CN108648204A - A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield - Google Patents
A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield Download PDFInfo
- Publication number
- CN108648204A CN108648204A CN201810371829.0A CN201810371829A CN108648204A CN 108648204 A CN108648204 A CN 108648204A CN 201810371829 A CN201810371829 A CN 201810371829A CN 108648204 A CN108648204 A CN 108648204A
- Authority
- CN
- China
- Prior art keywords
- safety check
- fluoroscopy images
- region
- image
- ray safety
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 26
- 238000002594 fluoroscopy Methods 0.000 claims abstract description 120
- 238000003708 edge detection Methods 0.000 claims abstract description 35
- 238000003062 neural network model Methods 0.000 claims abstract description 31
- 239000000284 extract Substances 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 11
- 238000007689 inspection Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 210000000746 body region Anatomy 0.000 claims 1
- 210000002569 neuron Anatomy 0.000 description 10
- 239000002184 metal Substances 0.000 description 7
- 230000002708 enhancing effect Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 238000010009 beating Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000032258 transport Effects 0.000 description 1
- 230000002747 voluntary effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The embodiment of the present application provides a kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield, including：Obtain the X-ray safety check fluoroscopy images of target body；Edge detection is carried out to X-ray safety check fluoroscopy images, extracts the image-region surrounded by closed edge；Image-region is identified using advance trained neural network model, judges whether image-region is sensitive body position；If image-region is for sensitive body position, the image-region in X-ray safety check fluoroscopy images is stamped into mosaic, the X-ray safety check fluoroscopy images after mosaic are stamped in display.The method and apparatus of the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application; the sensitive body position in the X-ray safety check fluoroscopy images of human body is identified by neural network model; and it will identify that mosaic is stamped at the sensitive body position come; avoid the sensitive body position for being shown in X-ray safety check fluoroscopy images and being detected human body; tested person's privacy is protected, safety check speed is improved.
Description
Technical field
This application involves image technique field more particularly to a kind of sides of human body safety check that realizing artificial intelligence regions shield
Method and device.
Background technology
The public transports such as subway, aircraft, railway are being taken, into public domains such as court, museums, come in and go out border
And when participating in some important public activities, safety inspection is carried out to human body itself and is all necessary, prevents someone from carrying
Weapon, combustible and explosive articles, tool used in crime, drugs, a large sum of prohibited items such as foreign exchange or die Konterbande.To the safety inspection hand of human body
Section mainly uses metal detector or metal detection door, gently pats clothing with police dog sniff and safety inspector and checks.But
Or being that these means are slipped through the net or false alarm possibility is larger or spends the time more, traffic efficiency is influenced, causes personnel stagnant
Stay congestion.
Perspective safety check speed is fast, personnel will not be caused to be detained, and can intuitively show that human body is portable various
The shape of foreign object, accuracy rate is high, and its dose of radiation is completely harmless to health.But perspective safety check also can directly be shown
The body part blocked by clothing is shown, this constitutes certain offend to the privacy and izzat of those who are investigated.In addition, X
X ray fluoroscopy x image is gray-scale map, and canescence is presented in the body of people, and grey black is presented in metal, and the two contrast is bigger, safety check
Member is still easy observation；But canescence is also presented in some nonmetallic carry-on articles, safety check little with the body colour contrast of people
Member's observation is just very laborious, if flow of the people is big, passage rate fast, safety inspector's continuous observation again, and not only eye fatigue, and hold very much
It easily misunderstands or leaks and see.
Invention content
In view of this, the purpose of the application is to propose a kind of side of specific objective identification and enhancing for safety check perspective
Method and system, come solve in the prior art, perspective safety check be easy exposure privacy places of human body, while pair with human body colour contrast compared with
The slow technical problem of small object identification.
Based on above-mentioned purpose, present applicant proposes a kind of method of human body safety check that realizing artificial intelligence regions shield, packets
It includes：
Obtain the X-ray safety check fluoroscopy images of target body；
Edge detection is carried out to the X-ray safety check fluoroscopy images, extracts the image-region surrounded by closed edge；
Described image region is identified using advance trained neural network model, judges that described image region is
No is sensitive body position；
If described image region is for sensitive body position, by the described image region in the X-ray safety check fluoroscopy images
Mosaic is stamped, and shows and stamps the X-ray safety check fluoroscopy images after mosaic.
In some embodiments, edge detection is carried out to the X-ray safety check fluoroscopy images described, extracted by closed edge
Before the image-region of encirclement, the method further includes：
Image enhancement, filtering removal noise jamming are carried out to the X-ray safety check fluoroscopy images, and remove artifact, then again
Enhance contrast, enhances the contrast in the X-ray safety check fluoroscopy images difference greyscale color region.
In some embodiments, described that edge detection is carried out to the X-ray safety check fluoroscopy images, it extracts by closed edge packet
The image-region enclosed specifically includes：
Edge detection is carried out to the X-ray safety check fluoroscopy images using canny edge detection operators, is extracted by closed edge
The image-region of encirclement.
In some embodiments, described that edge is carried out to the X-ray safety check fluoroscopy images using canny edge detection operators
The image-region surrounded by closed edge is extracted in detection, including：
Convolution is made to X-ray safety check fluoroscopy images and Gauss mask, the X-ray safety check fluoroscopy images are smoothed；
The gradient of each pixel of the X-ray safety check fluoroscopy images after smoothing processing is calculated using Sobel operators；
Retain the maximum of gradient intensity on each pixel of the X-ray safety check fluoroscopy images, deletes other values；
Set the threshold value upper bound of the maximum of gradient intensity and threshold value on each pixel of the X-ray safety check fluoroscopy images
Lower bound, the pixel that the maximum of gradient intensity is more than to the threshold value upper bound is confirmed as boundary, by the maximum of gradient intensity
The pixel for being less than the threshold value upper bound more than the threshold value lower bound is confirmed as weak boundary, and the maximum of gradient intensity is less than institute
The pixel for stating threshold value lower bound is confirmed as non-boundary；
The weak boundary being connected with the boundary is confirmed into boundary, if non-boundary is confirmed as on others boundary.
In some embodiments, described that described image region is known using advance trained neural network model
Not, judge whether described image region is sensitive body position, including：
According to the gray value of the pixel in described image region, angular second moment battle array, correlation, difference matrix, inverse differential are asked
Matrix and pixel entropy；
By the angular second moment battle array, correlation, difference matrix, contrast sub-matrix and pixel entropy composition described image region
Feature vector；
Described eigenvector is identified using neural network model trained in advance, whether judges described image region
For sensitive body position.
In some embodiments, the neural network model includes five input layers, four hidden layers and an output layer.
In some embodiments, the neural network model is the neural network model using back-propagation algorithm.
Based on above-mentioned purpose, the application also proposed a kind of device of human body safety check that realizing artificial intelligence regions shield,
Including：
Image collection module, the X-ray safety check fluoroscopy images for obtaining target body；
Image processing module, for carrying out edge detection to the X-ray safety check fluoroscopy images, extraction is surrounded by closed edge
Image-region；
Images match module, for matching described image region with pre-stored sensitive body's position template；
Image display stamps the described image region in the X-ray safety check fluoroscopy images if being used for successful match
Mosaic, and show and stamp the X-ray safety check fluoroscopy images after mosaic.
In some embodiments, further include：
Image pre-processing module, for carrying out image enhancement to the X-ray safety check fluoroscopy images, filtering removal noise is dry
It disturbs, and removes artifact, then enhance contrast again, enhance the contrast in the X-ray safety check fluoroscopy images difference greyscale color region.
In some embodiments, further include：
Memory module, for storing neural network model trained in advance.
The embodiment of the present application provides a kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield, wherein just
Method includes：Obtain the X-ray safety check fluoroscopy images of target body；To the X-ray safety check fluoroscopy images carry out edge detection, extraction by
The image-region that closed edge is surrounded；Described image region is identified using advance trained neural network model, is sentenced
Whether disconnected described image region is sensitive body position；If described image region is for sensitive body position, by the X-ray safety check
Mosaic is stamped in described image region in fluoroscopy images, and is shown and stamped the X-ray safety check fluoroscopy images after mosaic.The application
The method and apparatus of the human body safety check of the realization artificial intelligence regions shield of embodiment, by neural network model to the X of human body
Sensitive body position in light safety check fluoroscopy images is identified, and will identify that mosaic is stamped at the sensitive body position come,
The sensitive body position for being shown in X-ray safety check fluoroscopy images and being detected human body is avoided, tested person's privacy is protected, improves peace
Examine speed.
Description of the drawings
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon：
Fig. 1 is the flow chart of the method for the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application one；
Fig. 2 is the flow chart of the method for the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application two；
Fig. 3 is that the use canny edge detection operators of the embodiment of the present application three carry out side to the X-ray safety check fluoroscopy images
The flow chart of edge detection；
Fig. 4 is that the structure of the device of the human body safety check for realizing artificial intelligence regions shield of the embodiment of the present application four is shown
It is intended to；
Fig. 5 is that the structure of the device of the human body safety check for realizing artificial intelligence regions shield of the embodiment of the present application five is shown
It is intended to.
Specific implementation mode
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, is illustrated only in attached drawing and invent relevant part with related.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
The low dosage X-ray machine that perspective safety check is carried out specifically for human body has already been developed out, at home and abroad also once
Trial was used.Perspective safety check speed is fast, personnel will not be caused to be detained, and can intuitively show that human body is portable each
The drugs etc. carried secretly in the shape or even disjunctor of kind foreign object can find that accuracy rate is high, and its dose of radiation is strong to human body
Health is completely harmless.
But perspective safety check can also directly display out the body privacy places blocked by clothing, although strictly being only limitted to pacify
Inspection person can just see, but this constitutes certain offend to the privacy and izzat of those who are investigated.Even if more open
American-European countries, perspective screening machine could only use under the premise of examinee is completely informed and voluntary.China is passed due to society
Overall view thought is denseer, and perspective safety check is less received by the public, and a small number of units were tried out, because dispute is too greatly soon
It removes.
In order to change the status quo, present applicant proposes a kind of methods of human body safety check that realizing artificial intelligence regions shield, such as
It is the flow chart of the method for the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application one, in this reality shown in Fig. 1
It applies in example, the method for the human body safety check for realizing artificial intelligence regions shield includes the following steps：
S101：Obtain the X-ray safety check fluoroscopy images of target body.
In the present embodiment, the X-ray safety check fluoroscopy images of target body can be obtained by the X-ray machine of low dosage, had
Body, X-ray machine can be arranged in the both sides of Mag ＆ Bag Entrance, to being scanned by the human body of Mag ＆ Bag Entrance, to target body
When (the specific single human body i.e. in the stream of people) is scanned, the X-ray safety check fluoroscopy images of the target body can be obtained, with right
The article that the target body is carried or held carries out safety inspection, prevents any contraband in the article carried, including
Dangerous goods and smuggled goods.The X-ray machine can be door case type, and the both sides of the X-ray machine of the door case type are provided with X-ray machine
Scanning light source, when have human body by when, the X-ray machine is to complete scanning to human body.Here it is only exemplary to X-ray machine
Using illustrating, and it is understood not to the restriction to technical scheme.
S102：Edge detection is carried out to the X-ray safety check fluoroscopy images, extracts the image-region surrounded by closed edge.
It in the present embodiment, can be right after getting the X-ray safety check fluoroscopy images of the target body using X-ray machine
The X-ray safety check fluoroscopy images carry out edge detection, and extract the figure surrounded by closed edge in the X-ray safety check fluoroscopy images
As region.When human body personal effects, article and human body will present out respective profile in X-ray safety check fluoroscopy images,
Since human body is different with the composition material of article, so the color showed in X-ray safety check fluoroscopy images also differs, due to
The contrast of color makes human body and article edge obvious, and human body and article can be obtained in X-ray by image processing techniques
Edge in safety check fluoroscopy images.Meanwhile the different parts of human body color in X-ray safety check fluoroscopy images is also not quite similar,
There is more apparent edges, therefore, the profile at each position of human body can be shown in X-ray safety check fluoroscopy images.
After identifying the edge in the X-ray safety check fluoroscopy images, the image-region surrounded by closed edge may further be obtained.
The image-region surrounded by closed edge is identified, can human body position or human body carry article into people know
Not.For example, when human body carries metal product, in X-ray safety check fluoroscopy images, canescence, metal is presented in the body of people
Grey black is presented can protrude the edge of metal since the two contrast is bigger, to the image-region surrounded by the edge
It is identified, can confirm which kind of article the metal is.
S103：Described image region is identified using advance trained neural network model, judges described image
Whether region is sensitive body position.
In the present embodiment, after the image-region that extraction is surrounded by closed edge, advance trained god can be utilized
The image-region surrounded by closed edge extracted is identified through network model, is surrounded by closed edge described in judgement
Image-region whether be human body sensitive part image.It specifically, can be by the image surrounded by closed edge
Input of the region as the neural network model, and the convolution algorithm of the hidden layer by the neural network model, output
As a result, and determining whether described image region is sensitive body position according to the result of output.
S104：If described image region is for sensitive body position, by the described image in the X-ray safety check fluoroscopy images
Mosaic is stamped in region, and is shown and stamped the X-ray safety check fluoroscopy images after mosaic.
In the present embodiment, described image region is identified when by neural network model trained in advance, and really
Described image region is determined for behind sensitive body position, i.e., the image-region may relate to the individual privacy of tested human body, then by institute
It states the described image region in X-ray safety check fluoroscopy images and stamps mosaic, and show and stamp the X-ray safety check perspective view after mosaic
Picture.The above process is all completed on backstage, and the X-ray safety check fluoroscopy images that security staff sees are the X after beating mosaic processing
Light safety check fluoroscopy images, the sensitive body position of personnel is detected in image, and stamp has been handled, is not displayed.
The method of the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application, passes through neural network model pair
Sensitive body position in the X-ray safety check fluoroscopy images of human body is identified, and will identify that the sensitive body position come is stamped
Mosaic avoids the sensitive body position for being shown in X-ray safety check fluoroscopy images and being detected human body, protects tested person's privacy,
Improve safety check speed.
As shown in Fig. 2, being the stream of the method for the human body safety check of the realization artificial intelligence regions shield of the embodiment of the present application two
Cheng Tu, the method for the human body safety check of the realization artificial intelligence regions shield of the present embodiment, includes the following steps：
S101：Obtain the X-ray safety check fluoroscopy images of target body.
S201：Image enhancement, filtering removal noise jamming are carried out to the X-ray safety check fluoroscopy images, and remove artifact, so
Enhance contrast again afterwards, enhances the contrast in the X-ray safety check fluoroscopy images difference greyscale color region.
In the present embodiment, first saturating to the X-ray safety check after the X-ray safety check fluoroscopy images for obtaining the target body
Visible image carries out image enhancement, removes the noise in the X-ray safety check fluoroscopy images progress image using low-pass filtering, and use
High-pass filtering method enhances the high-frequency signals such as edge so that the X-ray safety check fluoroscopy images become more fully apparent, in favor of subsequently walking
Edge detection in rapid.After carrying out image enhancement and removing noise, the X-ray safety check fluoroscopy images are removed at artifact
Reason, then enhances contrast, enhances the contrast in the X-ray safety check fluoroscopy images difference greyscale color region so that the X-ray again
The edge of safety check fluoroscopy images each section is more prominent, to be easier to be detected.
S102：Edge detection is carried out to the X-ray safety check fluoroscopy images, extracts the image-region surrounded by closed edge.
Edge detection is carried out to the X-ray safety check fluoroscopy images handled through step 201, due to pacifying to the X-ray in step 201
Inspection fluoroscopy images have carried out image enhancement, denoising, go artifact and have enhanced the processing of contrast so as to the X-ray in this step
Safety check fluoroscopy images carry out edge detection and detect that edge is more acurrate, i.e., the quantity at weak edge is less.
S103：Described image region is identified using advance trained neural network model, judges described image
Whether region is sensitive body position.
S104：If described image region is for sensitive body position, by the described image in the X-ray safety check fluoroscopy images
Mosaic is stamped in region, and is shown and stamped the X-ray safety check fluoroscopy images after mosaic.
In the present embodiment, the step identical as embodiment one is no longer described in detail here, and the present embodiment is used for safety check
The method of the specific objective identification and enhancing of perspective, by carrying out image enhancement, filtering removal to the X-ray safety check fluoroscopy images
Noise jamming, and artifact is removed, then enhance contrast again, enhances the X-ray safety check fluoroscopy images difference greyscale color region
Contrast so that edge in the X-ray safety check fluoroscopy images is more prominent, is conducive to the accuracy rate for improving edge detection, together
When improve safety check speed.
As shown in figure 3, be the embodiment of the present application three using canny edge detection operators to the X-ray safety check perspective view
Flow chart as carrying out edge detection, as the alternative embodiment of the application, in use canny edge detection operators to institute
During stating X-ray safety check fluoroscopy images progress edge detection, edge detection, extraction are carried out to the X-ray safety check fluoroscopy images
The image-region surrounded by closed edge, specifically includes following steps：
S301：Convolution first is made to X-ray safety check fluoroscopy images and Gauss mask, the X-ray safety check fluoroscopy images are put down
Sliding processing.
In order to reduce influence of the noise to edge detection results as far as possible, so having to filter out noise to prevent from being drawn by noise
The error detection risen.For smoothed image, convolution is carried out using Gaussian filter and image, the step is by smoothed image, to subtract
Apparent influence of noise on few edge detector.Size be (2k+1) x (2k+1) Gaussian filter core growth equation formula by
It is given below：
Here is a sigma=1.4, and size is the example (it should be noted that normalization) of the Gaussian convolution core of 3x3：
If the window of a 3x3 is A in image, the pixel to be filtered is e, then passes through after gaussian filtering, pixel e
Brightness value be：
Wherein * is convolution symbol, and all elements are added summation in sum representing matrixes.
S302：The ladder of each pixel of the X-ray safety check fluoroscopy images after smoothing processing is calculated using Sobel operators
Degree.
Edge in image can be directed toward all directions, therefore Canny algorithms are come using four operators in detection image
Horizontal, vertical and diagonal edge.The operator (such as Roberts, Prewitt, Sobel etc.) of edge detection returns to horizontal Gx and vertical
Thus the first derivative values in the directions Gy can determine the gradient G and direction theta of pixel.
Wherein G is gradient intensity, and theta indicates gradient direction, and arctan is arctan function.Below with Sobel operators
For teach how to calculate gradient intensity and direction.
The Sobel operators in the directions x and y are respectively：
Wherein Sx indicates the Sobel operators in the directions x, the edge for detecting the directions y；Sy indicates that the Sobel in the directions y is calculated
Son, the edge for detecting the directions x (edge direction and gradient direction are vertical).
If the window of a 3x3 is A in image, the pixel that calculate gradient is e, then carries out convolution with Sobel operators
Later, Grad of the pixel e in the directions x and y is respectively：
Wherein * is convolution symbol, and all elements are added summation in sum representing matrixes.It can be calculated according to formula (3-2)
Go out gradient and the direction of pixel e.
S303：The maximum for retaining gradient intensity on each pixel of the X-ray safety check fluoroscopy images, deletes other
Value.
Gradient is divided into 8 directions, respectively E, NE, N, NW, W, SW, S, SE, wherein 0 represents 00~45o, 1 represents 450
~90o, 2 represent -900~-45o, and 3 represent -450~0o.The gradient direction of pixel P is theta, then pixel P1 and P2
Gradient linearity interpolation is：
Tan (θ)=Gy/Gx
Gp1=(1-tan (θ)) × E+tan (θ) × NE
Gp2=(1-tan (θ)) × W+tan (θ) × SW
S304：Set the threshold value upper bound of the maximum of gradient intensity on each pixel of the X-ray safety check fluoroscopy images
With threshold value lower bound, the pixel that the maximum of gradient intensity is more than to the threshold value upper bound is confirmed as boundary, by gradient intensity
Maximum is more than the threshold value lower bound and is confirmed as weak boundary less than the pixel in the threshold value upper bound, by the maximum of gradient intensity
Pixel less than the threshold value lower bound is confirmed as non-boundary.
After applying non-maxima suppression, remaining pixel can more accurately indicate the actual edge in image.So
And, however it remains some edge pixels caused by noise and color change.In order to solve these spurious responses, it is necessary to weak
Grad filters edge pixel, and retains the edge pixel with high gradient value, can be realized by selecting high-low threshold value.Such as
The Grad of fruit edge pixel is higher than high threshold, then is marked as strong edge pixel；If the Grad of edge pixel is less than
High threshold and be more than Low threshold, then be marked as weak edge pixel；If the Grad of edge pixel is less than Low threshold,
It can be suppressed.
S305：The weak boundary being connected with the boundary is confirmed into boundary, if non-boundary is confirmed as on others boundary.
By the above method, it can realize and edge detection is carried out to the X-ray safety check fluoroscopy images.
It is described to utilize advance trained neural network model to described image as the alternative embodiment of the application
Region is identified, and judges whether described image region is sensitive body position, including：
According to the gray value of the pixel in described image region, angular second moment battle array, correlation, difference matrix, inverse differential are asked
Matrix and pixel entropy；
By the angular second moment battle array, correlation, difference matrix, contrast sub-matrix and pixel entropy composition described image region
Feature vector；
Described eigenvector is identified using neural network model trained in advance, whether judges described image region
For sensitive body position.
About the training and identification of neural network model, specifically, some are chosen first and carries human body sensitive part
Fluoroscopy images as sample, and manually draw a circle to approve the sensitive image region in these sample images, extract quick in sample image
The features above vector for feeling image-region, forms feature vector sample, whole feature vector samples is updated to quick as human body
There is N number of input neuron, hidden layer to have for the convolutional neural networks of sensillary area domain identification model, the convolutional neural networks input layer
K hidden neuron, output layer have M output neuron, then calculate hidden layer successively and the numerical value of output layer is as follows：
Wherein w1nkIt is the weight between n-th of neuron of input layer and k-th of neuron of hidden layer, O1pkIt is hidden
Hide the output of k-th of neuron of layer；w2kmIt is the weight between m-th of neuron of k-th of neuron of hidden layer and output layer,
O2pmIt is the output of m-th of output layer neuron, activation primitiveIndicate the i-th wheel training；
(3) implementation deviation calculates：Judge whether the deviation of epicycle (the i-th wheel) is less than or equal to
Scheduled tolerance ε if the determination result is YES then stops iteration, if judging result is no, continues following flow；
(4) backwards calculation is executed：
Wherein learning rate is μ,
-δpm(i)=(tpm-O2pm(i))O2pm(i)(1-O2pm(i)),
It is as follows to change weights：
w1nk(i+1)=w1nk(i)+Δw1nk(i+1)
w2km(i+1)=w2km(i)+Δw2km(i+1)
(5) (2) step is returned, the study of i+1 wheel is re-started.
By repetition learning, the weighted value between neuron is constantly adjusted, is permitted equal to scheduled until deviation is less than
Perhaps deviation ε, then convolutional neural networks training are completed.To for each figure extracted in the fluoroscopy images that currently acquire in real time
As the features described above vector in region, it can be input to trained neural network model as an input vector, by this
The output of model is as the judgement for whether belonging to sensitive body position to each image-region.
As the alternative embodiment of the application, the neural network model includes five input layers, four
Hidden layer neuron and an output layer neuron.The neural network model can be the nerve net using back-propagation algorithm
Network model.
As shown in figure 4, being knot of the embodiment of the present application for realizing the device of the human body safety check of artificial intelligence regions shield
Structure schematic diagram.The device of the human body safety check for realizing artificial intelligence regions shield in the present embodiment, including：
Image collection module 401, the X-ray safety check fluoroscopy images for obtaining target body, described image acquisition module 401
It can be X-ray machine, can also be with other devices with X-ray machine similar functions.
Image processing module 402 is extracted for carrying out edge detection to the X-ray safety check fluoroscopy images by closed edge
The image-region of encirclement.
Images match module 403 is used for described image region and the progress of pre-stored sensitive body's position template
Match.
Image display 404, if successful match is used for, by the described image region in the X-ray safety check fluoroscopy images
Mosaic is stamped, and shows and stamps the X-ray safety check fluoroscopy images after mosaic.
The device of the human body safety check for realizing artificial intelligence regions shield in the present embodiment can obtain and above-mentioned side
The similar technique effect of method embodiment, which is not described herein again.
As shown in figure 5, being the device of the human body safety check for realizing artificial intelligence regions shield of the embodiment of the present application five
Structural schematic diagram, the device of the human body safety check for realizing artificial intelligence regions shield of the present embodiment includes：
Image collection module 401, the X-ray safety check fluoroscopy images for obtaining target body, described image acquisition module 401
It can be X-ray machine, can also be with other devices with X-ray machine similar functions.
Image processing module 402 is extracted for carrying out edge detection to the X-ray safety check fluoroscopy images by closed edge
The image-region of encirclement.
Images match module 403 is used for described image region and the progress of pre-stored sensitive body's position template
Match.
In addition, further including image pre-processing module 501, for carrying out image enhancement, filter to the X-ray safety check fluoroscopy images
Wave removes noise jamming, and removes artifact, then enhances contrast again, enhances the X-ray safety check fluoroscopy images difference gray scale face
The contrast in color region.
Memory module 502, for storing neural network model trained in advance.
The system of the specific objective identification and enhancing for safety check perspective of the present embodiment, since described image pre-processes mould
Block 501 first carries out image increasing after the X-ray safety check fluoroscopy images for obtaining the target body to the X-ray safety check fluoroscopy images
By force, the noise in the X-ray safety check fluoroscopy images progress image is removed using low-pass filtering, and side is enhanced using high-pass filtering method
The high-frequency signals such as edge so that the X-ray safety check fluoroscopy images become more fully apparent, in favor of the edge detection in subsequent step.
It is carrying out image enhancement and after removing noise, artifact processing is removed to the X-ray safety check fluoroscopy images, then enhancing again pair
Than degree, enhance the contrast in the X-ray safety check fluoroscopy images difference greyscale color region so that the X-ray safety check fluoroscopy images are each
Partial edge is more prominent, to be easier to be detected.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.People in the art
Member should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Other technical solutions of arbitrary combination and formation.Such as features described above has similar work(with (but not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (10)
1. a kind of method of human body safety check that realizing artificial intelligence regions shield, which is characterized in that including：
Obtain the X-ray safety check fluoroscopy images of target body；
Edge detection is carried out to the X-ray safety check fluoroscopy images, extracts the image-region surrounded by closed edge；
Described image region is identified using advance trained neural network model, judge described image region whether be
Sensitive body position；
If described image region is stamped the described image region in the X-ray safety check fluoroscopy images for sensitive body position
Mosaic, and show and stamp the X-ray safety check fluoroscopy images after mosaic.
2. according to the method described in claim 1, it is characterized in that, carrying out edge to the X-ray safety check fluoroscopy images described
Detection, before extracting the image-region surrounded by closed edge, the method further includes：
Image enhancement, filtering removal noise jamming are carried out to the X-ray safety check fluoroscopy images, and remove artifact, is then enhanced again
Contrast enhances the contrast in the X-ray safety check fluoroscopy images difference greyscale color region.
3. according to the method described in claim 1, it is characterized in that, described carry out edge inspection to the X-ray safety check fluoroscopy images
It surveys, extracts the image-region surrounded by closed edge and specifically include：
Edge detection is carried out to the X-ray safety check fluoroscopy images using canny edge detection operators, extraction is surrounded by closed edge
Image-region.
4. according to the method described in claim 3, it is characterized in that, described pacify the X-ray using canny edge detection operators
It examines fluoroscopy images and carries out edge detection, extract the image-region surrounded by closed edge, including：
Convolution is made to X-ray safety check fluoroscopy images and Gauss mask, the X-ray safety check fluoroscopy images are smoothed；
The gradient of each pixel of the X-ray safety check fluoroscopy images after smoothing processing is calculated using Sobel operators；
Retain the maximum of gradient intensity on each pixel of the X-ray safety check fluoroscopy images, deletes other values；
It sets on each pixel of the X-ray safety check fluoroscopy images under the threshold value upper bound and threshold value of the maximum of gradient intensity
Boundary, the pixel that the maximum of gradient intensity is more than to the threshold value upper bound is confirmed as boundary, and the maximum of gradient intensity is big
The pixel for being less than the threshold value upper bound in the threshold value lower bound is confirmed as weak boundary, the maximum of gradient intensity is less than described
The pixel of threshold value lower bound is confirmed as non-boundary；
The weak boundary being connected with the boundary is confirmed into boundary, if non-boundary is confirmed as on others boundary.
5. according to the method described in claim 1, it is characterized in that, described utilize advance trained neural network model to institute
It states image-region to be identified, judges whether described image region is sensitive body position, including：
According to the gray value of the pixel in described image region, angular second moment battle array, correlation, difference matrix, contrast sub-matrix are asked
With pixel entropy；
By the spy in the angular second moment battle array, correlation, difference matrix, contrast sub-matrix and pixel entropy composition described image region
Sign vector；
Described eigenvector is identified using neural network model trained in advance, judges whether described image region is quick
Touching body region.
6. according to the method described in claim 5, it is characterized in that, the neural network model include five input layers, four
Hidden layer and an output layer.
7. according to the method described in claim 5, it is characterized in that, the neural network model is using back-propagation algorithm
Neural network model.
8. a kind of device of human body safety check that realizing artificial intelligence regions shield, which is characterized in that including：
Image collection module, the X-ray safety check fluoroscopy images for obtaining target body；
Image processing module extracts the figure surrounded by closed edge for carrying out edge detection to the X-ray safety check fluoroscopy images
As region；
Images match module, for matching described image region with pre-stored sensitive body's position template；
Marseille is stamped in described image region in the X-ray safety check fluoroscopy images by image display if being used for successful match
Gram, and show and stamp the X-ray safety check fluoroscopy images after mosaic.
9. device according to claim 8, which is characterized in that further include：
Image pre-processing module, for carrying out image enhancement to the X-ray safety check fluoroscopy images, filtering removes noise jamming, and
Artifact is removed, then enhances contrast again, enhances the contrast in the X-ray safety check fluoroscopy images difference greyscale color region.
10. system according to claim 8, which is characterized in that further include：
Memory module, for storing neural network model trained in advance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810371829.0A CN108648204A (en) | 2018-04-24 | 2018-04-24 | A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810371829.0A CN108648204A (en) | 2018-04-24 | 2018-04-24 | A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108648204A true CN108648204A (en) | 2018-10-12 |
Family
ID=63747142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810371829.0A Pending CN108648204A (en) | 2018-04-24 | 2018-04-24 | A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108648204A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109597069A (en) * | 2018-12-25 | 2019-04-09 | 山东雷诚电子科技有限公司 | A kind of active MMW imaging method for secret protection |
| CN110298797A (en) * | 2019-06-12 | 2019-10-01 | 博微太赫兹信息科技有限公司 | A kind of millimeter-wave image processing method and system based on convolutional neural networks |
| CN110443748A (en) * | 2019-07-31 | 2019-11-12 | 思百达物联网科技(北京)有限公司 | Human body screen method, device and storage medium |
| CN111209793A (en) * | 2019-12-05 | 2020-05-29 | 重庆特斯联智慧科技股份有限公司 | Region shielding human body security check method and system based on artificial intelligence |
| CN111352171A (en) * | 2020-03-30 | 2020-06-30 | 重庆特斯联智慧科技股份有限公司 | Method and system for realizing artificial intelligence regional shielding security inspection |
| CN113379677A (en) * | 2021-05-08 | 2021-09-10 | 哈尔滨理工大学 | Static stack CO60 radioactive source early warning method based on artificial intelligence |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996314A (en) * | 2009-08-26 | 2011-03-30 | 厦门市美亚柏科信息股份有限公司 | Content-based human body upper part sensitive image identification method and device |
| CN102028482A (en) * | 2009-09-30 | 2011-04-27 | 同方威视技术股份有限公司 | Human body detection image processing method and human body detection apparatus |
| CN102708560A (en) * | 2012-02-29 | 2012-10-03 | 北京无线电计量测试研究所 | Privacy protection method based on millimeter-wave imaging |
| CN103942556A (en) * | 2014-04-22 | 2014-07-23 | 天津重方科技有限公司 | Human body private part recognition and processing method based on X-ray back scattering images |
| CN106447634A (en) * | 2016-09-27 | 2017-02-22 | 中国科学院上海微系统与信息技术研究所 | Private part positioning and protection method based on active millimeter wave imaging |
-
2018
- 2018-04-24 CN CN201810371829.0A patent/CN108648204A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996314A (en) * | 2009-08-26 | 2011-03-30 | 厦门市美亚柏科信息股份有限公司 | Content-based human body upper part sensitive image identification method and device |
| CN102028482A (en) * | 2009-09-30 | 2011-04-27 | 同方威视技术股份有限公司 | Human body detection image processing method and human body detection apparatus |
| CN102708560A (en) * | 2012-02-29 | 2012-10-03 | 北京无线电计量测试研究所 | Privacy protection method based on millimeter-wave imaging |
| CN103942556A (en) * | 2014-04-22 | 2014-07-23 | 天津重方科技有限公司 | Human body private part recognition and processing method based on X-ray back scattering images |
| CN106447634A (en) * | 2016-09-27 | 2017-02-22 | 中国科学院上海微系统与信息技术研究所 | Private part positioning and protection method based on active millimeter wave imaging |
Non-Patent Citations (1)
| Title |
|---|
| 王利: "基于局部特征与全局结构的轮廓检测算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109597069A (en) * | 2018-12-25 | 2019-04-09 | 山东雷诚电子科技有限公司 | A kind of active MMW imaging method for secret protection |
| CN110298797A (en) * | 2019-06-12 | 2019-10-01 | 博微太赫兹信息科技有限公司 | A kind of millimeter-wave image processing method and system based on convolutional neural networks |
| CN110298797B (en) * | 2019-06-12 | 2021-07-09 | 博微太赫兹信息科技有限公司 | Millimeter wave image processing method based on convolutional neural network |
| CN110443748A (en) * | 2019-07-31 | 2019-11-12 | 思百达物联网科技(北京)有限公司 | Human body screen method, device and storage medium |
| CN111209793A (en) * | 2019-12-05 | 2020-05-29 | 重庆特斯联智慧科技股份有限公司 | Region shielding human body security check method and system based on artificial intelligence |
| CN111352171A (en) * | 2020-03-30 | 2020-06-30 | 重庆特斯联智慧科技股份有限公司 | Method and system for realizing artificial intelligence regional shielding security inspection |
| CN111352171B (en) * | 2020-03-30 | 2023-01-24 | 重庆特斯联智慧科技股份有限公司 | Method and system for realizing artificial intelligence regional shielding security inspection |
| CN113379677A (en) * | 2021-05-08 | 2021-09-10 | 哈尔滨理工大学 | Static stack CO60 radioactive source early warning method based on artificial intelligence |
| CN113379677B (en) * | 2021-05-08 | 2024-07-12 | 南京新频点电子科技有限公司 | Static stacking CO60 radioactive source early warning method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108648204A (en) | A kind of method and apparatus of human body safety check that realizing artificial intelligence regions shield | |
| CN108549898B (en) | A kind of method and system of specific objective identification and enhancing for safety check perspective | |
| CN103984961B (en) | A kind of image detecting method for being used to detect underbody foreign matter | |
| WO2021203618A1 (en) | Image sample generating method and system, and target detection method | |
| CN105518709B (en) | The method, system and computer program product of face for identification | |
| CN104166841B (en) | The quick detection recognition methods of pedestrian or vehicle is specified in a kind of video surveillance network | |
| CN105389797B (en) | A kind of UAV Video small target detecting method based on super-resolution rebuilding | |
| CN108090888A (en) | The infrared image of view-based access control model attention model and the fusion detection method of visible images | |
| CN108038456A (en) | A kind of anti-fraud method in face identification system | |
| CN104134059B (en) | Keep the bad image detecting method under the mixing deformation model of colouring information | |
| CN107392885A (en) | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism | |
| CN108416336A (en) | A kind of method and system of intelligence community recognition of face | |
| Mendoza et al. | A new method for edge detection in image processing using interval type-2 fuzzy logic | |
| CN109858439A (en) | A kind of biopsy method and device based on face | |
| CN108764186A (en) | Personage based on rotation deep learning blocks profile testing method | |
| CN108154087A (en) | A kind of matched infrared human body target detection tracking method of feature based | |
| CN109102003A (en) | A kind of small target detecting method and system based on Infrared Physics Fusion Features | |
| Gunawan et al. | Artificial neural network based fast edge detection algorithm for mri medical images | |
| Rezai-Rad et al. | Comparison of SUSAN and sobel edge detection in MRI images for feature extraction | |
| CN108108740B (en) | Active millimeter wave human body image gender identification method | |
| CN112069907A (en) | X-ray machine image recognition method, device and system based on example segmentation | |
| CN102760295A (en) | Fire disaster image detection system for edge detection-based operator | |
| EP3021256A1 (en) | Method for image processing, presence detector and lighting system | |
| CN106650824B (en) | Moving object classification method based on support vector machines | |
| Bechar et al. | On-line video recognition and counting of harmful insects |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181012 |
|
| RJ01 | Rejection of invention patent application after publication |