-03-03 | 2016-06-01 | 中国民航大学 | Convolutional neural network model-based dangerous object image classification method |
| CN105740758A (en) * | 2015-12-31 | 2016-07-06 | 上海极链网络科技有限公司 | Internet video face recognition method based on deep learning |
| CN105809164A (en) * | 2016-03-11 | 2016-07-27 | 北京旷视科技有限公司 | Character identification method and device |
-
2016
- 2016-09-27 CN CN201610854506.8A patent/CN106485268B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102013098A (en) * | 2010-10-11 | 2011-04-13 | 公安部第一研究所 | Method for removing organic and inorganic matters from security inspection images |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105320945A (en) * | 2015-10-30 | 2016-02-10 | 小米科技有限责任公司 | Image classification method and apparatus |
| CN105740758A (en) * | 2015-12-31 | 2016-07-06 | 上海极链网络科技有限公司 | Internet video face recognition method based on deep learning |
| CN105631482A (en) * | 2016-03-03 | 2016-06-01 | 中国民航大学 | Convolutional neural network model-based dangerous object image classification method |
| CN105809164A (en) * | 2016-03-11 | 2016-07-27 | 北京旷视科技有限公司 | Character identification method and device |
Non-Patent Citations (4)
| Title |
|---|
| SHENXIAOLU1984: ""目标检测 Faster RCNN算法详解"", 《CSDN 博客》 * |
| 卢宏涛等: ""深度卷积神经网络在计算机视觉中的应用研究综述"", 《数据采集与处理》 * |
| 张艳珠等: ""基于物体材质的X射线安检图像分割算法"", 《装备制造技术》 * |
| 雷青等: ""基于深度学习的安卓APP视频枪支检测技术研究"", 《信息网络安全》 * |
Cited By (45)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106960186A (en) * | 2017-03-17 | 2017-07-18 | 王宇宁 | Ammunition recognition methods based on depth convolutional neural networks |
| CN108229523A (en) * | 2017-04-13 | 2018-06-29 | 深圳市商汤科技有限公司 | Image detection, neural network training method, device and electronic equipment |
| CN107273936A (en) * | 2017-07-07 | 2017-10-20 | 广东工业大学 | A kind of GAN image processing methods and system |
| CN107273936B (en) * | 2017-07-07 | 2020-09-11 | 广东工业大学 | GAN image processing method and system |
| CN107563290A (en) * | 2017-08-01 | 2018-01-09 | 中国农业大学 | A kind of pedestrian detection method and device based on image |
| CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图（厦门）科技有限公司 | The method and device of Face datection |
| CN107463965A (en) * | 2017-08-16 | 2017-12-12 | 湖州易有科技有限公司 | Fabric attribute picture collection and recognition methods and identifying system based on deep learning |
| CN107463965B (en) * | 2017-08-16 | 2024-03-26 | 湖州易有科技有限公司 | Deep learning-based fabric attribute picture acquisition and recognition method and recognition system |
| CN109557114A (en) * | 2017-09-25 | 2019-04-02 | 清华大学 | Inspection method and inspection equipment and computer-readable medium |
| CN109557114B (en) * | 2017-09-25 | 2021-07-16 | 清华大学 | Inspection method and inspection apparatus, and computer-readable medium |
| CN109583266A (en) * | 2017-09-28 | 2019-04-05 | 杭州海康威视数字技术股份有限公司 | A kind of object detection method, device, computer equipment and storage medium |
| CN107909093B (en) * | 2017-10-27 | 2021-02-02 | 浙江大华技术股份有限公司 | Method and equipment for detecting articles |
| CN107909093A (en) * | 2017-10-27 | 2018-04-13 | 浙江大华技术股份有限公司 | A kind of method and apparatus of Articles detecting |
| WO2019096181A1 (en) * | 2017-11-14 | 2019-05-23 | 深圳码隆科技有限公司 | Detection method, apparatus and system for security inspection, and electronic device |
| WO2019154383A1 (en) * | 2018-02-06 | 2019-08-15 | 同方威视技术股份有限公司 | Tool detection method and device |
| CN108647559A (en) * | 2018-03-21 | 2018-10-12 | 四川弘和通讯有限公司 | A kind of danger recognition methods based on deep learning |
| CN108510116A (en) * | 2018-03-29 | 2018-09-07 | 哈尔滨工业大学 | A kind of luggage space planning system based on mobile terminal |
| CN108510116B (en) * | 2018-03-29 | 2020-06-30 | 哈尔滨工业大学 | Case and bag space planning system based on mobile terminal |
| CN109001833A (en) * | 2018-06-22 | 2018-12-14 | 天和防务技术（北京）有限公司 | A kind of Terahertz hazardous material detection method based on deep learning |
| CN109034245A (en) * | 2018-07-27 | 2018-12-18 | 燕山大学 | A kind of object detection method merged using characteristic pattern |
| CN111103629A (en) * | 2018-10-25 | 2020-05-05 | 杭州海康威视数字技术股份有限公司 | Target detection method and device, NVR (network video recorder) equipment and security check system |
| CN111241893A (en) * | 2018-11-29 | 2020-06-05 | 阿里巴巴集团控股有限公司 | Identification recognition method, device and system |
| CN111241893B (en) * | 2018-11-29 | 2023-06-16 | 阿里巴巴集团控股有限公司 | Identification recognition method, device and system |
| WO2020134848A1 (en) * | 2018-12-28 | 2020-07-02 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detection method and device applied to millimeter wave security check instrument, and storage device |
| CN109816037A (en) * | 2019-01-31 | 2019-05-28 | 北京字节跳动网络技术有限公司 | The method and apparatus for extracting the characteristic pattern of image |
| WO2020173021A1 (en) * | 2019-02-25 | 2020-09-03 | 平安科技（深圳）有限公司 | Artificial intelligence-based forbidden object identification method, apparatus and device, and storage medium |
| CN111856445A (en) * | 2019-04-11 | 2020-10-30 | 杭州海康威视数字技术股份有限公司 | Target detection method, device, equipment and system |
| CN110245564A (en) * | 2019-05-14 | 2019-09-17 | 平安科技（深圳）有限公司 | A kind of pedestrian detection method, system and terminal device |
| CN110245564B (en) * | 2019-05-14 | 2024-07-09 | 平安科技（深圳）有限公司 | Pedestrian detection method, system and terminal equipment |
| CN110222641B (en) * | 2019-06-06 | 2022-04-19 | 北京百度网讯科技有限公司 | Method and apparatus for recognizing image |
| CN110222641A (en) * | 2019-06-06 | 2019-09-10 | 北京百度网讯科技有限公司 | The method and apparatus of image for identification |
| CN112185077A (en) * | 2019-07-01 | 2021-01-05 | 云丁网络技术(北京)有限公司 | Intelligent reminding method, device and system and camera equipment |
| CN110459225A (en) * | 2019-08-14 | 2019-11-15 | 南京邮电大学 | A kind of speaker identification system based on CNN fusion feature |
| CN110781911B (en) * | 2019-08-15 | 2022-08-19 | 腾讯科技（深圳）有限公司 | Image matching method, device, equipment and storage medium |
| CN110781911A (en) * | 2019-08-15 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Image matching method, device, equipment and storage medium |
| CN110909604B (en) * | 2019-10-23 | 2024-04-19 | 深圳市重投华讯太赫兹科技有限公司 | Security check image detection method, terminal equipment and computer storage medium |
| CN110909604A (en) * | 2019-10-23 | 2020-03-24 | 深圳市华讯方舟太赫兹科技有限公司 | Security image detection method, terminal device and computer storage medium |
| CN112730468A (en) * | 2019-10-28 | 2021-04-30 | 同方威视技术股份有限公司 | Article detection device and method for detecting article |
| CN110942453A (en) * | 2019-11-21 | 2020-03-31 | 山东众阳健康科技集团有限公司 | CT image lung lobe identification method based on neural network |
| CN110796127A (en) * | 2020-01-06 | 2020-02-14 | 四川通信科研规划设计有限责任公司 | Embryo prokaryotic detection system based on occlusion sensing, storage medium and terminal |
| CN111340775B (en) * | 2020-02-25 | 2023-09-29 | 湖南大学 | Parallel method, device and computer equipment for acquiring ultrasonic standard section |
| CN111340775A (en) * | 2020-02-25 | 2020-06-26 | 湖南大学 | Parallel method and device for acquiring ultrasonic standard tangent plane and computer equipment |
| CN112215095A (en) * | 2020-09-24 | 2021-01-12 | 西北工业大学 | Contraband detection method, device, processor and security inspection system |
| CN114549900A (en) * | 2022-02-23 | 2022-05-27 | 智慧航安（北京）科技有限公司 | Article classification method, device and system |
| CN114693612A (en) * | 2022-03-16 | 2022-07-01 | 深圳大学 | Knee joint bone tumor detection method based on deep learning and related device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106485268B (en) | 2020-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106485268A (en) | A kind of image-recognizing method and device | |
| CN108154168B (en) | Comprehensive cargo inspection system and method | |
| US20230162342A1 (en) | Image sample generating method and system, and target detection method | |
| US12067760B2 (en) | Systems and methods for image processing | |
| CN104636707B (en) | The method of automatic detection cigarette | |
| CN109902643A (en) | Intelligent safety inspection method, device, system and its electronic equipment based on deep learning | |
| CN104751163B (en) | The fluoroscopic examination system and method for automatic Classification and Identification are carried out to cargo | |
| CN104165896B (en) | Liquid goods safety inspection method and device | |
| US10042079B2 (en) | Image-based object detection and feature extraction from a reconstructed charged particle image of a volume of interest | |
| US10436932B2 (en) | Inspection systems for quarantine and methods thereof | |
| CN108664971A (en) | Pulmonary nodule detection method based on 2D convolutional neural networks | |
| CN107209944A (en) | The correction of beam hardening pseudomorphism in the sample microtomography being imaged in a reservoir | |
| Rogers et al. | Threat Image Projection (TIP) into X-ray images of cargo containers for training humans and machines | |
| CN110488368A (en) | A kind of contraband recognition methods and device based on dual intensity X-ray screening machine | |
| CN110186940A (en) | Safety check recognition methods, device, computer equipment and storage medium | |
| CN106651841B (en) | Analysis method for security inspection image complexity | |
| DE102014205447A1 (en) | Detection of objects in an object | |
| US20080253653A1 (en) | Systems and methods for improving visibility of scanned images | |
| CN102608135B (en) | Method and equipment for confirming CT (Computed Tomography) scanning position in dangerous goods inspection system | |
| Visser et al. | Automated comparison of X-ray images for cargo scanning | |
| CN116129153A (en) | Intelligent analysis processing method, system, device, processor and computer readable storage medium for forbidden band article identification | |
| CN111539251B (en) | Security check article identification method and system based on deep learning | |
| Thawornwong et al. | Lumber value differences from reduced CT spatial resolution and simulated log sawing | |
| CN114676759A (en) | Method and device for detecting contraband in security inspection image | |
| Gu | Research and implementation of automatic cutlery recognition method based on X-ray security inspection image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |