CN109740478A - Vehicle detection and recognition methods, device, computer equipment and readable storage medium storing program for executing - Google Patents
Vehicle detection and recognition methods, device, computer equipment and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN109740478A CN109740478A CN201811596741.5A CN201811596741A CN109740478A CN 109740478 A CN109740478 A CN 109740478A CN 201811596741 A CN201811596741 A CN 201811596741A CN 109740478 A CN109740478 A CN 109740478A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- detected
- image
- feature
- sliding window
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Image Analysis (AREA)
Abstract
The present invention is suitable for technical field of image processing, provides a kind of vehicle detection and recognition methods, device, computer equipment and readable storage medium storing program for executing, wherein the described method includes: acquisition current video image；According to the video image, the location information of vehicle to be detected is obtained, and intercepts vehicle image to be detected；According to the vehicle image to be detected, it is based on sliding window mechanism, the feature obtained in sliding window region determines vehicle overall appearance information to be detected in conjunction with preset data collection；According to the feature in the sliding window region, the local feature of vehicle to be detected is obtained, and the local feature of the vehicle to be detected is classified and identified；According to the vehicle image to be detected, passes through HSV model analysis and determine the automobile body color to be detected.The present invention can differentiate the specific brand of vehicle, model and color, can effectively mitigate existing manual processing effort, realize more efficient automated vehicle management objectives.
Description
Technical field
The invention belongs to technical field of image processing more particularly to a kind of vehicle detection and recognition methods, device, computers
Equipment and readable storage medium storing program for executing.
Background technique
With the fast development of Chinese society economy in recent years, China's vehicle year sales volume and ownership oneself leapt to world head
Position, but vehicle explosive growth takes place frequently but also relating to vehicle criminal case, and new challenge is brought to public safety.With city
Development and public safety demand, emerged in multitude using the traffic block port of video monitoring.Video monitoring is in inspection violating the regulations, accident
Escape chases, suspected vehicles distributed monitoring, and fake license plate vehicle identifies, improves traffic administration and investigation case handling efficiency etc. all has
There are very important realistic meaning and very wide application prospect.
However, the picture of bayonet monitoring at present is mainly by each category feature of vehicle needed for artificial interpretation, heavy workload and efficiency
Lowly；Or Recognition Algorithm of License Plate is relied only on, detection efficiency is low, and can not effectively detect unlicensed, deck or deliberately block vehicle
The vehicle of board, leads to missing inspection, increases vehicle management difficulty.
Summary of the invention
The embodiment of the present invention provides a kind of vehicle detection and recognition methods, it is intended to solve above-mentioned technical problem.
The embodiments of the present invention are implemented as follows, a kind of vehicle detection and recognition methods, which comprises
Vehicle identification request to be detected is responded, current video image is acquired；
According to the video image, the location information of vehicle to be detected is obtained, and intercepts vehicle image to be detected；
According to the vehicle image to be detected, it is based on sliding window mechanism, obtains the feature in sliding window region, in conjunction with
Preset data collection determines vehicle overall appearance information to be detected；
According to the feature in the sliding window region, the local feature of vehicle to be detected is obtained, and to described to be detected
The local feature of vehicle is classified and is identified；
According to the vehicle image to be detected, passes through HSV model analysis and determine the automobile body color to be detected；
According to vehicle overall appearance information, local feature and the body color to be detected, carried out with preset data collection
Matching exports the recognition result of the vehicle to be detected.
The embodiment of the present invention also provides a kind of vehicle detection and identification device, described device include:
Video acquisition unit acquires current video image for responding vehicle identification request to be detected；
Position determination unit, for obtaining the location information of vehicle to be detected, and intercept to be checked according to the video image
Survey vehicle image；
Appearance determination unit obtains sliding window for being based on sliding window mechanism according to the vehicle image to be detected
Feature in region determines vehicle overall appearance information to be detected in conjunction with preset data collection；
Tagsort and recognition unit, for obtaining vehicle to be detected according to the feature in the sliding window region
Local feature, and the local feature of the vehicle to be detected is classified and identified；
Color recognition unit, for passing through HSV model analysis and determining described to be checked according to the vehicle image to be detected
Measuring car body color；And
Output unit, it is and pre- for according to vehicle overall appearance information, local feature and the body color to be detected
If data set is matched, the recognition result of the vehicle to be detected is exported.
A kind of computer equipment provided in an embodiment of the present invention, which is characterized in that the computer equipment includes processor,
The step of processor is for realizing above-mentioned vehicle detection and recognition methods when executing the computer program stored in memory.
A kind of computer readable storage medium provided in an embodiment of the present invention, is stored thereon with computer program, feature
It is: realizes when the computer program is executed by processor such as the step of above-mentioned vehicle detection and recognition methods.
In embodiments of the present invention, according to current video image, the location information of vehicle to be detected is obtained, and is intercepted to be checked
Survey vehicle image；According to the vehicle image to be detected, it is based on sliding window mechanism, obtains the feature in sliding window region,
In conjunction with preset data collection, vehicle overall appearance information to be detected is determined；According to the feature in the sliding window region, obtain to
The local feature of vehicle is detected, and the local feature of the vehicle to be detected is classified and identified；According to described to be detected
Vehicle image passes through HSV model analysis and determines the automobile body color to be detected；It is whole outer according to the vehicle to be detected
Information, local feature and body color are seen, is matched with preset data collection, the identification knot of the vehicle to be detected is exported
Fruit；Compared with vehicle recongnition technique in the prior art, the present invention can sentence the specific brand of vehicle, model and color
Not, and the speed and precision of existing algorithm are improved, can effectively mitigates existing manual processing effort, while fortune can also be effectively reduced
Cost is sought, the gentle economic benefit of Automated water is improved, realizes more efficient automated vehicle management objectives.
Detailed description of the invention
Fig. 1 is the implementation flow chart of a kind of vehicle detection provided in an embodiment of the present invention and recognition methods；
Fig. 2 is the implementation flow chart of another vehicle detection provided in an embodiment of the present invention and recognition methods；
Fig. 3 is a kind of pyramid sampling flow diagram provided in an embodiment of the present invention；
Fig. 4 is that a kind of vehicle location provided in an embodiment of the present invention returns schematic diagram；
Fig. 5 is the implementation flow chart of another vehicle detection provided in an embodiment of the present invention and recognition methods；
Fig. 6 is a kind of structural schematic diagram of Faster-RCNN model provided in an embodiment of the present invention；
Fig. 7 is the implementation flow chart of another vehicle detection provided in an embodiment of the present invention and recognition methods；
Fig. 8 is a kind of vehicle detection of optimization provided in an embodiment of the present invention and the implementation flow chart of recognition methods；
Fig. 9 is the vehicle detection of another optimization provided in an embodiment of the present invention and the implementation flow chart of recognition methods；
Figure 10 is the structural schematic diagram of a kind of vehicle detection provided in an embodiment of the present invention and identification device.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, to this
Invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, not
For limiting the present invention.
The term used in embodiments of the present invention is only to be not intended to be limiting merely for for the purpose of describing particular embodiments
The present invention.Packet is also intended in the "an" and "the" of the embodiment of the present invention and singular used in the attached claims
Most forms are included, unless the context clearly indicates other meaning.It is also understood that term "and/or" used herein is
Refer to and includes that one or more associated any or all of project listed may combine.
The technical means and efficacy taken in order to which the present invention is further explained for the predetermined goal of the invention of realization, below in conjunction with
Attached drawing and preferred embodiment, to specific embodiment, structure, feature and its effect according to the present invention, detailed description are as follows.
In real life, the clue often known is very clear, for example oneself knows illegal vehicle model, color, Che Shang
Personnel characteristics, this clue of license plate easily lacks that (when such as case occurs, common people are difficult conscious to go to remember a vehicle instead
Board.License plate can be deliberately blocked before many criminal's crimes or uses personation, deck license plate, but it can not change vehicle easily
Shape and color).Therefore the embodiment of the present invention directly utilizes the existing image information of abundant excavation of computer rapidly and efficiently, realizes
The retrieval and identification of type of vehicle and feature determine vehicle brand, color, concrete model or even other details feature, this nothing
Doubt inspection violating the regulations, hit-and-run are chased, suspected vehicles distributed monitoring, fake license plate vehicle identifies, and accelerates criminal investigation and case detection
Efficiency and speed all have very important realistic meaning and very wide application prospect.Specifically, the present invention is quasi- from intelligence
Energy traffic system and traffic block port monitoring actual demand are set out, and the automotive vehicle feature inspection based on computer vision technique is studied
It surveys and identification technology improves treatment effeciency to mitigate artificial treatment burden, answered for other in public security prevention and control and intelligent transportation
With offer technical foundation and support.
A kind of vehicle detection provided in an embodiment of the present invention and recognition methods, by being obtained from video image collected
The location information of vehicle to be detected, and then determine overall appearance information, local feature and the body color of vehicle to be detected, with
Preset data collection is matched, and the recognition results such as model, the color of the vehicle to be detected are exported；With vehicle in the prior art
Identification technology is compared, and the present invention can differentiate the specific brand of vehicle, model and color, and improve the speed of existing algorithm
It with precision, can effectively mitigate existing manual processing effort, while operation cost can also be effectively reduced, it is gentle to improve Automated water
Economic benefit realizes more efficient automated vehicle management objectives.
Fig. 1 shows the implementation process of a kind of vehicle detection provided in an embodiment of the present invention and recognition methods, for the ease of
Illustrate, only show part related to the embodiment of the present invention, details are as follows:
In step s101, vehicle identification request to be detected is responded, current video image is acquired.
In embodiments of the present invention, video image can be sets in the regions such as current road or certain parking lots, square
The video image that the monitor or video camera set or camera etc. are produced；The response vehicle identification request to be detected, is adopted
Collect current video image, refer specifically to: responding vehicle identification request to be detected, when vehicle to be detected enters video monitoring regional,
Then acquire current video image.
In step s 102, according to the video image, the location information of vehicle to be detected is obtained, and intercepts measuring car to be checked
Image.
In embodiments of the present invention, video image can be divided into several frame image compositions, and in each frame video image
Vehicle location or vehicle size all can be inevitable distinct, i.e., vehicle to be detected possibly is present at any one of image
Position, and the size of vehicle in the picture may be arbitrary size.
In embodiments of the present invention, as shown in Fig. 2, shown step S102, specifically includes the following steps:
In step s 201, each frame video image is subjected to length and width classification diminution processing according to preset ratio.
In actual vehicle position fixing process, there is any positions that target vehicle possibly is present at image in image to be detected
It sets, and vehicle may be the problem of arbitrary size, the general multiple dimensioned exhaustive search method for using pyramid.
In embodiments of the present invention, due between image otherness it is very big, and include in each image vehicle number, Che great
Small and truck position is all different.In order to which system can be positioned in arbitrary image and identifies different size of vehicle, make
With based on pyramidal image sampling method.Firstly, chromatic image is carried out gray processing, become the image of 256 gray levels, if
The image of input has been gray level image then without gray processing processing, and most of texture that gray processing can both retain image is empty
Between etc. information, it is also possible that calculation amount substantially reduces；Pyramid as shown in Figure 3 samples flow diagram, inputs to a width
When image carries out pyramid sampling, need to carry out length and width classification according to a fixed ratio to input picture to reduce, until contracting
Length and width after small are less than fixed test window size.
In step S202, by default detection window, treated that video image sequentially carries out at slip scan to described
Reason, and obtain vehicle location coordinate information to be detected.
In embodiments of the present invention, the size for presetting detection window is arranged according to the big freedom in minor affairs of real image, only needs to protect
After card input picture carries out length and width classification diminution processing in proportion, the size of the detection window can be less than.It is described by pre-
If treated that video image sequentially carries out slip scan processing to described for detection window, and obtains vehicle location coordinate to be detected
Information, comprising: in the image of each ratio, sequentially carry out slip scan with the detection window of a constant size, record should
The position of window, meanwhile, do not stop to extract scanning and obtain window, scanning while identifies the window, judges that window is vehicle also right and wrong
Vehicle is input in deep neural network and is differentiated, record is identified as the window and the window's position of vehicle, so as to subsequent carry out vehicle
Positioning；Processing is merged to all windows for being identified as vehicle, last recognition result is obtained, according to the position of correlation window
Information positions vehicle.
In step S203, according to the vehicle location coordinate information to be detected, the vehicle location to be detected is fitted
The nonlinear function of information, and intercept vehicle image to be detected.
In embodiments of the present invention, after obtaining video image, the image of any one frame is taken, vehicle to be detected is likely to occur
In any one position of image, and the size of vehicle in the picture may be arbitrary size, thus proposed adoption to detection window into
Row Aspect Ratio arbitrarily scales and modifies network losses parameter, fits the non-linear function regression problem of vehicle location.
Vehicle location as shown in Figure 4 returns schematic diagram, and original image extracts convolution feature first, then big using 3*3
Small rectangle frame is scanned, and the characteristic pattern of 3*3 is one piece of bigger fixed area, is carried out respectively to length and width on this region
Scaling then can produce shown in right half of region of 9 different rectangular areas figures as above, to the different rectangular areas in upper figure
Object judgement prediction and rectangle frame position and size fine tuning are carried out, optimal position coordinates are exported, this also just solves bat
Act as regent the deformation problems of object caused by setting difference.
After searching out vehicle location, for the video image of acquisition, the time interval very little of every two frame can be with
Think that travel speed of the vehicle in this two frame is very slow, and assume that it moves with uniform velocity, therefore Kalman filter can be used
Wave device, state model are as follows:
Xt=AXt-1+Wt
Wherein, XtIt is the state of t moment vehicle, Xt-1It is the state of t-1 moment vehicle, A is sytem matrix, WtFor white Gaussian
Noise meets following formula:
P(Wt)~N (0, Q)
Wherein, Q is the covariance matrix of motion artifacts.
Define Xt=(xv, dxv, yv, dyv).Wherein, xv, dxv, yv, dyvThe vehicle center side x respectively in vehicle travel process
To measurement amount, measurement amount, the speed in speed and the direction y, the matrix of system are as follows:
Wherein, Δ t is the acquisition process time of sensing data.
The measurement model of Kalman filter are as follows:
Zt=CXt+Vt
Wherein, ZtFor observation vector, C is observing matrix vector, VtFor white Gaussian noise, meet following formula:
P(Vt)~N (0, R)
Wherein, R is the covariance matrix for measuring noise.
The observation vector of definition system is Zt=(xv, yv)T.Observing matrix is initialized as:
Given Initial state estimation error co-variance matrix is P0=I4×4, the covariance matrix of initial procedure system noise
Q0=I4×4, the covariance matrix of initial measurement noise is R0=I2×4.Wherein I is unit matrix.
In step s 103, according to the vehicle image to be detected, it is based on sliding window mechanism, obtains sliding window region
Interior feature determines vehicle overall appearance information to be detected in conjunction with preset data collection.
In embodiments of the present invention, preset data collection is expanded on the basis of existing Vehicle system data collection, is built
Vertical a set of more targeted, more complete vehicle detection and automatic identification data set, are mainly realized by following approach:
1) the CompCars data that domestic more perfect data set is established at present for Hong Kong Chinese University Tang Xiaoou team
Collection contains the image of 163 vehicle brands in data set, totally 1716 kinds of vehicles, 136727 vehicle vehicle pictures and
The picture of 27618 vehicle parts, in addition to this, data set also cover the various different perspectivess of vehicle, vehicle interior feature,
The picture of outside vehicle feature can be used for learning Vehicle Detail feature, such as the car door number of vehicle, vehicle window number etc..Number
It is formed according to all data are concentrated by obtaining on network and in monitor video.It can carry out vehicle training, vehicle brand instruction
Experienced and vehicle attribute training etc..
2) Internet resources are downloaded automatically by " crawler software ", since Internet resources are not necessarily accurate, under downloading
The picture come is not necessarily auto graph, so needing to carry out data cleansing to it, i.e., artificially tests to every picture.
3) it acquires and image is shot to the automobile of the scenes such as cell, highway.
In embodiments of the present invention, the specific aim and reasonability of data set directly influence the essence of detection and recognizer
Degree, picture number is bigger in training set, type is more, and the accuracy and robustness of algorithm are then stronger.
In embodiments of the present invention, as shown in figure 5, the step S103, specifically includes:
In step S501, greyscale transformation and normalized are carried out to the vehicle image to be detected.
In embodiments of the present invention, it is equal to specifically include size normalization, intensity profile normalization histogram for normalized
Weighing apparatusization, direction normalization；The reason of image need to being normalized, is: due to vehicle self character such as size, new and old
There is difference in degree, position etc., while distance, angle difference, the environment illumination intensity etc. when camera shooting sample are also deposited
In difference, all there is difference in obtained vehicle image, therefore carries out feature extraction to sample and identify that requiring one before returns
One changes preprocessing process.
In step S502, according to treated the vehicle image to be detected, it is based on sliding window mechanism, obtains sliding
The recurrence coordinate value of feature and corresponding region position in window area.
In embodiments of the present invention, according to treated the vehicle image to be detected, it is based on sliding window mechanism, is obtained
The recurrence coordinate value of feature and corresponding region position in sliding window region, specifically: it is based on Faster R-CNN network
Vehicle image to be measured is handled, Faster R-CNN model is with convolutional neural networks frame caffe
Based on (convolutional architecture for fast feature embedding), mainly by two moulds
Block is constituted: RPN candidate frame extraction module and FastR-CNN target (vehicle) detection module, model framework chart are as shown in Figure 6.RPN is
Full convolutional neural networks find out the position that vehicle is likely to occur in figure: FastR- for extracting the candidate frame of high quality in advance
CNN is detected based on the candidate frame that RPN is extracted and is identified the vehicle in candidate frame.By ImageNet pre-training model ZF most
A convolutional layer and two parallel full articulamentums are added after later layer convolution characteristic pattern to construct RPN.RPN is with original image
Whether the convolution characteristic pattern matrix that is extracted exports a series of rectangle candidate frame and the rectangle candidate region as input
Structural schematic diagram for the score of target, RPN is as shown in Figure 6.Sliding window mechanism is used in RPN, in convolutional neural networks
Increase a small sliding window on the characteristic pattern of the last one shared convolutional layer, the sliding window and input feature vector figure n × n are big
Small space connects entirely, each sliding window is mapped to the short amount an of low-dimensional, which is input to two
Parallel fully connected network network layers, the feature that a network layer is responsible for exporting in sliding window region belong to image background or mesh
Mark, another network layer are responsible for exporting the recurrence coordinate value of the regional location.
In step S503, according to the recurrence coordinate value of feature and corresponding region position in the dynamic window area,
Feature vector is formed, and dimensionality reduction is carried out to described eigenvector using Principal Component Analysis Algorithm.
In embodiments of the present invention, Principal Component Analysis Algorithm PCA (Principal Component Analysis) is one
The common data analysing method of kind.Initial data is transformed to the expression of one group of each dimension linear independence by linear transformation by PCA,
The main feature component that can be used for extracting data, is usually used in the dimensionality reduction of high dimensional data.
In embodiments of the present invention, vehicle body is extracted in conjunction with the methods of Gabor wavelet transformation and Faster-RCNN deep learning
Local feature (Haar feature, CNN feature, LBP feature, HOG feature etc.), and block statistics histogram is carried out, by histogram sequence
Column are jointly formed the feature vector to sample image description, finally using PCA algorithm to feature vector dimensionality reduction.According to above-mentioned, work as n
The sliding window of × n when being slided on convolution characteristic pattern matrix, each position of sliding original image correspond to k it is different
Anchor frame (candidate region in original image artificially assumed), therefore a full articulamentum exports 2 × k dimensional vector, corresponding k is a
The score of anchor frame target and background, another full articulamentum export 4 × k dimensional vector, indicate that k anchor frame corresponds to real goal frame
Transformation parameter.Each anchor frame respectively corresponds a kind of scale and length-width ratio centered on the center of current sliding window mouth.
In embodiments of the present invention, dimensionality reduction this means that information loss, to the loss function of an image is defined as:
Wherein: i is the candidate frame index chosen in a batch processing, piIt is the probability of target for candidate frame i.If candidate
Frame is a positive label, corresponding real estate labelIt is 1, otherwiseIt is 0.tiIndicate 4 ginsengs of the bounding box of prediction
Numberization coordinate, vector,It is the coordinate vector of corresponding real estate bounding box.Classification Loss Lcls, it is for two classes
The not logarithm loss of (target and non-targeted), is defined as:
It is lost for returning, is defined as:
Wherein smoothL1(x) are as follows:
For returning, using the parameter of 4 coordinates:
Wherein: (x, y) is the centre coordinate of the bounding box of prediction；(xa, ya) be candidate frame coordinate；(x*, y*) it is true
The coordinate of region bounding box；W and h is respectively the width and height of bounding box；NclsAnd NregFor normalized parameter；λ is balance factor.
In step S504, vehicle overall appearance letter to be detected is determined in conjunction with preset data collection according to described eigenvector
Breath.
In embodiments of the present invention, according to described eigenvector, in conjunction with preset data collection, determine that vehicle to be detected is whole outer
Information is seen, specifically: features described above vector is matched and is carried out with preset data collection the judgement of similarity；The preset data collection
Same as above, details are not described herein.
In step S104, according to the feature in the sliding window region, the local feature of vehicle to be detected is obtained, and
The local feature of the vehicle to be detected is classified and identified.
In embodiments of the present invention, as shown in fig. 7, the step S104, specifically includes the following steps:
In step s 701, according to the feature in the sliding window region, the local feature figure of vehicle to be detected is acquired
Picture.
In embodiments of the present invention, local feature image, which generally requires, uses textural characteristics, SIFT (Scale-
Invariant feature transform) feature, HOG (Histogram of Oriented Gradient) feature etc..
In step S702, by the local feature image be based on FasterR-CNN network pre-process, and obtain to
Detect the local feature of vehicle.
It is right in conjunction with preset vehicle local feature database by using support vector machines as classifier in step S703
The local feature of the vehicle to be detected is classified and is identified.
In embodiments of the present invention, support vector machines (Support Vector Machine, SVM) is a kind of new machine
Learning method, classification capacity are also significantly better than some conventional learning algorithms, can be used to solve classification, outlier detection, recurrence
And the problem of cluster etc., the algorithm show many peculiar in solution small sample, the identification of non-linear and high dimensional pattern
Advantage, quilt is simultaneously widely used in all kinds of pattern recognition problems.Using SVM classifier, any M class problem be can be broken into
(M-1) a two class identifies problem.
In embodiments of the present invention, svm classifier algorithm is developed from linear classifier.Linear classifier is
Refer to the boundary using a linear function as judgement, two class samples is distinguished, the classification method model is simple, is easy to count
It calculates.Linear discriminant function can be written as follow form:
Wx+b=0
W determines the direction of this hyperplane of decision boundary, and b determines the position of decision surface.Because meeting condition
There are many straight line, it is therefore desirable to design certain algorithm and find out optimal classification straight line, make at this point, supporting vector will be leaned on to play
With.Target is to find out an optimal separating hyper plane to separate sample set, and the hyperplane for separating sample set is possible to more than one,
An optimal hyperlane should be found thus, which can not only correctly classify two class sample points, but also should to separate
Sample point it is maximum to the distance of hyperplane or interval.In order to maximize interval, two parallel hyperplane are defined.
Wx+b=1
Wx+b=-1
To guarantee that no training sample appears among the two hyperplane, for all sample point xiMust satisfy with
Lower inequality.
ωi(w·x+b)≥1
Distance between the two hyperplane is 2/ | | w | |, to maximize interval, need to minimize | | w | |, and make its satisfaction
On, formula condition.This is the quadratic programming problem of a belt restraining in fact, is a convex problem, does not have locally optimal solution.For
This can be solved with method of Lagrange multipliers, Lagrangian objective function is defined as:
Wherein aiIt is Lagrange multiplier, to allow L to minimize about w, b, local derviation can be sought to it:
Bringing two formula into former formula again can be obtained sample point xi, xiAnd correspond to classification or label Wi, WjBetween pass
System:
The final formula for needing to optimize known to as a result, is:
In embodiments of the present invention, preset vehicle local feature database includes that such as logo, car light, vehicle bumper are each
Vehicle local message.
In embodiments of the present invention, on the basis of the vehicle characteristics above-mentioned based on FasterR-CNN network detect, increase
Support vector machines (Support Vector Machine, SVM) (such as logo, car light, bumper and keeps out the wind to vehicle local feature
Glass etc.) classified and is identified.
In step s105, according to the vehicle image to be detected, pass through HSV model analysis and determine the measuring car to be checked
Body color.
In embodiments of the present invention, HSV (Hue, Saturation, Value) is created according to the intuitive nature of color
A kind of color space, also referred to as hexagonal pyramid model.The parameter of color is respectively in this model: tone (H), and saturation degree (S) is bright
It spends (V)；Wherein, tone (H) refers to is measured with angle, and value range is 0 °~360 °, is counted counterclockwise since red
It calculates, red is 0 °, and green is 120 °, and blue is 240 °, and complementary color is: yellow is 60 °, and cyan is 180 °, and magenta is 300 °；Saturation
Degree S refers to that value range is 0.0~1.0, and value is bigger, and color is more saturated；Brightness V refers to that value range is 0 (black)~255
(white).
In embodiments of the present invention, as shown in figure 8, the step S105, specifically includes the following steps:
In step S801, the vehicle image to be detected is subjected to greyscale transform process.
In embodiments of the present invention, vehicle image to be detected is subjected to greyscale transform process, it is made to contain only the bright of image
Information is spent, to save the processing time of detection, while image raw information can also be retained to the maximum extent, it is specific to convert
Are as follows:
Cmax (x, y)=max (r (x, y), g (x, y), b (x, y))
Cmin (x, y)=min (r (x, y), g (x, y), b (x, y))
Define the RGB maximum value and minimum value of each point in image.
γ (x, y)=cmax (x, y)-cmin (x, y)
γ (x, y) indicates the gray scale of image slices vegetarian refreshments, to entire gray level image linear stretch, is conducive to image point in this way
It cuts.
In step S802, by thresholding method, treated that vehicle image to be detected is split processing by described,
And identify the RGB feature of car engine cover area to be detected.
In embodiments of the present invention, thresholding method is maximum variance between clusters (OTSU), and OTSU is mainly according to figure
As the gamma characteristic of itself, finds maximum between-cluster variance and divide the image into background and target.For image I (x, y), target and back
The segmentation threshold value of scape is denoted as T, and the ratio that the pixel number of target accounts for entire image is denoted as w0, average gray μ0；Background pixel point
The ratio that number accounts for entire image is w1, average gray μ1；The overall average gray scale of image is denoted as μ, and inter-class variance is denoted as g.General
Have in rate statistics:
μ=w0μ0+w1μ1
The definition such as formula of its inter-class variance g:
G=w0(μ1-μ)+w1(μ0-μ)
Following formula can be obtained:
G=w0w1(μ0-μ1)
When variance g maximum, it is believed that foreground and background difference is maximum at this time, and gray scale T at this time is optimal threshold, and
Image segmentation is carried out with it.
The RGB information of vehicle body is extracted using cut zone, Cregion is used as segmentation vehicle body region, if image x, y point
Rgb value are as follows:
Rgb space structure does not simultaneously meet people to the subjective judgement of color similarity, and HSV more meets people to the master of color
See understanding.The RGB color of image is converted into hsv color space.
TR is transfer function of the RGB to hsv color space.
In step S803, according to the RGB feature, based on HSV model analysis and corresponding color is identified, by display frequency
The maximum color of number is determined as automobile body color to be detected.
In practical applications, rgb space structure and people are not met to the subjective judgement of color similarity, HSV more meets
Subjective understanding of the people to color, so the RGB color of image is converted to hsv color space, finally to HSV model into
Row analysis, can show the frequency that various colors is shown using histogram, select mass-tone of the maximum color of frequency as image,
Using mass-tone as the foundation for judging vehicle color.
In step s 106, and pre- according to vehicle overall appearance information, local feature and the body color to be detected
If data set is matched, the recognition result of the vehicle to be detected is exported.
In embodiments of the present invention, as shown in figure 9, the step S106, specifically includes the following steps:
In step S901, result is exported according to the vehicle overall appearance information to be detected, the confidence level of local feature
The sum of, determine vehicle model to be detected.
In step S902, according to the vehicle model to be detected and body color, concentrated with preset data corresponding
Information is matched, and the recognition result of the vehicle to be detected is exported.
A kind of vehicle detection provided in an embodiment of the present invention and recognition methods, by being obtained from video image collected
The location information of vehicle to be detected, and then determine overall appearance information, local feature and the body color of vehicle to be detected, with
Preset data collection is matched, and the recognition results such as model, the color of the vehicle to be detected are exported；With vehicle in the prior art
Identification technology is compared, and the present invention can differentiate the specific brand of vehicle, model and color, and based on Faster R-CNN and
SVM even depth learning ways substantially increase the speed and precision of existing algorithm, can effectively mitigate existing artificial treatment work
Amount, while operation cost can also be effectively reduced, the gentle economic benefit of Automated water is improved, realizes more efficient automated vehicle pipe
Manage target.
Figure 10 shows the structure of a kind of vehicle detection provided in an embodiment of the present invention and identification device, for the ease of saying
It is bright, part related to the embodiment of the present invention is only shown, details are as follows:
Vehicle detection and identification device, including video acquisition unit 100, position determination unit 200, appearance determination unit
300, tagsort and recognition unit 400, color recognition unit 500 and output unit 600.
Video acquisition unit 100 acquires current video image for responding vehicle identification request to be detected.
In embodiments of the present invention, video acquisition unit 100 works as forward sight for responding vehicle identification request to be detected, acquisition
Frequency image；Wherein, the video image can be the monitoring being arranged in the regions such as current road or certain parking lots, square
The video image that device or video camera or camera etc. are produced；Forward sight is worked as in the response vehicle identification request to be detected, acquisition
Frequency image, refers specifically to: responding vehicle identification request to be detected, when vehicle to be detected enters video monitoring regional, then acquisition is worked as
Preceding video image.
Position determination unit 200, for obtaining the location information of vehicle to be detected, and intercept according to the video image
Vehicle image to be detected.
In embodiments of the present invention, position determination unit 200 is used to obtain vehicle to be detected according to the video image
Location information, and intercept vehicle image to be detected；Wherein, the video image can be divided into several frame image compositions, and each
Vehicle location or vehicle size in frame video image all can be inevitable distinct, i.e., vehicle to be detected possibly is present at
Any one position of image, and the size of vehicle in the picture may be arbitrary size.
Appearance determination unit 300 obtains sliding for being based on sliding window mechanism according to the vehicle image to be detected
Feature in window area determines vehicle overall appearance information to be detected in conjunction with preset data collection.
In embodiments of the present invention, appearance determination unit 300 is used to be based on sliding window according to the vehicle image to be detected
Mouth mechanism, the feature obtained in sliding window region determine vehicle overall appearance information to be detected in conjunction with preset data collection；Its
In, preset data collection is expanded on the basis of existing Vehicle system data collection, foundation it is a set of more targetedly, it is completeer
Whole vehicle detection and automatic identification data set；It is described that sliding window mechanism is based on according to the vehicle image to be detected, it obtains
Feature in sliding window region determines vehicle overall appearance information to be detected, specifically includes: to institute in conjunction with preset data collection
It states vehicle image to be detected and carries out greyscale transformation and normalized；According to treated the vehicle image to be detected, it is based on
Sliding window mechanism obtains the feature in sliding window region and the recurrence coordinate value of corresponding region position；According to described dynamic
The recurrence coordinate value of feature and corresponding region position in window area is formed feature vector, and is calculated using principal component analysis
Method carries out dimensionality reduction to described eigenvector；According to described eigenvector, in conjunction with preset data collection, determine that vehicle to be detected is whole outer
See information.
Tagsort and recognition unit 400, for obtaining vehicle to be detected according to the feature in the sliding window region
Local feature, and the local feature of the vehicle to be detected is classified and is identified.
In embodiments of the present invention, tagsort and recognition unit 400 are used for according to the spy in the sliding window region
Sign, obtains the local feature of vehicle to be detected, and the local feature of the vehicle to be detected is classified and identified；Described
According to the feature in the sliding window region, the local feature of vehicle to be detected is obtained, and to the part of the vehicle to be detected
Feature is classified and is identified, specifically includes: according to the feature in the sliding window region, acquiring the part of vehicle to be detected
Characteristic image；The local feature image is based on Faster R-CNN network to pre-process, and obtains vehicle to be detected
Local feature；By using support vector machines as classifier, in conjunction with preset vehicle local feature database, to the measuring car to be checked
Local feature classified and identified.
Color recognition unit 500, for passing through described in HSV model analysis and determination according to the vehicle image to be detected
Automobile body color to be detected.
In embodiments of the present invention, color recognition unit 500 is used to pass through HSV mould according to the vehicle image to be detected
Type analysis simultaneously determines the automobile body color to be detected；It is described according to the vehicle image to be detected, pass through HSV model analysis
And determine the automobile body color to be detected, it specifically includes: the vehicle image to be detected is subjected to greyscale transform process；It is logical
Crossing thresholding method, treated that vehicle image to be detected is split processing by described, and identifies car engine cover region to be detected
The RGB feature in domain；According to the RGB feature, based on HSV model analysis and corresponding color is identified, will show that frequency is maximum
Color is determined as automobile body color to be detected.
Output unit 600 is used for according to vehicle overall appearance information, local feature and the body color to be detected,
It is matched with preset data collection, exports the recognition result of the vehicle to be detected.
In embodiments of the present invention, output unit 600 is used for according to vehicle overall appearance information, the locally spy to be detected
Sign and body color, are matched with preset data collection, export the recognition result of the vehicle to be detected；It is described according to
Vehicle overall appearance information, local feature and body color to be detected, are matched with preset data collection, are exported described to be checked
The recognition result of measuring car, specifically includes: being exported according to the vehicle overall appearance information to be detected, the confidence level of local feature
The sum of as a result, determine vehicle model to be detected；According to the vehicle model to be detected and body color, concentrated with preset data
Corresponding informance matched, export the recognition result of the vehicle to be detected.
A kind of vehicle detection provided in an embodiment of the present invention and identification device, by being obtained from video image collected
The location information of vehicle to be detected, and then determine overall appearance information, local feature and the body color of vehicle to be detected, with
Preset data collection is matched, and the recognition results such as model, the color of the vehicle to be detected are exported；With vehicle in the prior art
Identification technology is compared, and the present invention can differentiate the specific brand of vehicle, model and color, and based on FasterR-CNN and
SVM even depth learning ways substantially increase the speed and precision of existing algorithm, can effectively mitigate existing artificial treatment work
Amount, while operation cost can also be effectively reduced, the gentle economic benefit of Automated water is improved, realizes more efficient automated vehicle pipe
Manage target.
The embodiment of the invention also provides a kind of computer equipment, which includes processor, and processor is used for
Vehicle detection and recognition methods that above-mentioned each embodiment of the method provides are realized when executing the computer program stored in memory
The step of.
The embodiments of the present invention also provide a kind of computer readable storage medium, it is stored thereon with computer program/refer to
Enable, the computer program/instruction realized when being executed by above-mentioned processor above-mentioned each embodiment of the method offer vehicle detection and
The step of recognition methods.
Illustratively, computer program can be divided into one or more modules, one or more module is stored
In memory, and by processor it executes, to complete the present invention.One or more modules, which can be, can complete specific function
Series of computation machine program instruction section, the instruction segment is for describing implementation procedure of the computer program in computer equipment.Example
Such as, the computer program can be divided into the step of vehicle detection and recognition methods that above-mentioned each embodiment of the method provides
Suddenly.
It will be understood by those skilled in the art that the description of above-mentioned computer equipment is only example, do not constitute to calculating
The restriction of machine equipment may include component more more or fewer than foregoing description, perhaps combine certain components or different portions
Part, such as may include input-output equipment, network access equipment, bus etc..
Alleged processor can be central processing unit (Central ProcessingUnit, CPU), can also be other
General processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng the processor is the control centre of the computer equipment, utilizes various interfaces and the entire user terminal of connection
Various pieces.
The memory can be used for storing the computer program and/or module, and the processor is by operation or executes
Computer program in the memory and/or module are stored, and calls the data being stored in memory, described in realization
The various functions of computer equipment.The memory can mainly include storing program area and storage data area, wherein storage program
It area can application program (such as sound-playing function, image player function etc.) needed for storage program area, at least one function
Deng；Storage data area, which can be stored, uses created data (such as audio data, phone directory etc.) etc. according to mobile phone.In addition,
Memory may include high-speed random access memory, can also include nonvolatile memory, such as hard disk, memory, grafting
Formula hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card
(Flash Card), at least one disk memory, flush memory device or other volatile solid-state parts.
If the integrated module/unit of the computer equipment is realized in the form of SFU software functional unit and as independent
Product when selling or using, can store in a computer readable storage medium.Based on this understanding, the present invention is real
All or part of the process in existing above-described embodiment method, can also instruct relevant hardware come complete by computer program
At the computer program can be stored in a computer readable storage medium, which is being executed by processor
When, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program code, described
Computer program code can be source code form, object identification code form, executable file or certain intermediate forms etc..The meter
Calculation machine readable medium may include: can carry the computer program code any entity or device, recording medium, USB flash disk,
Mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), random access memory
Device (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (10)
1. a kind of vehicle detection and recognition methods, which is characterized in that the described method includes:
Vehicle identification request to be detected is responded, current video image is acquired；
According to the video image, the location information of vehicle to be detected is obtained, and intercepts vehicle image to be detected；
According to the vehicle image to be detected, it is based on sliding window mechanism, obtains the feature in sliding window region, in conjunction with default
Data set determines vehicle overall appearance information to be detected；
According to the feature in the sliding window region, the local feature of vehicle to be detected is obtained, and to the vehicle to be detected
Local feature classified and identified；
According to the vehicle image to be detected, passes through HSV model analysis and determine the automobile body color to be detected；
According to vehicle overall appearance information, local feature and the body color to be detected, matched with preset data collection,
Export the recognition result of the vehicle to be detected.
2. vehicle detection as described in claim 1 and recognition methods, which is characterized in that the response vehicle identification to be detected is asked
It asks, acquires current video image, specifically include:
Vehicle identification request to be detected is responded, judges whether the vehicle to be detected enters video monitoring regional；
When the vehicle to be detected enters video monitoring regional, then current video image is acquired.
3. vehicle detection as described in claim 1 and recognition methods, which is characterized in that it is described according to the video image, it obtains
The location information of vehicle to be detected is taken, and intercepts vehicle image to be detected, is specifically included:
Each frame video image is subjected to length and width classification diminution processing according to preset ratio；
By default detection window, treated that video image sequentially carries out slip scan processing to described, and obtains measuring car to be checked
Location coordinate information；
According to the vehicle location coordinate information to be detected, the nonlinear function of the vehicle position information to be detected is fitted,
And intercept vehicle image to be detected.
4. vehicle detection as described in claim 1 and recognition methods, which is characterized in that described according to the vehicle figure to be detected
Picture is based on sliding window mechanism, and the feature obtained in sliding window region determines that vehicle to be detected is whole in conjunction with preset data collection
Body appearance information, specifically includes:
Greyscale transformation and normalized are carried out to the vehicle image to be detected；
According to treated the vehicle image to be detected, it is based on sliding window mechanism, obtains the feature in sliding window region
And the recurrence coordinate value of corresponding region position；
According to the recurrence coordinate value of feature and corresponding region position in the dynamic window area, feature vector is formed, and is adopted
Dimensionality reduction is carried out to described eigenvector with Principal Component Analysis Algorithm；
Vehicle overall appearance information to be detected is determined in conjunction with preset data collection according to described eigenvector.
5. vehicle detection as described in claim 1 and recognition methods, which is characterized in that described according to the sliding window region
Interior feature obtains the local feature of vehicle to be detected, and the local feature of the vehicle to be detected is classified and identified,
It specifically includes:
According to the feature in the sliding window region, the local feature image of vehicle to be detected is acquired；
The local feature image is based on Faster R-CNN network to pre-process, and the part for obtaining vehicle to be detected is special
Sign；
By using support vector machines as classifier, in conjunction with preset vehicle local feature database, to the vehicle to be detected
Local feature is classified and is identified.
6. vehicle detection as described in claim 1 and recognition methods, which is characterized in that described according to the vehicle figure to be detected
Picture passes through HSV model analysis and determines the automobile body color to be detected, specifically includes:
The vehicle image to be detected is subjected to greyscale transform process；
By thresholding method, treated that vehicle image to be detected is split processing by described, and identifies that vehicle to be detected draws
Hold up the RGB feature of cover area；
According to the RGB feature, based on HSV model analysis and corresponding color is identified, will show that the maximum color of frequency determines
For automobile body color to be detected.
7. vehicle detection as described in claim 1 and recognition methods, which is characterized in that described whole according to the vehicle to be detected
Body appearance information, local feature and body color are matched with preset data collection, export the knowledge of the vehicle to be detected
Not as a result, specifically including:
According to the sum of the vehicle overall appearance information to be detected, the confidence level of local feature output result, measuring car to be checked is determined
Model；
According to the vehicle model to be detected and body color, the corresponding informance concentrated with preset data is matched, and is exported
The recognition result of the vehicle to be detected.
8. a kind of vehicle detection and identification device, which is characterized in that described device includes:
Video acquisition unit acquires current video image for responding vehicle identification request to be detected；
Position determination unit, for obtaining the location information of vehicle to be detected, and intercept measuring car to be checked according to the video image
Image；
Appearance determination unit obtains sliding window region for being based on sliding window mechanism according to the vehicle image to be detected
Interior feature determines vehicle overall appearance information to be detected in conjunction with preset data collection；
Tagsort and recognition unit, for obtaining the part of vehicle to be detected according to the feature in the sliding window region
Feature, and the local feature of the vehicle to be detected is classified and identified；
Color recognition unit, for passing through HSV model analysis and determining the measuring car to be checked according to the vehicle image to be detected
Body color；And
Output unit is used for according to vehicle overall appearance information, local feature and the body color to be detected, with present count
It is matched according to collection, exports the recognition result of the vehicle to be detected.
9. a kind of computer equipment, which is characterized in that the computer equipment includes processor, and the processor is deposited for executing
It is realized when the computer program stored in reservoir such as the step of any one of claim 1-7 the method.
10. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that: the computer program
It is realized when being executed by processor such as the step of any one of claim 1-7 the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596741.5A CN109740478B (en) | 2018-12-26 | 2018-12-26 | Vehicle detection and identification method, device, computer equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596741.5A CN109740478B (en) | 2018-12-26 | 2018-12-26 | Vehicle detection and identification method, device, computer equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740478A true CN109740478A (en) | 2019-05-10 |
| CN109740478B CN109740478B (en) | 2023-04-28 |
Family
ID=66359964
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811596741.5A Active CN109740478B (en) | 2018-12-26 | 2018-12-26 | Vehicle detection and identification method, device, computer equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740478B (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110322522A (en) * | 2019-07-11 | 2019-10-11 | 山东领能电子科技有限公司 | A kind of vehicle color identification method based on the interception of target identification region |
| CN110378381A (en) * | 2019-06-17 | 2019-10-25 | 华为技术有限公司 | Object detecting method, device and computer storage medium |
| CN110675373A (en) * | 2019-09-12 | 2020-01-10 | 珠海格力智能装备有限公司 | Component installation detection method, device and system |
| CN110751053A (en) * | 2019-09-26 | 2020-02-04 | 高新兴科技集团股份有限公司 | Vehicle color identification method, device, equipment and storage medium |
| CN111144372A (en) * | 2019-12-31 | 2020-05-12 | 上海眼控科技股份有限公司 | Vehicle detection method, device, computer equipment and storage medium |
| CN111191539A (en) * | 2019-12-20 | 2020-05-22 | 江苏常熟农村商业银行股份有限公司 | Certificate authenticity verification method and device, computer equipment and storage medium |
| CN111340775A (en) * | 2020-02-25 | 2020-06-26 | 湖南大学 | Parallel method and device for acquiring ultrasonic standard tangent plane and computer equipment |
| CN111353444A (en) * | 2020-03-04 | 2020-06-30 | 上海眼控科技股份有限公司 | Marker lamp monitoring method and device, computer equipment and storage medium |
| CN111982415A (en) * | 2019-05-24 | 2020-11-24 | 杭州海康威视数字技术股份有限公司 | Pipeline leakage detection method and device |
| CN112016502A (en) * | 2020-09-04 | 2020-12-01 | 平安国际智慧城市科技股份有限公司 | Safety belt detection method and device, computer equipment and storage medium |
| CN112036421A (en) * | 2019-05-16 | 2020-12-04 | 搜狗（杭州）智能科技有限公司 | Image processing method and device and electronic equipment |
| CN112101282A (en) * | 2020-09-25 | 2020-12-18 | 北京瞰天科技有限公司 | Aquatic target identification method and device, electronic equipment and storage medium |
| CN112164223A (en) * | 2020-02-27 | 2021-01-01 | 周国霞 | Intelligent traffic information processing method and device based on cloud platform |
| CN114648713A (en) * | 2020-12-18 | 2022-06-21 | 北京字节跳动网络技术有限公司 | Video classification method and device, electronic equipment and computer-readable storage medium |
| CN115626159A (en) * | 2021-07-01 | 2023-01-20 | 信扬科技(佛山)有限公司 | Vehicle warning system and method and automobile |
| CN115880565A (en) * | 2022-12-06 | 2023-03-31 | 江苏凤火数字科技有限公司 | Neural network-based scraped vehicle identification method and system |
| CN117269180A (en) * | 2023-11-24 | 2023-12-22 | 成都数之联科技股份有限公司 | Vehicle appearance detection method, device, server and computer readable storage medium |
| CN117392857A (en) * | 2023-10-13 | 2024-01-12 | 深圳市平安顺科技有限公司 | Truck type recognition system and recognition method based on Bluetooth network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184393A (en) * | 2011-06-20 | 2011-09-14 | 苏州两江科技有限公司 | Method for judging automobile type according to license plate recognition |
| CN107730905A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Multitask fake license plate vehicle vision detection system and method based on depth convolutional neural networks |
| CN107798335A (en) * | 2017-08-28 | 2018-03-13 | 浙江工业大学 | A kind of automobile logo identification method for merging sliding window and Faster R CNN convolutional neural networks |
| CN108875754A (en) * | 2018-05-07 | 2018-11-23 | 华侨大学 | A kind of vehicle recognition methods again based on more depth characteristic converged network |
-
2018
- 2018-12-26 CN CN201811596741.5A patent/CN109740478B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184393A (en) * | 2011-06-20 | 2011-09-14 | 苏州两江科技有限公司 | Method for judging automobile type according to license plate recognition |
| CN107730905A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Multitask fake license plate vehicle vision detection system and method based on depth convolutional neural networks |
| CN107798335A (en) * | 2017-08-28 | 2018-03-13 | 浙江工业大学 | A kind of automobile logo identification method for merging sliding window and Faster R CNN convolutional neural networks |
| CN108875754A (en) * | 2018-05-07 | 2018-11-23 | 华侨大学 | A kind of vehicle recognition methods again based on more depth characteristic converged network |
Non-Patent Citations (3)
| Title |
|---|
| 于明月: ""基于属性的车辆检索算法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 张飞云: ""基于深度学习的车辆定位及车型识别研究"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 顾人舒: ""交通卡口车辆检测与自动识别技术研究"", 《中国优秀硕士论文全文数据库 信息科技辑》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036421A (en) * | 2019-05-16 | 2020-12-04 | 搜狗（杭州）智能科技有限公司 | Image processing method and device and electronic equipment |
| CN111982415A (en) * | 2019-05-24 | 2020-11-24 | 杭州海康威视数字技术股份有限公司 | Pipeline leakage detection method and device |
| CN110378381A (en) * | 2019-06-17 | 2019-10-25 | 华为技术有限公司 | Object detecting method, device and computer storage medium |
| CN110378381B (en) * | 2019-06-17 | 2024-01-19 | 华为技术有限公司 | Object detection method, device and computer storage medium |
| CN110322522A (en) * | 2019-07-11 | 2019-10-11 | 山东领能电子科技有限公司 | A kind of vehicle color identification method based on the interception of target identification region |
| CN110322522B (en) * | 2019-07-11 | 2023-06-16 | 山东领能电子科技有限公司 | Vehicle color recognition method based on target recognition area interception |
| CN110675373A (en) * | 2019-09-12 | 2020-01-10 | 珠海格力智能装备有限公司 | Component installation detection method, device and system |
| CN110751053A (en) * | 2019-09-26 | 2020-02-04 | 高新兴科技集团股份有限公司 | Vehicle color identification method, device, equipment and storage medium |
| CN110751053B (en) * | 2019-09-26 | 2022-02-22 | 高新兴科技集团股份有限公司 | Vehicle color identification method, device, equipment and storage medium |
| CN111191539B (en) * | 2019-12-20 | 2021-01-29 | 江苏常熟农村商业银行股份有限公司 | Certificate authenticity verification method and device, computer equipment and storage medium |
| CN111191539A (en) * | 2019-12-20 | 2020-05-22 | 江苏常熟农村商业银行股份有限公司 | Certificate authenticity verification method and device, computer equipment and storage medium |
| CN111144372A (en) * | 2019-12-31 | 2020-05-12 | 上海眼控科技股份有限公司 | Vehicle detection method, device, computer equipment and storage medium |
| CN111340775B (en) * | 2020-02-25 | 2023-09-29 | 湖南大学 | Parallel method, device and computer equipment for acquiring ultrasonic standard section |
| CN111340775A (en) * | 2020-02-25 | 2020-06-26 | 湖南大学 | Parallel method and device for acquiring ultrasonic standard tangent plane and computer equipment |
| CN112164223B (en) * | 2020-02-27 | 2022-04-29 | 浙江恒隆智慧科技集团有限公司 | Intelligent traffic information processing method and device based on cloud platform |
| CN112164223A (en) * | 2020-02-27 | 2021-01-01 | 周国霞 | Intelligent traffic information processing method and device based on cloud platform |
| CN111353444A (en) * | 2020-03-04 | 2020-06-30 | 上海眼控科技股份有限公司 | Marker lamp monitoring method and device, computer equipment and storage medium |
| CN112016502A (en) * | 2020-09-04 | 2020-12-01 | 平安国际智慧城市科技股份有限公司 | Safety belt detection method and device, computer equipment and storage medium |
| CN112016502B (en) * | 2020-09-04 | 2023-12-26 | 平安国际智慧城市科技股份有限公司 | Safety belt detection method, safety belt detection device, computer equipment and storage medium |
| CN112101282A (en) * | 2020-09-25 | 2020-12-18 | 北京瞰天科技有限公司 | Aquatic target identification method and device, electronic equipment and storage medium |
| CN112101282B (en) * | 2020-09-25 | 2024-04-26 | 北京瞰天科技有限公司 | Water target identification method and device, electronic equipment and storage medium |
| CN114648713A (en) * | 2020-12-18 | 2022-06-21 | 北京字节跳动网络技术有限公司 | Video classification method and device, electronic equipment and computer-readable storage medium |
| CN115626159A (en) * | 2021-07-01 | 2023-01-20 | 信扬科技(佛山)有限公司 | Vehicle warning system and method and automobile |
| CN115880565B (en) * | 2022-12-06 | 2023-09-05 | 江苏凤火数字科技有限公司 | Neural network-based scraped vehicle identification method and system |
| CN115880565A (en) * | 2022-12-06 | 2023-03-31 | 江苏凤火数字科技有限公司 | Neural network-based scraped vehicle identification method and system |
| CN117392857A (en) * | 2023-10-13 | 2024-01-12 | 深圳市平安顺科技有限公司 | Truck type recognition system and recognition method based on Bluetooth network |
| CN117269180A (en) * | 2023-11-24 | 2023-12-22 | 成都数之联科技股份有限公司 | Vehicle appearance detection method, device, server and computer readable storage medium |
| CN117269180B (en) * | 2023-11-24 | 2024-03-12 | 成都数之联科技股份有限公司 | Vehicle appearance detection method, device, server and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740478B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109740478A (en) | Vehicle detection and recognition methods, device, computer equipment and readable storage medium storing program for executing | |
| CN109993056B (en) | Method, server and storage medium for identifying vehicle illegal behaviors | |
| Jia et al. | Region-based license plate detection | |
| US9842266B2 (en) | Method for detecting driver cell phone usage from side-view images | |
| JP4723582B2 (en) | Traffic sign detection method | |
| US7734097B1 (en) | Detecting objects in images with covariance matrices | |
| CN104091147B (en) | A kind of near-infrared eyes positioning and eye state identification method | |
| Wang et al. | An effective method for plate number recognition | |
| CN105404886B (en) | Characteristic model generation method and characteristic model generating means | |
| Derpanis et al. | Classification of traffic video based on a spatiotemporal orientation analysis | |
| CN101715111B (en) | Method for automatically searching abandoned object in video monitoring | |
| US20090148046A1 (en) | Error propogation and variable-bandwidth mean shift for feature space analysis | |
| CN102509112A (en) | Number plate identification method and identification system thereof | |
| CN107194393B (en) | Method and device for detecting temporary license plate | |
| CN106778742B (en) | Car logo detection method based on Gabor filter background texture suppression | |
| KR101178333B1 (en) | A method for detecting human using wavelet-based cs-lbp and a cascade of random forests | |
| Nurhadiyatna et al. | Gabor filtering for feature extraction in real time vehicle classification system | |
| Liu et al. | Multi-type road marking recognition using adaboost detection and extreme learning machine classification | |
| CN106557740A (en) | The recognition methods of oil depot target in a kind of remote sensing images | |
| Li et al. | Multi-view vehicle detection based on fusion part model with active learning | |
| CN108073940A (en) | A kind of method of 3D object instance object detections in unstructured moving grids | |
| CN107301421A (en) | The recognition methods of vehicle color and device | |
| Shang et al. | A novel method for vehicle headlights detection using salient region segmentation and PHOG feature | |
| Yousefi et al. | Water-body segmentation in satellite imagery applying modified kernel kmeans | |
| Zheng et al. | Shadow removal for pedestrian detection and tracking in indoor environments |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20230410 Address after: No. 17, Unit 1, Building 42, Chenguang Community, Fushan District, Yantai City, Shandong Province, 264006 Applicant after: Yang Xianming Address before: No.1 Lanhai Road, hi tech Zone, Yantai City, Shandong Province Applicant before: SHANDONG CHUANGKE AUTOMATION TECHNOLOGY Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |