CN108764202A - Airport method for recognizing impurities, device, computer equipment and storage medium - Google Patents
Airport method for recognizing impurities, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108764202A CN108764202A CN201810574129.1A CN201810574129A CN108764202A CN 108764202 A CN108764202 A CN 108764202A CN 201810574129 A CN201810574129 A CN 201810574129A CN 108764202 A CN108764202 A CN 108764202A
- Authority
- CN
- China
- Prior art keywords
- image
- foreign matter
- detection
- detection image
- identification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 239000012535 impurity Substances 0.000 title claims abstract description 26
- 238000003860 storage Methods 0.000 title claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 167
- 239000000203 mixture Substances 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims description 65
- 238000012360 testing method Methods 0.000 claims description 45
- 230000002708 enhancing effect Effects 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 26
- 238000012545 processing Methods 0.000 claims description 22
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 16
- 210000001525 retina Anatomy 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 5
- 239000011248 coating agent Substances 0.000 claims description 5
- 238000000576 coating method Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 abstract description 9
- 230000008859 change Effects 0.000 abstract description 5
- 238000005286 illumination Methods 0.000 abstract description 5
- 238000007689 inspection Methods 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 10
- 239000000284 extract Substances 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 239000003086 colorant Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000001105 regulatory effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 208000002173 dizziness Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 229920003023 plastic Polymers 0.000 description 1
- 239000004033 plastic Substances 0.000 description 1
- 229920000573 polyethylene Polymers 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
- -1 wooden unit Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/35—Categorising the entire scene, e.g. birthday party or wedding scene
- G06V20/38—Outdoor scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of airport method for recognizing impurities, device, computer equipment and storage medium, there are the continuous predetermined frame image for when foreign matter, obtaining the detection image, composition identification image sets being gone out in detection image by foreign bodies detection model inspection.By identifying the feature vector similarity of the feature vector and reference characteristic vector of foreign matter corresponding position in image set, to obtain comparison result, recognition result is finally generated according to comparison result.It can be influenced and the erroneous judgement caused by recognition result by surrounding environment change (illumination, shade etc.) to avoid detection image, to screen out part misjudgement sample in the foreign matter identification process of airport, to improve the accuracy of identification of airport foreign matter.
Description
Technical field
The present invention relates to field of image recognition more particularly to a kind of airport method for recognizing impurities, device, computer equipment and
Storage medium.
Background technology
It often will appear various abnormal objects in airport runways, be referred to as FOD (Foreign Object Debris),
FOD refers to certain the external substance that may damage aircraft or system, frequently referred to runway foreign matter.The type of FOD is quite a lot of, such as
Aircraft and engine connector (nut, screw, washer, fuse etc.), machine tool, flight article (nail, personalized documents,
Pen, pencil etc.), wild animal, leaf, stone and sand, road plane materiel material, wooden unit, plastics or polythene material, paper products,
Ice fragment of Operational Zone etc..Experiment and case all show that the exotic on airfield pavement can be easy to be inhaled into engine,
Lead to power failure.Fragment can be also deposited in mechanical device, influence the normal operation of the equipment such as undercarriage, wing flap.
And due to the development of artificial intelligence, it begins attempt to be realized to airport foreign matter with deep learning object detection model
Detection.However, existing deep learning object detection model is broadly divided into two steps detection (Two stage detecotr) model
(FastRCNN, FasterRCNN etc.) and single step detect (Single stage detector) model (FCN, SSD etc.) two classes.
In the case that two traditional step detection models are for object scene accounting rate extremely low (insufficient one thousandth), region choose it is difficult and
Arithmetic speed is slow.Be not suitable for the scene for having certain requirement of real-time.And traditional single step detection model is inadequate for small items
Sensitivity, and for small items, final test position easy tos produce deviation.
Invention content
Based on this, it is necessary in view of the above technical problems, provide a kind of airport that can improve airport foreign matter accuracy of identification
Method for recognizing impurities, device, computer equipment and storage medium.
A kind of airport method for recognizing impurities, including：
Detection image is obtained, the detection image is detected using foreign bodies detection model, obtains testing result；
If the testing result is that there are foreign matters in the detection image, the foreign matter is obtained in the detection image
Position the feature vector of the foreign matter is extracted according to the reference position as reference position, as reference characteristic vector；
Continuous predetermined frame image, composition identification image set are obtained according to the detection image；
The feature vector of each identification image in the identification image set is extracted according to the reference position, and is compared each
It identifies the feature vector similarity of the feature vector and reference characteristic vector of image, obtains comparison result；
Recognition result is generated according to the comparison result, the recognition result includes being confirmed as foreign matter and being confirmed as non-different
Object.
A kind of airport foreign matter identification device, including：
Testing result acquisition module carries out the detection image using foreign bodies detection model for obtaining detection image
Detection obtains testing result；
Reference characteristic vector acquisition module obtains if being that there are foreign matters in the detection image for the testing result
Position of the foreign matter in the detection image is taken, as reference position, the foreign matter is extracted according to the reference position
Feature vector, as reference characteristic vector；
Image set comprising modules are identified, for obtaining continuous predetermined frame image, composition identification figure according to the detection image
Image set；
Comparison result acquisition module, for extracting each identification image in the identification image set according to the reference position
Feature vector, and compare it is each identification image feature vector and reference characteristic vector feature vector similarity, obtain
Take comparison result；
Recognition result acquisition module, for generating recognition result according to the comparison result, the recognition result includes true
Think foreign matter and is confirmed as non-foreign matter.
A kind of computer equipment, including memory, processor and be stored in the memory and can be in the processing
The computer program run on device, the processor realize above-mentioned airport method for recognizing impurities when executing the computer program
Step.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter
The step of calculation machine program realizes above-mentioned airport method for recognizing impurities when being executed by processor.
In above-mentioned airport method for recognizing impurities, device, computer equipment and storage medium, examined by foreign bodies detection model
It measures there are when foreign matter in detection image, by obtaining the continuous predetermined frame image of the detection image, composition identifies image set.It is logical
The feature vector similarity for crossing the feature vector and reference characteristic vector of foreign matter corresponding position in identification image set, compares to obtain
As a result, finally generating recognition result according to comparison result.It can be to avoid detection image by surrounding environment change (illumination, shade etc.)
It influences and the erroneous judgement caused by recognition result, to screen out part misjudgement sample in foreign matter identification process, to improve machine
The accuracy of identification of field foreign matter.
Description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the present invention
Example, for those of ordinary skill in the art, without having to pay creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is an application environment schematic diagram of airport method for recognizing impurities in one embodiment of the invention；
Fig. 2 is an exemplary plot of airport method for recognizing impurities in one embodiment of the invention；
Fig. 3 is an exemplary plot of the step S10 of airport method for recognizing impurities in one embodiment of the invention；
Fig. 4 is an exemplary plot of the step S11 of airport method for recognizing impurities in one embodiment of the invention；
Fig. 5 is an exemplary plot of the step S12 of airport method for recognizing impurities in one embodiment of the invention；
Fig. 6 is a functional block diagram of foreign matter identification device in airport in one embodiment of the invention；
Fig. 7 is a schematic diagram of one embodiment of the invention Computer equipment.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, the every other implementation that those of ordinary skill in the art are obtained without creative efforts
Example, shall fall within the protection scope of the present invention.
Airport method for recognizing impurities provided by the present application, can be applicable in the application environment such as Fig. 1, wherein client (meter
Calculate machine equipment) it is communicated with server-side by network.Client is sent in detection image to server-side, and server-side schemes detection
As being identified, recognition result is generated.Wherein, client (computer equipment) can be, but not limited to be various personal computers,
Laptop, smart mobile phone, tablet computer, video capture device and portable wearable device.Server-side can be used independent
The server cluster of server either multiple servers composition realize.
In one embodiment, it as shown in Fig. 2, providing a kind of airport method for recognizing impurities, applies in this way in Fig. 1
It illustrates, includes the following steps for server-side：
S10：Detection image is obtained, detection image is detected using foreign bodies detection model, obtains testing result.
Wherein, detection image refers to that the video data in the monitor video by airport is divided into according to certain time interval
The image of predetermined frame and formed.Preferably, detection image is ranked up according to chronological order, then foreign matter is used to examine
It surveys model to be detected detection image, obtains testing result.Optionally, testing result include in detection image there are foreign matter and
There is no two kinds of situations of foreign matter.Foreign bodies detection model is an advance trained identification model, optionally, foreign bodies detection model
Two steps detection (Two stage detecotr) model (FastRCNN, FasterRCNN etc.) or single step detection may be used
(Single stage detector) model (FCN, SSD etc.) is realized.By advance trained foreign bodies detection model to detection
Image is detected, and foreign bodies detection model exports a testing result.
S20：If testing result is that there are foreign matters in detection image, position of the foreign matter in detection image is obtained, as
Reference position extracts the feature vector of foreign matter according to reference position, as reference characteristic vector.
Detection image is detected using foreign bodies detection model, if foreign bodies detection model output the result is that the detection figure
There are foreign matters as in, then obtain position of the foreign matter in detection image, and extract corresponding feature vector conduct based on the position
Reference characteristic vector.Specifically, the foreign matter detected first can uniformly be zoomed to predefined size (such as 32*32), then carried
Feature vector is taken, as reference characteristic vector.It is alternatively possible to extract the color histogram and histograms of oriented gradients of foreign matter
(HOG, Histogram of Gradient) forms reference characteristic vector.
S30：Continuous predetermined frame image, composition identification image set are obtained according to detection image.
Corresponding continuous predetermined frame image, composition identification image set are obtained based on the detection image.Continuous predetermined frame image
It refer to adjacent with detection image in the video data where detection image and continuous predetermined frame image.For example, obtaining the inspection
The corresponding lower 20 frame images of altimetric image, composition identification image set.
S40：The feature vector of each identification image in identification image set is extracted according to reference position, and compares each identification
The feature vector similarity of feature vector and the reference characteristic vector of image obtains comparison result.
The feature vector of each identification image in identification image set is got according to reference position, and by each identification image
Feature vector and reference characteristic vector be compared, obtain feature vector similarity.Specifically, may be used Ming Shi distance,
Euclidean distance or mahalanobis distance scheduling algorithm calculate the feature vector of the feature vector and reference characteristic vector of each identification image
Similarity.Wherein, the feature vector similarity being calculated and preset similarity threshold are compared, and show that one compares
As a result, being specifically as follows：Phase Sihe is dissimilar.Such as：When feature vector similarity is greater than or equal to similarity threshold, compare
As a result it is similar；When feature vector similarity is less than similarity threshold, comparison result is dissmilarity.
S50：Recognition result is generated according to comparison result, recognition result includes being confirmed as foreign matter and being confirmed as non-foreign matter.
The comparison result of each identification image in statistics identification image set, when similar quantity is more than or waits in comparison result
When decision threshold, recognition result is to be confirmed as foreign matter.When similar quantity is less than decision threshold in comparison result, identification knot
Fruit is to be confirmed as non-foreign matter.The decision threshold can be set by identifying the quantity of the image in image set, for example, decision threshold
Value is 60%, 80% or the 90% of the quantity of the image in identification image set.
When in detection image, it may be possible to scheme in detection because of the influence of surrounding environment change (illumination, shade etc.)
Shade is formd as in.Using foreign bodies detection model detection image is detected when, it is possible to by the shade identification at
Foreign matter.It is to pass through the continuous predetermined frame figure of further Determination image there are when foreign matter in the testing result of foreign bodies detection model
The feature vector similarity of corresponding position as in, to exclude influence of the shade present in detection image to recognition result.
In the present embodiment, go out in detection image there are when foreign matter by foreign bodies detection model inspection, be somebody's turn to do by obtaining
The continuous predetermined frame image of detection image, composition identification image set.By identify image set in foreign matter corresponding position feature to
The feature vector similarity of amount and reference characteristic vector finally generates recognition result to obtain comparison result according to comparison result.
It can be influenced and the erroneous judgement caused by recognition result by surrounding environment change (illumination, shade etc.) to avoid detection image, with different
Part misjudgement sample is screened out in object identification process, to improve the accuracy of identification of airport foreign matter.
In one embodiment, as shown in figure 3, being detected to detection image using foreign bodies detection model, detection knot is obtained
Fruit includes the following steps：
S11：Detection image is pre-processed, images to be recognized is obtained.
It refers to carrying out enhancing processing to detection image to carry out pretreatment to detection image, to improve subsequent accuracy of detection.
When obtaining detection image, many because being known as of detection image are influenced, such as：Illuminance is uneven, collecting device limitation and
The difference etc. of acquisition environment will cause the clarity of detection image inadequate, lead to the reduction of subsequent accuracy of identification.Therefore, exist
By being pre-processed to detection image in the step, to improve subsequent accuracy of detection.It is alternatively possible to be adopted to detection image
Global enhancing or local enhancement processing are carried out with algorithm for image enhancement, then processing is sharpened to enhanced detection image,
Obtain images to be recognized.Preferably, algorithm for image enhancement can be multiple dimensioned retina algorithm, adaptive histogram equalization calculation
Method or optimization contrast algorithm etc..By carrying out, with after processing, just obtaining images to be recognized to detection image.
S12：Images to be recognized is input in complete poor-pyramid character network identification model and is identified, classification is obtained
Confidence map.
Wherein, complete poor-pyramid character network identification model refers to according to coding-decoded model, using complete poor
(DenseNet, Densely Connected Convolutional Networks) is used as coding network, using pyramid spy
The neural network recognization that sign (RefineNet, Multi-Path Refinement Networks) is constituted as decoding network
Model.
Specifically, complete poor network is in neural network model by the way that the network of different layers is done a splicing so that each
The input of layer network includes the output of all layer networks in front, can be lost in this way to avoid small items in model upsampling process
It loses.Complete poor network can promote the efficiency of transmission of information and gradient in a network, and every layer directly can take ladder from loss function
Degree, and input signal is directly obtained, it can thus train deeper network, this network structure to also have the effect of regularization,
Complete poor network promotes network performance from the angle of feature reuse.Therefore, model up-sampling is not only reduced using complete poor network
The phenomenon that small items are lost in the process, while also improving training speed and reducing over-fitting.
Pyramid character network is the improvement network of a multipath, extracts all information during down-sampling, is used
Long-range network connection obtains high-resolution prediction network.The feature of pyramid character network detailed level so that high-rise
Semantic information can be improved.More RCU (residual error connection unit, residual have been used in pyramid character network
Connection units) so that the connection of short-range is formd inside pyramid character network, it is beneficial to training.This
Outside, pyramid character network also forms the connection of long-range with complete poor network, and gradient can be effectively transmitted to entirely
In network, influence of the low-level image feature to final result is increased, effectively increases the positioning accuracy of object (airport foreign matter).
Confidence map of classifying refers to being marked in different ways to different classes of in image after being detected to images to be recognized
Outpour the image for coming and presenting.It is alternatively possible to be distinguished to different classes of in images to be recognized using different colors.Example
Such as：In detection image, it is possible to which the object of appearance is runway, lawn, airport equipment (non-foreign matter) and airport foreign matter etc..Cause
This, can be that above-mentioned different classes of object assigns different colors in advance.Images to be recognized is being input to complete poor-pyramid
After being identified in character network identification model, complete poor-pyramid character network identification model according in images to be recognized not
Different judging results with region form classification confidence map in conjunction with different colors is assigned in advance.
In a specific embodiment, airport foreign matter can also be labeled with more specifical object, such as：Hair
Motivation connector (nut, screw, washer, fuse etc.), machine tool, flight article (nail, personalized documents, pen, pencil
Deng) and animal etc..And these are all belonged to the classification of airport foreign matter, in this way, can be further when identifying airport foreign matter
It determines specific foreign matter type, facilitates and formulate suitable treatment measures.
S13：Testing result is obtained according to classification confidence map, testing result includes that there are foreign matters and detection figure in detection image
Foreign matter is not present as in.
After obtaining classification confidence map, testing result can be obtained according to the different colours on classification confidence map, examined
It includes that there are foreign matter is not present in foreign matter and detection image in detection image to survey result.If for example, in advance setting, by machine
Foreign matter is set as red, then after getting classification confidence map, in the confidence map that judges to classify with the presence or absence of red area from
And obtain different testing results.If illustrating that there are foreign matters in detection image, at this time there are red area in confidence map of classifying
Testing result is that there are foreign matters in detection image.If red area is not present in confidence map of classifying, illustrate in detection image not
There are foreign matters, and testing result is that there are foreign matters in detection image at this time.Optionally, testing result can by word, voice or
The modes such as person's signal lamp embody, and can also be the combination of at least two, word, voice or signal lamp.For example, working as testing result
For in detection image, there are when foreign matter, voice prompt can be sent, and prompted by the way of warning lamp, preferably to carry
Awake related personnel is handled.
In one embodiment, when in confidence map of classifying, there are when red area, the airport foreign matter can also be obtained
Location information, testing result further includes the location information of airport foreign matter at this time.Specifically, can be each images to be recognized in advance
An identification marking is assigned, the image sources for positioning the images to be recognized, such as it is which to be navigated to by the identification marking
What the photographic device of position collected.In this way, when in confidence map of classifying there are when red area, can be by obtaining the red
Position of the region in images to be recognized, in conjunction with the identification marking of the images to be recognized, to show that the red area corresponds to
Airport foreign matter in airport actual position.
The present embodiment obtains images to be recognized, to improve subsequent accuracy of detection by being pre-processed to detection image.
And images to be recognized is identified using complete poor-pyramid character network identification model, it ensure that in identification process to micro-
The accuracy of identification and positioning accuracy of wisp, also improve recognition efficiency.
In one embodiment, as shown in figure 4, in step S11, i.e., detection image is pre-processed, obtains figure to be identified
Picture specifically comprises the following steps：
S111：Global enhancing processing is carried out to detection image using multiple dimensioned retina algorithm.
Wherein, multiple dimensioned retina (Multi-Scale Retinex, MSR) algorithm is a kind of calculation of image enhancement processing
Method, the influence of the various factors (such as interference noise, edge details lack) for weakening untreated original image.Using more
Scale retina algorithm carries out enhancing processing to detection image, by removing the illumination image of detection image, retains reflectogram
Picture, and the gray scale dynamic range of detection image is adjusted, the reflective information of the corresponding reflected image of detection image is obtained, according to
This reaches enhancing effect.
Preferably, global enhancing processing is carried out to detection image using multiple dimensioned retina algorithm, specifically included：
Global enhancing processing is carried out to detection image using following formula：
Wherein, N is the number of scale, and (x, y) is the coordinate value of detection image pixel, and G (x, y) calculates for multiple dimensioned retina
The input of method, the i.e. gray value of detection image, R (x, y) are the output of multiple dimensioned retina algorithm, i.e., treated for global enhancing
The gray value of detection image, wnFor the weight factor of scale, constraints isFn(x, y) is n-th of center
Surround function, expression formula are：
In formula, σnFor the scale parameter of n-th of center surround function, COEFFICIENT KnIt must meet：
Specifically, the gray value G (x, y) that detection image is obtained by image information acquisition tool, according in the n of input
The scale parameter σ of heart surround functionnValue, determine meetKnValue, then will
Center surround function FnAfter (x, y) and G (x, y) are calculated according to following formula, global enhancing is obtained treated detection figure
The gray value R (x, y) of picture：
Wherein, σnDecision center surround function Size of Neighborhood, size determine the quality of detection image, σnWhen taking larger,
For the contiguous range of selection with regard to larger, detection image pixel is influenced smaller, the part of prominent detection image by its surrounding pixel
Details.
In a specific embodiment, the number n=3 for choosing scale, is correspondingly arranged:
σ1=30, σ2=110, σ3=200；
Wherein, σ1、σ2And σ3Low gray scale, middle gray scale and the height ash of the gray value interval [0,255] of detection image are corresponded to respectively
Degree, and w is set1=w2=w3=1/3.According to the setting of above-mentioned parameter, multiple dimensioned retina algorithm has taken into account low ash simultaneously
Degree, middle gray scale and high gray scale this 3 grey scales, to obtain preferable effect.Multiple dimensioned retina algorithm is more by combining
Good adaptivity may be implemented in a scale, highlights dark picture areas grain details, and dynamic range of images may be implemented
It adjusts and then achievees the purpose that image enhancement.
S112：Using Laplace operator, to overall situation enhancing, treated that detection image is sharpened processing, obtains waiting knowing
Other image.
Laplace operator (Laplacian operator) is a kind of Second Order Differential Operator, is suitable for improving because of light
Diffusing reflection caused by image it is fuzzy.Image can be reduced to image progress Laplace operator sharpening transformation to obscure, and improved
The clarity of image.Therefore, by the way that overall situation enhancing treated detection image is sharpened processing, prominent overall situation enhancing is handled
The edge details feature of detection image afterwards improves global enhancing treated the contour sharpness of detection image.At sharpening
Reason refers to the transformation being sharpened to image, for reinforcing the object boundary in image and image detail.After global enhancing processing
Detection image after Laplace operator Edge contrast, image edge detailss feature also weakens light while being reinforced
It is dizzy, to protect the details of detection image.
Laplace operator based on second-order differential is defined as：
For overall situation enhancing treated detection image R (x, y), second dervative is：
Therefore, Laplace operatorFor:
Obtain Laplace operatorLater, Laplace operator is usedTo overall situation enhancing treated detection image
Each grey scale pixel value of R (x, y) is all sharpened according to following formula, the grey scale pixel value after being sharpened, in formula, g (x,
Y) it is the grey scale pixel value after sharpening.
The gray scale that grey scale pixel value after sharpening is replaced at former (x, y) pixel is worth to images to be recognized.
In a specific embodiment, Laplace operatorFour neighborhoods are selected to sharpen pattern matrix H:
Sharpening pattern matrix H using four neighborhoods, to carry out Laplace operator to overall situation enhancing treated detection image sharp
Change.
In the present embodiment, global enhancing processing is carried out to detection image using multiple dimensioned retina algorithm, it will be through excessive
Detection image after scale retina algorithm enhancing processing is sharpened using Laplace operator, image edge detailss feature
Also halation is weakened while being reinforced, to protect the details of detection image.In addition, above-mentioned steps are not only simple and convenient,
It is more prominent clear that images to be recognized edge details feature is obtained after processing, enhances the textural characteristics of images to be recognized, favorably
In the accuracy rate for improving images to be recognized identification.
In one embodiment, as shown in figure 5, identifying mould images to be recognized is input to complete poor-pyramid character network
Before the step of being identified in type, obtaining classification confidence map, which further includes：
S121：Training sample set is obtained, classification annotation is carried out to the training image that training sample is concentrated.
Wherein, training sample concentration includes training image, and training image refers to for the complete poor-pyramid feature net of training
The sample image of network identification model.Optionally, which can be by being arranged video capture device in airport different location
Either image capture device collects correspondence to acquire corresponding data to obtain video capture device or image capture device
Data after be sent to server-side.It, can be to video data according to scheduled frame if what server-side was got is video data
Rate carries out sub-frame processing to obtain training image.It refers to the different objects in training image to carry out classification annotation to training image
Classify.For example, in training image, it is possible to which the object of appearance is runway, lawn, airport equipment (non-foreign matter) and airport
Foreign matter etc..By assigning different markup informations to the different objects in training image to complete the contingency table to training image
Note.
S122：The complete poor network of training image training that classification annotation is concentrated using training sample, obtains target output vector.
In this step, the training image of classification annotation is concentrated using training sample to train complete poor network, and complete poor
In network, setting training image input is x0, complete poor network is by L layers of structure composition, and poor network all includes one non-to each layer entirely
Linear transformation Hl(·).Optionally, nonlinear transformation may include ReLU (activation primitive, Rectified Linear Units)
With Pooling (pond) or BN (standardization layer, Batch Normalization), ReLU and convolutional layer or BN, ReLU
And Pooling.Wherein, BN is exactly by standardization means, the distribution tune of the input value of the arbitrary neuron of every layer of neural network
The standardized normal distribution that whole is 0 to mean value and variance is 1 so that activation input value falls more sensitive to inputting in nonlinear function
Region, allow gradient to become larger, avoid gradient disappearance problem generate, training speed can be greatly speeded up.ReLU is a piecewise linearity
Function and a kind of unilateral inhibition function, it is 0 that can all export all negative values of input, and the positive value inputted is then kept
It is constant.By ReLU may be implemented it is sparse after model can preferably excavate correlated characteristic, be fitted training data.
In the present embodiment, if l layers of output is x in complete poor networkl, each layer is all with front in poor network entirely
Layer be directly connected to, i.e.,：
xl=Hl([x0,x1,...,xl-1])；
And the output of respective layer just constitutes target output vector in complete poor network, for subsequently using the target export to
Amount training pyramid character network.
S123：Pyramid character network is trained using target output vector, obtains complete poor-pyramid character network identification mould
Type.
In pyramid character network, each layer output in the target output vector in complete poor network can respectively and pyramid
The RCU units of character network connect.There is the number of plies with the target output vector in complete poor network i.e. in pyramid character network
Identical RCU units.
RCU units refer to the cellular construction extracted from complete poor network, have specifically included ReLU, convolution sum summation etc.
Part.By the way that each layer target output vector obtained in complete poor network is passed through ReLU, convolution sum sum operation respectively.RCU is mono-
Each layer output of member is all handled using Multi-resolution fusion (multi-resolution Fusion), to obtain difference
Output characteristic pattern：Self-adaptive processing all first is carried out with a convolutional layer to each layer of RCU units output characteristic pattern, then is carried out
Sample the maximum resolution of this layer.Chained residual pooling (chain type residual error pond) differentiate the difference of input
The output characteristic pattern of rate is upsampled to the size equal with maximum output characteristic pattern and then is overlapped.It finally will be defeated after superposition
Go out characteristic pattern and carries out convolution using RCU to get to a fine-feature figure.
The effect of pyramid character network is exactly that the characteristic pattern of different resolution is merged.First the complete of training in advance
Poor network is divided into several complete poor blocks by the resolution ratio of characteristic pattern, if then to the right this several blocks respectively as
A dry path is merged by pyramid character network, finally obtain a fine-feature figure (follow-up softmax layers of connection,
It is exported again by bilinear interpolation).
In pyramid character network, pyramid character network is trained to be formed by the target output vector in complete poor network
One initial training network, then pyramid character network is verified and adjusted using verification sample, until obtaining preset
Classification accuracy, then training terminate.Wherein, preset classification accuracy can be set according to the needs of actual identification model
It sets.
In this embodiment, it trains to obtain complete poor-pyramid feature by using the training sample set after classification annotation
Network Recognition model ensure that the accuracy of identification and speed of the complete poor-pyramid character network identification model.
In one embodiment, the complete poor network of training, specifically includes：
The initial convolutional layer of complete poor network is set, and down-sampling is carried out using the maximum pond layer in complete poor network.
Convolutional layer for feature extraction to input picture, and initial convolutional layer extraction be exactly training image feature,
Optionally, initial convolutional layer uses the convolution kernel of 7*7.Down-sampling is carried out using the maximum pond layer in complete poor network, is being carried out
It is down-sampling if new sample rate is less than former sample rate during sampling.Maximum pond (max-pooling) refers to sampling
Function takes the maximum value of all neurons in region.Maximum pondization processing is carried out by the input picture by initial convolutional layer,
Feature Compression is carried out, extracts main feature, and simplify network calculations complexity.
Three layers of complete poor network module are set, and each complete poor network module includes a complete poor convolutional layer and a complete poor activation
Layer, the activation primitive in poor active coating is using linear activation primitive entirely.
In this three layers complete poor network module, the output of each complete poor network module is all module outputs in front
In conjunction with that is,：
xl=Hl([x0,x1,...,xl-1]), l=1,2,3；
Wherein, each Hl() is all the combination of two layer operation of convolutional layer and active coating：Conv->ReLU.Optionally,
Convolution kernel size in complete poor convolutional layer is 3*3.Every HlThe feature quantity of () output is characterized growth rate, optionally,
It is 16 that feature growth rate, which is arranged, then the output feature quantity of three layers of complete poor network module output is 48.And linear activation primitive public affairs
Formula is：
By the conversion of linear activation primitive, the time of training process can be made rapidly to restrain.
Transport layer is set between complete poor network module, and each transport layer includes standardization layer, transmission active coating and average
Pond layer.
In complete poor network module, the output feature of each complete poor network module be increased, in above-mentioned setting,
If feature growth rate is 16, the output of three layers of complete poor network module output is characterized as 48.In this way, calculation amount is to incrementally increase
, therefore transport layer is introduced, configured transmission is set to indicate the input by transport layer narrows down to original how many times.Example
The input of transport layer is narrowed down to original 0.6 by ground, configured transmission 0.6.
In the present embodiment, by the setting to network structure and each parameter in complete poor network, it ensure that complete poor network
Training speed and precision.
In one embodiment, during training complete poor-pyramid character network identification model, loss function uses
Focal Loss functions are realized：
FL(pt)=- (1-pt)γlog(pt)；
Wherein,ptIt is complete poor-pyramid character network identification model to the pre- of training image
Measured value, p are estimated probability of the model for training image y=1, and p ∈ [0,1], y are the mark value of training image, and γ is to adjust
Parameter.
Loss function refer to it is a kind of by an event (element in a sample space) be mapped to one expression with
A kind of function on the real number of the relevant economic cost of its event or opportunity cost.In the present embodiment, in complete poor-golden word of training
When tower character network identification model, the good of this complete poor-pyramid character network identification model prediction is weighed using loss function
Bad, loss function is smaller, and the predictive ability of the identification model is better.In embodiments of the present invention, the training figure of training sample set
The sample image quantity of each classification as in may be unbalanced, and the training image for especially including airport foreign matter may be less, is
It preferably promotes the predictive ability of complete poor-pyramid character network identification model and loss function is selected.
Therefore loss function uses the realization of Focal Loss functions, and Focal loss functions increase a regulatory factor
(1-pt)γ, wherein the value of adjustment parameter γ is between [0,5].Y is the mark value of training image, illustratively, for training
The mark of foreign matter in image is then labeled as y=1 if it is foreign matter, and non-foreign matter is then labeled as y=-1.When a training image
When by misclassification, PtVery little, at this time regulatory factor (1-pt)γClose to 1, which will not have a significant impact；Work as PtValue is very big, becomes
When being bordering on 1, the value of regulatory factor levels off to 0, therefore the loss values of the sample for correctly classifying are reduced.
In the present embodiment, Focal Loss are used during training complete poor-pyramid character network identification model
Function can reduce influence of the classification samples unevenness to the complete poor-pyramid character network identification model of training, also reach and carry
The effect of high subsequent detection precision.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
In one embodiment, a kind of airport foreign matter identification device is provided, the airport foreign matter identification device and above-described embodiment
Middle airport method for recognizing impurities corresponds.As shown in fig. 6, the airport foreign matter identification device includes testing result acquisition module
10, reference characteristic vector acquisition module 20, identification image set comprising modules 30, comparison result acquisition module 40 and recognition result obtain
Modulus block 50.Detailed description are as follows for each function module：
Testing result acquisition module 10 examines detection image using foreign bodies detection model for obtaining detection image
It surveys, obtains testing result.
Reference characteristic vector acquisition module 20 obtains foreign matter if being that there are foreign matters in detection image for testing result
Position in detection image extracts the feature vector of foreign matter according to reference position as reference position, as reference characteristic to
Amount.
Image set comprising modules 30 are identified, for obtaining continuous predetermined frame image, composition identification image according to detection image
Collection.
Comparison result acquisition module 40, the feature for extracting each identification image in identification image set according to reference position
Vector, and the feature vector similarity of the feature vector and reference characteristic vector of each identification image is compared, obtain comparison result.
Recognition result acquisition module 50, for generating recognition result according to comparison result, recognition result is different including being confirmed as
Object and it is confirmed as non-foreign matter.
Preferably, testing result acquisition module 10 includes images to be recognized acquiring unit 11, classification confidence map acquiring unit
12 and testing result acquiring unit 13.
Images to be recognized acquiring unit 11 obtains images to be recognized for being pre-processed to detection image.
Classification confidence map acquiring unit 12 identifies mould for images to be recognized to be input to complete poor-pyramid character network
It is identified in type, obtains classification confidence map.
Testing result acquiring unit 13, for obtaining testing result according to classification confidence map, testing result includes that detection is schemed
There are foreign matter is not present in foreign matter and detection image as in.
Preferably, images to be recognized acquiring unit 11 includes that global enhancing handles subelement 111 and Edge contrast subelement
112。
Overall situation enhancing processing subelement 111, for carrying out global enhancing to original image using multiple dimensioned retina algorithm
Processing.
Edge contrast unit 112, for being carried out to overall situation enhancing treated original image using Laplace operator
Edge contrast obtains images to be recognized.
Preferably, which further includes training sample set acquisition module 121, target output vector acquisition
Module 122 and identification model acquisition module 123.
Training sample set acquisition module 121 carries out the training image that training sample is concentrated for obtaining training sample set
Classification annotation.
Target output vector acquisition module 122, for concentrating the training image training of classification annotation complete using training sample
Poor network, obtains target output vector.
Identification model acquisition module 123, for training pyramid character network using target output vector, obtain it is complete it is poor-
Pyramid character network identification model.
Specific about airport foreign matter identification device limits the limit that may refer to above for airport method for recognizing impurities
Fixed, details are not described herein.Modules in above-mentioned airport foreign matter identification device can fully or partially through software, hardware and its
It combines to realize.Above-mentioned each module can be embedded in or in the form of hardware independently of in the processor in computer equipment, can also
It is stored in a software form in the memory in computer equipment, in order to which processor calls the above modules of execution corresponding
Operation.
In one embodiment, a kind of computer equipment is provided, which can be server-side, internal junction
Composition can be as shown in Figure 7.The computer equipment include the processor connected by system bus, memory, network interface and
Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment
Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data
Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The database of machine equipment is for storing detection image and foreign bodies detection model data.The network interface of the computer equipment be used for
External terminal is communicated by network connection.To realize a kind of airport foreign matter identification side when the computer program is executed by processor
Method.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, processor realize following steps when executing computer program：
Detection image is obtained, detection image is detected using foreign bodies detection model, obtains testing result.
If the testing result is that there are foreign matters in detection image, position of the foreign matter in detection image is obtained, as
Reference position extracts the feature vector of foreign matter according to reference position, as reference characteristic vector.
Continuous predetermined frame image, composition identification image set are obtained according to detection image.
The feature vector of each identification image in identification image set is extracted according to reference position, and compares each identification image
Feature vector and reference characteristic vector feature vector similarity, obtain comparison result.
Recognition result is generated according to comparison result, recognition result includes being confirmed as foreign matter and being confirmed as non-foreign matter.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
Machine program realizes following steps when being executed by processor：
Detection image is obtained, detection image is detected using foreign bodies detection model, obtains testing result.
If the testing result is that there are foreign matters in detection image, position of the foreign matter in detection image is obtained, as
Reference position extracts the feature vector of foreign matter according to reference position, as reference characteristic vector.
Continuous predetermined frame image, composition identification image set are obtained according to detection image.
The feature vector of each identification image in identification image set is extracted according to reference position, and compares each identification image
Feature vector and reference characteristic vector feature vector similarity, obtain comparison result.
Recognition result is generated according to comparison result, recognition result includes being confirmed as foreign matter and being confirmed as non-foreign matter.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the flow of the embodiment of above-mentioned each method.Wherein,
Any reference to memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each work(
Can unit, module division progress for example, in practical application, can be as needed and by above-mentioned function distribution by different
Functional unit, module are completed, i.e., the internal structure of described device are divided into different functional units or module, more than completion
The all or part of function of description.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations；Although with reference to aforementioned reality
Applying example, invention is explained in detail, it will be understood by those of ordinary skill in the art that：It still can be to aforementioned each
Technical solution recorded in embodiment is modified or equivalent replacement of some of the technical features；And these are changed
Or replace, the spirit and scope for various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution should all
It is included within protection scope of the present invention.
Claims (10)
1. a kind of airport method for recognizing impurities, which is characterized in that including：
Detection image is obtained, the detection image is detected using foreign bodies detection model, obtains testing result；
If the testing result is that there are foreign matters in the detection image, position of the foreign matter in the detection image is obtained
It sets, as reference position, the feature vector of the foreign matter is extracted according to the reference position, as reference characteristic vector；
Continuous predetermined frame image, composition identification image set are obtained according to the detection image；
The feature vector of each identification image in the identification image set is extracted according to the reference position, and compares each identification
The feature vector similarity of the feature vector of image and reference characteristic vector obtains comparison result；
Recognition result is generated according to the comparison result, the recognition result includes being confirmed as foreign matter and being confirmed as non-foreign matter.
2. airport method for recognizing impurities as described in claim 1, which is characterized in that described to use foreign bodies detection model to described
Detection image is detected, and is obtained testing result, is specifically included：
The detection image is pre-processed, images to be recognized is obtained；
The images to be recognized is input in complete poor-pyramid character network identification model and is identified, classification confidence is obtained
Figure；
Testing result is obtained according to the classification confidence map, the testing result includes that there are foreign matters and detection figure in detection image
Foreign matter is not present as in.
3. airport method for recognizing impurities as claimed in claim 2, which is characterized in that described to be located in advance to the detection image
Reason, obtains images to be recognized, specifically includes：
Global enhancing processing is carried out to the detection image using multiple dimensioned retina algorithm；
Using Laplace operator, to overall situation enhancing, treated that the detection image is sharpened processing, obtains figure to be identified
Picture.
4. airport method for recognizing impurities as claimed in claim 2, which is characterized in that input the images to be recognized described
To before the step of being identified in complete poor-pyramid character network identification model, obtain classification confidence map, the airport foreign matter
Recognition methods further includes：
Training sample set is obtained, classification annotation is carried out to the training image that the training sample is concentrated；
Using the complete poor network of training image training of classification annotation described in the training sample set, target output vector is obtained；
Pyramid character network is trained using the target output vector, obtains the complete poor-pyramid character network identification mould
Type.
5. airport method for recognizing impurities as claimed in claim 4, which is characterized in that the training poor network entirely specifically includes：
The initial convolutional layer of the complete poor network is set, and down-sampling is carried out using the maximum pond layer in the complete poor network；
Three layers of complete poor network module are set, and each complete poor network module includes a complete poor convolutional layer and a complete poor activation
Layer, the activation primitive in the complete poor active coating uses linear activation primitive；
Transport layer is set between the complete poor network module, each transport layer include standardization layer, transmission active coating and
Average pond layer.
6. airport method for recognizing impurities as claimed in claim 4, which is characterized in that in the complete poor-pyramid character network of training
During identification model, loss function is realized using Focal Loss functions：
FL(pt)=- (1-pt)γlog(pt)；
Wherein,ptIt is the complete poor-pyramid character network identification model to the training image
Predicted value, p be model for the training image y=1 estimated probability, p ∈ [0,1], y be the training image mark
Value, γ is adjustment parameter.
7. a kind of airport foreign matter identification device, which is characterized in that including：
Testing result acquisition module is detected the detection image using foreign bodies detection model for obtaining detection image,
Obtain testing result；
Reference characteristic vector acquisition module obtains institute if being that there are foreign matters in the detection image for the testing result
Position of the foreign matter in the detection image is stated, as reference position, the feature of the foreign matter is extracted according to the reference position
Vector, as reference characteristic vector；
Image set comprising modules are identified, for obtaining continuous predetermined frame image, composition identification image set according to the detection image；
Comparison result acquisition module, the spy for extracting each identification image in the identification image set according to the reference position
Sign vector, and the feature vector similarity of the feature vector and reference characteristic vector of each identification image is compared, obtain ratio
Relatively result；
Recognition result acquisition module, for generating recognition result according to the comparison result, the recognition result includes being confirmed as
Foreign matter and it is confirmed as non-foreign matter.
8. airport foreign matter identification device as claimed in claim 7, which is characterized in that testing result acquisition module includes：
Images to be recognized acquiring unit obtains images to be recognized for being pre-processed to the detection image；
Classification confidence map acquiring unit, for the images to be recognized to be input to complete poor-pyramid character network identification model
In be identified, obtain classification confidence map；
Testing result acquiring unit, for obtaining testing result according to the classification confidence map, the testing result includes detection
There are foreign matter is not present in foreign matter and detection image in image.
9. a kind of computer equipment, including memory, processor and it is stored in the memory and can be in the processor
The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to
The step of any one of 6 airport method for recognizing impurities.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, feature to exist
In realizing the airport method for recognizing impurities as described in any one of claim 1 to 6 when the computer program is executed by processor
Step.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810574129.1A CN108764202B (en) | 2018-06-06 | 2018-06-06 | Airport foreign matter identification method and device, computer equipment and storage medium |
| PCT/CN2018/092614 WO2019232831A1 (en) | 2018-06-06 | 2018-06-25 | Method and device for recognizing foreign object debris at airport, computer apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810574129.1A CN108764202B (en) | 2018-06-06 | 2018-06-06 | Airport foreign matter identification method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764202A true CN108764202A (en) | 2018-11-06 |
| CN108764202B CN108764202B (en) | 2023-04-18 |
Family
ID=63999142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810574129.1A Active CN108764202B (en) | 2018-06-06 | 2018-06-06 | Airport foreign matter identification method and device, computer equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN108764202B (en) |
| WO (1) | WO2019232831A1 (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109218233A (en) * | 2018-11-14 | 2019-01-15 | 国家电网有限公司 | OFDM channel estimation method based on depth characteristic converged network |
| CN109583502A (en) * | 2018-11-30 | 2019-04-05 | 天津师范大学 | A kind of pedestrian's recognition methods again based on confrontation erasing attention mechanism |
| CN109766884A (en) * | 2018-12-26 | 2019-05-17 | 哈尔滨工程大学 | A kind of airfield runway foreign matter detecting method based on Faster-RCNN |
| CN109765557A (en) * | 2018-12-30 | 2019-05-17 | 上海微波技术研究所（中国电子科技集团公司第五十研究所） | The recognition methods of FOD objective self-adapting Fast Classification, system and medium based on distribution character |
| CN109829501A (en) * | 2019-02-01 | 2019-05-31 | 北京市商汤科技开发有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN110268420A (en) * | 2019-05-09 | 2019-09-20 | 京东方科技集团股份有限公司 | In the picture detect background objects on exotic computer implemented method, in the picture detect background objects on exotic equipment and computer program product |
| CN111160074A (en) * | 2018-11-08 | 2020-05-15 | 欧菲影像技术（广州）有限公司 | Method and device for identifying foreign matters in equipment, computer equipment and storage medium |
| CN111553277A (en) * | 2020-04-28 | 2020-08-18 | 电子科技大学 | Chinese signature identification method and terminal introducing consistency constraint |
| CN111680609A (en) * | 2020-06-03 | 2020-09-18 | 合肥中科类脑智能技术有限公司 | Foreign matter identification system and method based on image registration and target detection |
| CN111862105A (en) * | 2019-04-29 | 2020-10-30 | 北京字节跳动网络技术有限公司 | Image area processing method and device and electronic equipment |
| CN112037197A (en) * | 2020-08-31 | 2020-12-04 | 中冶赛迪重庆信息技术有限公司 | Hot-rolled bar cold-shearing material accumulation detection method, system and medium |
| CN112580509A (en) * | 2020-12-18 | 2021-03-30 | 中国民用航空总局第二研究所 | Logical reasoning type road surface detection method and system |
| CN112686172A (en) * | 2020-12-31 | 2021-04-20 | 上海微波技术研究所（中国电子科技集团公司第五十研究所） | Method and device for detecting foreign matters on airport runway and storage medium |
| CN113177922A (en) * | 2021-05-06 | 2021-07-27 | 中冶赛迪重庆信息技术有限公司 | Raw material foreign matter identification method, system, medium and electronic terminal |
| CN113393401A (en) * | 2021-06-24 | 2021-09-14 | 上海科技大学 | Object detection hardware accelerators, systems, methods, apparatus, and media |
| WO2023118937A1 (en) * | 2021-12-20 | 2023-06-29 | Sensetime International Pte. Ltd. | Object recognition method, apparatus, device and storage medium |
| CN117690164A (en) * | 2024-01-30 | 2024-03-12 | 成都欣纳科技有限公司 | Airport bird identification and driving method and system based on edge calculation |
Families Citing this family (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919869B (en) * | 2019-02-28 | 2021-06-04 | 腾讯科技（深圳）有限公司 | Image enhancement method and device and storage medium |
| CN111178200B (en) * | 2019-12-20 | 2023-06-02 | 海南车智易通信息技术有限公司 | Method for identifying instrument panel indicator lamp and computing equipment |
| CN111259763B (en) * | 2020-01-13 | 2024-02-02 | 华雁智能科技(集团)股份有限公司 | Target detection method, target detection device, electronic equipment and readable storage medium |
| CN111539443B (en) * | 2020-01-22 | 2024-02-09 | 北京小米松果电子有限公司 | Image recognition model training method and device and storage medium |
| CN111310645B (en) * | 2020-02-12 | 2023-06-13 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for warning overflow bin of goods accumulation |
| CN111340796B (en) * | 2020-03-10 | 2023-07-21 | 创新奇智(成都)科技有限公司 | Defect detection method and device, electronic equipment and storage medium |
| CN111462059B (en) * | 2020-03-24 | 2023-09-29 | 湖南大学 | Parallel processing method and device for intelligent target detection of fetal ultrasonic image |
| CN113449554B (en) * | 2020-03-25 | 2024-03-08 | 北京灵汐科技有限公司 | Target detection and identification method and system |
| CN111340137A (en) * | 2020-03-26 | 2020-06-26 | 上海眼控科技股份有限公司 | Image recognition method, device and storage medium |
| CN111897025A (en) * | 2020-08-06 | 2020-11-06 | 航泰众联(北京)科技有限公司 | Airport pavement foreign matter detection equipment and system based on 3D/2D integration detection |
| CN112308073B (en) * | 2020-11-06 | 2023-08-25 | 中冶赛迪信息技术(重庆)有限公司 | Method, system, equipment and medium for identifying loading and unloading and transferring states of scrap steel train |
| CN112396005A (en) * | 2020-11-23 | 2021-02-23 | 平安科技（深圳）有限公司 | Biological characteristic image recognition method and device, electronic equipment and readable storage medium |
| CN112668423B (en) * | 2020-12-18 | 2024-05-28 | 平安科技（深圳）有限公司 | Corridor sundry detection method and device, terminal equipment and storage medium |
| CN113159089A (en) * | 2021-01-18 | 2021-07-23 | 安徽建筑大学 | Pavement damage identification method, system, computer equipment and storage medium |
| CN112766151B (en) * | 2021-01-19 | 2022-07-12 | 北京深睿博联科技有限责任公司 | Binocular target detection method and system for blind guiding glasses |
| CN112949474B (en) * | 2021-02-26 | 2023-01-31 | 山东鹰格信息工程有限公司 | Airport FOD monitoring method, equipment, storage medium and device |
| CN113111703B (en) * | 2021-03-02 | 2023-07-28 | 郑州大学 | Airport pavement disease foreign matter detection method based on fusion of multiple convolutional neural networks |
| CN113096080B (en) * | 2021-03-30 | 2024-01-16 | 四川大学华西第二医院 | Image analysis method and system |
| CN113111810B (en) * | 2021-04-20 | 2023-12-08 | 北京嘀嘀无限科技发展有限公司 | Target identification method and system |
| CN113177459A (en) * | 2021-04-25 | 2021-07-27 | 云赛智联股份有限公司 | Intelligent video analysis method and system for intelligent airport service |
| CN113344900B (en) * | 2021-06-25 | 2023-04-18 | 北京市商汤科技开发有限公司 | Airport runway intrusion detection method, airport runway intrusion detection device, storage medium and electronic device |
| CN113780074A (en) * | 2021-08-04 | 2021-12-10 | 五邑大学 | Method and device for detecting quality of wrapping paper and storage medium |
| CN113658135B (en) * | 2021-08-17 | 2024-02-02 | 中国矿业大学 | Fuzzy PID-based self-adaptive dimming belt foreign matter detection method and system |
| CN113807227B (en) * | 2021-09-11 | 2023-07-25 | 浙江浙能嘉华发电有限公司 | Safety monitoring method, device, equipment and storage medium based on image recognition |
| CN114155746B (en) * | 2021-12-01 | 2022-09-13 | 南京莱斯电子设备有限公司 | FOD alarm accuracy rate and FOD alarm false alarm rate calculation method |
| CN114821194B (en) * | 2022-05-30 | 2023-07-25 | 深圳市科荣软件股份有限公司 | Equipment running state identification method and device |
| CN115147770B (en) * | 2022-08-30 | 2022-12-02 | 山东千颐科技有限公司 | Belt foreign matter vision recognition system based on image processing |
| CN116269378B (en) * | 2023-01-09 | 2023-11-17 | 西安电子科技大学 | Psychological health state detection device based on skin nicotinic acid response video analysis |
| CN116704446B (en) * | 2023-08-04 | 2023-10-24 | 武汉工程大学 | Real-time detection method and system for foreign matters on airport runway pavement |
| CN117523318B (en) * | 2023-12-26 | 2024-04-16 | 宁波微科光电股份有限公司 | Anti-light interference subway shielding door foreign matter detection method, device and medium |
| CN117726882A (en) * | 2024-02-07 | 2024-03-19 | 杭州宇泛智能科技有限公司 | Tower crane object identification method, system and electronic equipment |
| CN117994753B (en) * | 2024-04-03 | 2024-06-07 | 浙江浙能数字科技有限公司 | Vision-based device and method for detecting abnormality of entrance track of car dumper |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017107188A1 (en) * | 2015-12-25 | 2017-06-29 | 中国科学院深圳先进技术研究院 | Method and apparatus for rapidly recognizing video classification |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| CN107545263A (en) * | 2017-08-02 | 2018-01-05 | 清华大学 | A kind of object detecting method and device |
| CN107562900A (en) * | 2017-09-07 | 2018-01-09 | 广州辰创科技发展有限公司 | Method and system for analyzing airfield runway foreign matter based on big data mode |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105160362B (en) * | 2015-10-22 | 2018-10-09 | 中国民用航空总局第二研究所 | A kind of runway FOD image detection method and devices |
| CN105740910A (en) * | 2016-02-02 | 2016-07-06 | 北京格灵深瞳信息技术有限公司 | Vehicle object detection method and device |
| CN107766864B (en) * | 2016-08-23 | 2022-02-01 | 斑马智行网络(香港)有限公司 | Method and device for extracting features and method and device for object recognition |
| EP3151164A3 (en) * | 2016-12-26 | 2017-04-12 | Argosai Teknoloji Anonim Sirketi | A method for foreign object debris detection |
| CN107481233A (en) * | 2017-08-22 | 2017-12-15 | 广州辰创科技发展有限公司 | A kind of image-recognizing method being applied in FOD foreign bodies detection radars |
-
2018
- 2018-06-06 CN CN201810574129.1A patent/CN108764202B/en active Active
- 2018-06-25 WO PCT/CN2018/092614 patent/WO2019232831A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017107188A1 (en) * | 2015-12-25 | 2017-06-29 | 中国科学院深圳先进技术研究院 | Method and apparatus for rapidly recognizing video classification |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| CN107545263A (en) * | 2017-08-02 | 2018-01-05 | 清华大学 | A kind of object detecting method and device |
| CN107562900A (en) * | 2017-09-07 | 2018-01-09 | 广州辰创科技发展有限公司 | Method and system for analyzing airfield runway foreign matter based on big data mode |
Non-Patent Citations (1)
| Title |
|---|
| IVAN KRESO等: "Ladder-style DenseNets for Semantic Segmentation of Large Natural Images" * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160074A (en) * | 2018-11-08 | 2020-05-15 | 欧菲影像技术（广州）有限公司 | Method and device for identifying foreign matters in equipment, computer equipment and storage medium |
| CN109218233A (en) * | 2018-11-14 | 2019-01-15 | 国家电网有限公司 | OFDM channel estimation method based on depth characteristic converged network |
| CN109218233B (en) * | 2018-11-14 | 2021-07-20 | 国家电网有限公司 | OFDM channel estimation method based on depth feature fusion network |
| CN109583502A (en) * | 2018-11-30 | 2019-04-05 | 天津师范大学 | A kind of pedestrian's recognition methods again based on confrontation erasing attention mechanism |
| CN109583502B (en) * | 2018-11-30 | 2022-11-18 | 天津师范大学 | Pedestrian re-identification method based on anti-erasure attention mechanism |
| CN109766884A (en) * | 2018-12-26 | 2019-05-17 | 哈尔滨工程大学 | A kind of airfield runway foreign matter detecting method based on Faster-RCNN |
| CN109765557A (en) * | 2018-12-30 | 2019-05-17 | 上海微波技术研究所（中国电子科技集团公司第五十研究所） | The recognition methods of FOD objective self-adapting Fast Classification, system and medium based on distribution character |
| CN109829501A (en) * | 2019-02-01 | 2019-05-31 | 北京市商汤科技开发有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN111862105A (en) * | 2019-04-29 | 2020-10-30 | 北京字节跳动网络技术有限公司 | Image area processing method and device and electronic equipment |
| WO2020223963A1 (en) * | 2019-05-09 | 2020-11-12 | Boe Technology Group Co., Ltd. | Computer-implemented method of detecting foreign object on background object in image, apparatus for detecting foreign object on background object in image, and computer-program product |
| CN110268420A (en) * | 2019-05-09 | 2019-09-20 | 京东方科技集团股份有限公司 | In the picture detect background objects on exotic computer implemented method, in the picture detect background objects on exotic equipment and computer program product |
| US11238301B2 (en) | 2019-05-09 | 2022-02-01 | Boe Technology Group Co., Ltd. | Computer-implemented method of detecting foreign object on background object in an image, apparatus for detecting foreign object on background object in an image, and computer-program product |
| CN111553277A (en) * | 2020-04-28 | 2020-08-18 | 电子科技大学 | Chinese signature identification method and terminal introducing consistency constraint |
| CN111553277B (en) * | 2020-04-28 | 2022-04-26 | 电子科技大学 | Chinese signature identification method and terminal introducing consistency constraint |
| CN111680609A (en) * | 2020-06-03 | 2020-09-18 | 合肥中科类脑智能技术有限公司 | Foreign matter identification system and method based on image registration and target detection |
| CN111680609B (en) * | 2020-06-03 | 2023-02-07 | 合肥中科类脑智能技术有限公司 | Foreign matter identification system and method based on image registration and target detection |
| CN112037197A (en) * | 2020-08-31 | 2020-12-04 | 中冶赛迪重庆信息技术有限公司 | Hot-rolled bar cold-shearing material accumulation detection method, system and medium |
| CN112580509A (en) * | 2020-12-18 | 2021-03-30 | 中国民用航空总局第二研究所 | Logical reasoning type road surface detection method and system |
| CN112580509B (en) * | 2020-12-18 | 2022-04-15 | 中国民用航空总局第二研究所 | Logical reasoning type road surface detection method and system |
| CN112686172A (en) * | 2020-12-31 | 2021-04-20 | 上海微波技术研究所（中国电子科技集团公司第五十研究所） | Method and device for detecting foreign matters on airport runway and storage medium |
| CN113177922A (en) * | 2021-05-06 | 2021-07-27 | 中冶赛迪重庆信息技术有限公司 | Raw material foreign matter identification method, system, medium and electronic terminal |
| CN113393401A (en) * | 2021-06-24 | 2021-09-14 | 上海科技大学 | Object detection hardware accelerators, systems, methods, apparatus, and media |
| CN113393401B (en) * | 2021-06-24 | 2023-09-05 | 上海科技大学 | Object detection hardware accelerator, system, method, apparatus and medium |
| WO2023118937A1 (en) * | 2021-12-20 | 2023-06-29 | Sensetime International Pte. Ltd. | Object recognition method, apparatus, device and storage medium |
| CN117690164A (en) * | 2024-01-30 | 2024-03-12 | 成都欣纳科技有限公司 | Airport bird identification and driving method and system based on edge calculation |
| CN117690164B (en) * | 2024-01-30 | 2024-04-30 | 成都欣纳科技有限公司 | Airport bird identification and driving method and system based on edge calculation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764202B (en) | 2023-04-18 |
| WO2019232831A1 (en) | 2019-12-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764202A (en) | Airport method for recognizing impurities, device, computer equipment and storage medium | |
| CN109086656A (en) | Airport foreign matter detecting method, device, computer equipment and storage medium | |
| Zavrtanik et al. | Draem-a discriminatively trained reconstruction embedding for surface anomaly detection | |
| CN109800736B (en) | Road extraction method based on remote sensing image and deep learning | |
| CN107316294B (en) | Lung nodule feature extraction method based on improved depth Boltzmann machine | |
| CN108416307A (en) | A kind of Aerial Images road surface crack detection method, device and equipment | |
| US11308714B1 (en) | Artificial intelligence system for identifying and assessing attributes of a property shown in aerial imagery | |
| CN105303150B (en) | Realize the method and system of image procossing | |
| CN111723657B (en) | River foreign matter detection method and device based on YOLOv3 and self-optimization | |
| US11436443B2 (en) | Testing machine learning (ML) models for robustness and accuracy using generative deep learning | |
| CN103020582A (en) | Method for computer to identify vehicle type by video image | |
| CN109815945A (en) | A kind of respiratory tract inspection result interpreting system and method based on image recognition | |
| CN110852358A (en) | Vehicle type distinguishing method based on deep learning | |
| Ju et al. | Classification of jujube defects in small data sets based on transfer learning | |
| Yang et al. | Instance segmentation and classification method for plant leaf images based on ISC-MRCNN and APS-DCCNN | |
| CN114882215A (en) | Shape selection identification method for particle aggregate region of photoelectric coal gangue sorting image | |
| CN105718947B (en) | Lung Cancer Images sophisticated category method based on LBP and wavelet moment fusion feature | |
| CN114565798A (en) | Power device wear fault diagnosis method and system based on ferrographic image analysis | |
| Jing et al. | Island road centerline extraction based on a multiscale united feature | |
| Sui et al. | ECGAN: An improved conditional generative adversarial network with edge detection to augment limited training data for the classification of remote sensing images with high spatial resolution | |
| CN114220004A (en) | Artificial pasture land parcel identification method and system based on remote sensing image | |
| Chen et al. | Deep-learning-based road crack detection frameworks for dashcam-captured images under different illumination conditions | |
| CN113902978A (en) | Interpretable SAR image target detection method and system based on deep learning | |
| Bachhal et al. | Real-time disease detection system for maize plants using deep convolutional neural networks | |
| Wu et al. | MMFS: a grape disease recognition method based on multi-feature fusion and SVM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |