CN108681707A - Wide-angle model recognizing method and system based on global and local Fusion Features - Google Patents

Wide-angle model recognizing method and system based on global and local Fusion Features Download PDF

Info

Publication number
CN108681707A
CN108681707A CN201810463134.5A CN201810463134A CN108681707A CN 108681707 A CN108681707 A CN 108681707A CN 201810463134 A CN201810463134 A CN 201810463134A CN 108681707 A CN108681707 A CN 108681707A
Authority
CN
China
Prior art keywords
vehicle
image
feature
global
image block
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201810463134.5A
Other languages
Chinese (zh)
Inventor
蒋行国
万今朝
李海鸥
李琦
张法碧
肖功利
陈永和
傅涛
孙堂友
蔡晓东
苏欣欣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guilin University of Electronic Technology
Original Assignee
Guilin University of Electronic Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guilin University of Electronic Technology filed Critical Guilin University of Electronic Technology
Priority to CN201810463134.5A priority Critical patent/CN108681707A/en
Publication of CN108681707A publication Critical patent/CN108681707A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/08—Detecting or categorising vehicles

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Multimedia (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

The present invention discloses a kind of wide-angle model recognizing method and system based on global and local Fusion Features, the image comprising vehicle is intercepted out when vehicle passes through vehicle cab recognition acquisition zone, the image comprising vehicle of interception is cut first, the vehicle pictures of removal complex background are obtained, vehicle pictures are divided into vehicle face image piecemeal, tailstock image block and wheel image block.Vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image block are imported into deep layer multiple-limb convolutional neural networks the global and local feature to vehicle again and carry out feature training, and vehicle pictures feature and each blocking characteristic are subjected to Fusion Features.Classification and Identification is carried out to the feature after fusion by grader afterwards.The present invention merges the global and local feature of wide-angle vehicle, can significantly improve the accuracy rate of wide-angle vehicle cab recognition.

Description

Wide-angle model recognizing method and system based on global and local Fusion Features
Technical field
The present invention relates to intelligent identification technology fields, and in particular to a kind of wide-angle based on global and local Fusion Features Model recognizing method and system.
Background technology
Vehicle model identification technology has become the needs of social development, is one in intelligent transportation system and important grinds Study carefully field, at the same be also artificial intelligence image recognition, image procossing and pattern identification research heat subject, intelligent transportation with And the tracking etc. of crime vehicle has important application value and research significance.
Traditional vehicle model identification always relies on manual identified and Car license recognition, and all uses bayonet picture, wherein The method of manual identified is extremely inefficient, and recognition accuracy is not also high, and the method for Car license recognition is difficult to the car plate to forgery, set The delinquent vehicle such as board is effectively identified.In addition, the scene residing for wide-angle vehicle is changeable, vehicle body appearance is changeable, than Such as color, shape, size, in addition the distance of each camera shooting is different, and shooting angle difference can all influence final vehicle Type identifier result.
Invention content
The present invention has that recognition accuracy is not high for existing vehicle model recognition methods, provides one kind and is based on The wide-angle model recognizing method and system of global and local Fusion Features.
To solve the above problems, the present invention is achieved by the following technical solutions：
Wide-angle model recognizing method based on global and local Fusion Features, including steps are as follows：
Step 1, when vehicle pass through vehicle cab recognition acquisition zone when, intercept out the image comprising vehicle；
Step 2 pre-processes the image comprising vehicle of interception：
Step 2.1 first cuts image the image of acquisition using callout box, obtains the vehicle of removal complex background Picture；
Step 2.2, vehicle face part, the tail part that vehicle pictures are detected with vehicle respectively using target detection SSD algorithms And wheel portion, obtain vehicle face image piecemeal, tailstock image block and wheel image block；
Step 2.3, to vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image block carry out rotation and Mirror image enhances for data, and enhanced vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image are divided Block is cut into unified size, builds vehicle database；
Constructed vehicle database is imported into the deep layer multiple-limb convolutional neural networks to vehicle by step 3 Global and local feature carry out feature training；
Feature is trained obtained vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image by step 4 The feature of piecemeal is weighted Fusion Features；
Step 5 carries out Classification and Identification by grader to the feature after fusion.
In above-mentioned steps 2.2, target detection SSD algorithms are as follows：
Step 2.2.1, the vehicle pictures of input removal complex background carry out propagated forward and obtain the essential characteristic of vehicle；
Step 2.2.2, the candidate region of different size, different length-width ratios is set in each position of feature；
Step 2.2.3, candidate region and true frame are matched；
Step 2.2.4, the position offset of each candidate region is carried out by prediction by fallout predictor and classification confidence level is defeated Go out；
Step 2.2.5, by the loss function of multitask by the reversed weight relayed calculating and adjust each layer.
In above-mentioned steps 2.2.5, the loss function of multitask is the sum of position loss function and confidence level loss function.
In above-mentioned steps 4, when being weighted Fusion Features to feature,
Vehicle image piecemeal weight coefficient w1 is：
W1=k1/ (k1+k2+k3+k4)；
Vehicle face image piecemeal weight coefficient w2 is：
W2=k2/ (k1+k2+k3+k4)；
Tailstock image block weight coefficient w3 is：
W3=k3/ (k1+k2+k3+k4)；
Wheel image block weight coefficient w4 is：
W4=k4/ (k1+k2+k3+k4)；
Wherein, k1 is the weight of vehicle image piecemeal, and k2 is the weight of vehicle face image piecemeal, and k3 is tailstock image block Weight, k4 are the weights of wheel image block.
Based on a kind of wide-angle model recognition system based on global and local Fusion Features designed by the above method, by Sequentially connected video image acquisition module, vehicle pictures detection module, vehicle pictures segmentation module, vehicle global and local figure As characteristic extracting module, vehicle global and local image co-registration module and wide-angle vehicle model identification module composition；
Video image acquisition module：For obtaining vehicle monitoring video；
Vehicle pictures detection module：It is examined for the vehicle in the road traffic crossroad monitor video to acquisition It surveys, and intercepts out the vehicle in video monitoring；
Vehicle pictures divide module：Vehicle face, the tailstock, wheel portion for detecting vehicle, by the vehicle face of vehicle, the tailstock, Wheel portion is split.
Vehicle global and local image characteristics extraction module；By depth multiple-limb convolutional neural networks respectively to vehicle figure Picture, vehicle face image piecemeal, tailstock image block, wheel image block carry out feature extraction；
Vehicle global and local image co-registration module；Then the feature each branch extracted carries out Fusion Features, to The more rich feature of vehicle can be obtained.
Wide-angle vehicle model identification module：For wide-angle vehicle model to be identified in the feature after fusion, from And identify the model of vehicle in vehicle testing library.
In said program, video image acquisition module is 4 high-definition cameras being arranged on 4 diagonal lines.
Compared with traditional model recognizing method, the present invention carries out local segmentation to vehicle, passes through deep layer multiple-limb convolution Global and local feature is extracted, merged and classified by neural network so that is had greatly improved in terms of recognition accuracy.
Description of the drawings
Fig. 1 is the wide-angle model recognizing method flow chart based on global and local Fusion Features.
Fig. 2 is a kind of wide-angle model recognition system structure chart based on global and local Fusion Features.
Fig. 3 is deep layer multiple-limb convolutional neural networks structure chart.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific example, and with reference to attached Figure, the present invention is described in more detail.
As shown in Figure 1, a kind of wide-angle model recognizing method and system based on global and local Fusion Features, specific to walk It is rapid as follows：
Step S1：When vehicle passes through vehicle cab recognition acquisition zone (road traffic crossroad), the position of positioning vehicle, and And intercept out the image comprising vehicle.
Step S2：The image comprising vehicle of interception is pre-processed.
Step S21：Obtain the information that road traffic crossroad includes vehicle pictures.
Step S22：The image of acquisition cuts image using callout box, obtains the vehicle figure of removal complex background Piece.
Step S23：After obtaining vehicle pictures, vehicle face part, the vehicle of vehicle are detected respectively using target detection SSD algorithms Portion and wheel portion obtain vehicle face image piecemeal, tailstock image block, wheel image block.
The target detection SSD algorithms specifically include：
(1) vehicle pictures of input removal complex background carry out propagated forward and obtain the essential characteristic of vehicle；
(2) candidate region of different size, different length-width ratios is set in each position of feature；
(3) candidate region and true frame are matched；
(4) the position offset prediction by fallout predictor by each candidate region and classification confidence level export；
(5) by the loss function of multitask by the reversed weight relayed calculating and adjust each layer.Wherein multitask Loss function is position loss function (lloc) and confidence level loss function (lconf) sum.
Multitask loss function formula is as follows：
Wherein, N represents the number of matched candidate frame region,Represent weight factor.
Confidence level loss function formula is as follows：
Wherein, c indicates confidence level.
Position loss function formula is as follows：
Wherein, l represents prediction block, and g represents actual value, and d represents candidate region frame, and (cx, cy) represents central point, width w, A height of h.
Step S24：Vehicle pictures, vehicle face image piecemeal, tailstock image block, wheel image block are rotated respectively And mirror image, enhance for data, unified size is cut into enhanced vehicle pictures and each piecemeal picture, builds vehicle number According to library.
Step S3：Deep layer multiple-limb convolutional neural networks are built, by vehicle pictures, vehicle face image piecemeal, tailstock image point Block, wheel image block imported into the global and local feature in the deep layer multiple-limb convolutional neural networks to vehicle and carry out Feature is trained, and vehicle pictures feature and each blocking characteristic are then carried out Fusion Features.
The deep layer multiple-limb convolutional neural networks are as shown in figure 3, it is specifically included：Vehicle pictures branch is sequentially connected Convolutional layer Conv1, pond layer Pool1, convolutional layer Conv2, pond layer Pool2, convolutional layer Conv3, pond layer Pool3, volume Lamination Conv4, pond layer Pool4, the sequentially connected convolutional layer Conv5 of vehicle face image piecemeal branch, pond layer Pool5, convolution Layer Conv6, pond layer Pool6, convolutional layer Conv7, pond layer Pool7, convolutional layer Conv8, pond layer Pool8, tailstock image The sequentially connected convolutional layer Conv9 of piecemeal branch, pond layer Pool9, convolutional layer Conv10, pond layer Pool10, convolutional layer Conv11, pond layer Pool11, convolutional layer Conv12, pond layer Pool12, the sequentially connected convolution of wheel image block branch Layer Conv13, pond layer Pool13, convolutional layer Conv14, pond layer Pool14, convolutional layer Conv15, pond layer Pool15, volume Lamination Conv16, pond layer Pool16 further include that the pond layer Pool4 is connected by full articulamentum Fc1 and articulamentum Concat It connects, the pond layer Pool8 is connect by full articulamentum Fc2 with articulamentum Concat, and the pond layer Pool12 by connecting entirely Layer Fc3 is met to connect with articulamentum Concat, the pond layer Pool16 is connect by full articulamentum Fc4 with articulamentum Concat, Further include that the classification layer Softmax+coco_loss is connect by full articulamentum Fc5 with articulamentum Concat.
The convolutional layer Conv carries out convolution algorithm by convolution kernel to the image of input, then uses neuronal activation letter Number calculates the output valve of convolution.
The feature that the pond layer Pool exports convolutional layer is compressed, and the computation complexity of depth network is simplified, and And extraction main feature.
The full articulamentum Fc is that each node of last layer is connected with all nodes of adjacent layer.
The articulamentum Concat is to merge the feature of above-mentioned each full articulamentum output.
The classification layer Softmax+coco_loss is that the feature for inputting articulamentum is classified.
The Fusion Features specifically include：
Vehicle image piecemeal weight coefficient is：W1=k1/ (k1+k2+k3+k4)；
Vehicle face image piecemeal weight coefficient is：W2=k2/ (k1+k2+k3+k4)；
Tailstock image block weight coefficient is：W3=k2/ (k1+k2+k3+k4)；
Wheel image block weight coefficient is：W4=k2/ (k1+k2+k3+k4)；
Wherein k1 is the weight of vehicle image piecemeal, and k2 is the weight of vehicle face image piecemeal, and k3 is tailstock image block Weight, k4 are that the weight definition vehicle image piecemeal output of wheel image block is x1, and the output of vehicle face image piecemeal is x2, the tailstock Image block output is x3, and the output of wheel image is x4, and the output of final Fusion Features is：Y=w1*x1+w2*x2+w3*x3+ w4*x4。
Step S4：By Softmax loss Classification Loss functions and coco_loss loss functions to the feature after fusion Carry out feature learning.
The probability output of every one kind is calculated using following formula for the Softmax loss functions
Wherein, xiFor Softmax i-th of nodal value of layer, yiFor i-th of output valve, the node number that n is Softmax layers.
The coco_loss loss functions mainly further the feature of similar sample, zoom out the feature of different classifications sample.
(1) input feature vector and central feature normalization
Wherein, ckFor kth classification target eigencenter, f(i)Indicate input feature vector, i=1 ..., N, i.e. batch_size are N。
(2) COS distance of the feature and each eigencenter of input is calculated
COS distance is defined asValue range is [- 1 ,+1], and it is higher to be worth bigger expression similarity.
(3) coco_loss loss functions are calculated
Wherein, B indicates entire batch.Molecule item indicates input feature vector f(i)Cosine between corresponding central feature away from From；Denominator term indicates input feature vector to all central feature sum of the distance.
The multi-section that the present invention can extract wide-angle vehicle by deep layer multiple-limb convolutional neural networks divides abundant feature, Then these abundant features are merged, is had greatly improved in terms of the accuracy rate of identification wide-angle vehicle.
Based on a kind of wide-angle model recognition system based on global and local Fusion Features designed by the above method, such as Shown in Fig. 2, including sequentially connected video image acquisition module, vehicle pictures detection module, vehicle pictures segmentation module, vehicle Global and local image characteristics extraction module, vehicle global and local multi-features module, the identification of wide-angle vehicle model Module.
Video image acquisition module：Vehicle monitoring for obtaining vehicle cab recognition acquisition zone (road traffic crossroad) regards Frequently, it can be four high-definition cameras.
Vehicle pictures detection module：It is examined for the vehicle in the road traffic crossroad monitor video to acquisition It surveys, and intercepts out the vehicle in video monitoring.
Vehicle pictures divide module：Vehicle face, the tailstock, wheel portion for detecting vehicle, by the vehicle face of vehicle, the tailstock, Wheel portion is split.
Vehicle global and local image characteristics extraction module：By depth multiple-limb convolutional neural networks respectively to vehicle figure Picture, vehicle face image piecemeal, tailstock image block, wheel image block carry out feature extraction；
Vehicle global and local multi-features module：Then the feature each branch extracted carries out Fusion Features, So as to obtain the more rich feature of vehicle；
Wide-angle vehicle model identification module：For wide-angle vehicle model to be identified in the feature after fusion, from And identify the model of vehicle in vehicle testing library.
The position of positioning vehicle when vehicle passes through traffic intersection, and the image comprising vehicle is intercepted out, to interception Including the image of vehicle is pre-processed, the image comprising vehicle of interception is cut using callout box first, is gone Except the vehicle pictures of complex background, vehicle pictures are divided into vehicle face image piecemeal, tailstock image block, wheel image block. Deep layer multiple-limb convolutional neural networks are built, by vehicle pictures, vehicle face image piecemeal, tailstock image block, wheel image block It imported into the global and local feature in the deep layer multiple-limb convolutional neural networks to vehicle and carries out feature training, then will Vehicle pictures feature and each blocking characteristic carry out Fusion Features, then carry out classification knowledge to the feature after fusion by grader Not.The method have the advantage is capable of the global and local feature of wide-angle vehicle is merged, with existing tradition Model recognizing method is compared, and the accuracy rate of wide-angle vehicle cab recognition can be significantly improved.
It should be noted that although the above embodiment of the present invention is illustrative, this is not to the present invention Limitation, therefore the invention is not limited in above-mentioned specific implementation mode.Without departing from the principles of the present invention, every The other embodiment that those skilled in the art obtain under the inspiration of the present invention is accordingly to be regarded as within the protection of the present invention.

Claims (6)

1. the wide-angle model recognizing method based on global and local Fusion Features, characterized in that including steps are as follows：
Step 1, when vehicle pass through vehicle cab recognition acquisition zone when, intercept out the image comprising vehicle；
Step 2 pre-processes the image comprising vehicle of interception：
Step 2.1 first cuts image the image of acquisition using callout box, obtains the vehicle figure of removal complex background Piece；
Step 2.2, vehicle face part, tail part and the vehicle that vehicle pictures are detected with vehicle respectively using target detection SSD algorithms Part is taken turns, vehicle face image piecemeal, tailstock image block and wheel image block are obtained；
Step 2.3 carries out rotation and mirror image to vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image block, And enhanced vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image block are cut into unified size, structure Build vehicle database；
Constructed vehicle database is imported into the deep layer multiple-limb convolutional neural networks to the complete of vehicle by step 3 Office and local feature carry out feature training；
Feature is trained obtained vehicle pictures, vehicle face image piecemeal, tailstock image block and wheel image block by step 4 Feature be weighted Fusion Features；
Step 5 carries out Classification and Identification by grader to the feature after fusion.
2. the wide-angle model recognizing method according to claim 1 based on global and local Fusion Features, characterized in that In step 2.2, target detection SSD algorithms are as follows：
Step 2.2.1, the vehicle pictures of input removal complex background carry out propagated forward and obtain the essential characteristic of vehicle；
Step 2.2.2, the candidate region of different size, different length-width ratios is set in each position of feature；
Step 2.2.3, candidate region and true frame are matched；
Step 2.2.4, the position offset of each candidate region is carried out by prediction by fallout predictor and classification confidence level exports；
Step 2.2.5, by the loss function of multitask by the reversed weight relayed calculating and adjust each layer.
3. the wide-angle model recognizing method according to claim 2 based on global and local Fusion Features, characterized in that In step 2.2.5, the loss function of multitask is the sum of position loss function and confidence level loss function.
4. the wide-angle model recognizing method according to claim 1 based on global and local Fusion Features, characterized in that In step 4, when being weighted Fusion Features to feature,
Vehicle image piecemeal weight coefficient w1 is：
W1=k1/ (k1+k2+k3+k4)；
Vehicle face image piecemeal weight coefficient w2 is：
W2=k2/ (k1+k2+k3+k4)；
Tailstock image block weight coefficient w3 is：
W3=k3/ (k1+k2+k3+k4)；
Wheel image block weight coefficient w4 is：
W4=k4/ (k1+k2+k3+k4)；
Wherein, k1 is the weight of vehicle image piecemeal, and k2 is the weight of vehicle face image piecemeal, and k3 is the power of tailstock image block Weight, k4 is the weight of wheel image block.
5. the wide-angle model recognition system based on global and local Fusion Features, characterized in that by sequentially connected video figure As acquisition module, vehicle pictures detection module, vehicle pictures segmentation module, vehicle global and local image characteristics extraction module, Vehicle global and local multi-features module and wide-angle vehicle model identification module composition；
Video image acquisition module：For obtaining vehicle monitoring video；
Vehicle pictures detection module：It is detected for the vehicle in the road traffic crossroad monitor video to acquisition, and And intercept out the vehicle in video monitoring；
Vehicle pictures divide module：Vehicle face, the tailstock, wheel portion for detecting vehicle, by vehicle face, the tailstock, the wheel of vehicle Part is split；
Vehicle global and local image characteristics extraction module：By depth multiple-limb convolutional neural networks respectively to vehicle image, Vehicle face image piecemeal, tailstock image block, wheel image block carry out feature extraction；
Vehicle global and local multi-features module：Then the feature each branch extracted carries out Fusion Features, to The more rich feature of vehicle can be obtained；
Wide-angle vehicle model identification module：For wide-angle vehicle model to be identified in the feature after fusion, to know Do not go out the model of vehicle in vehicle testing library.
6. the wide-angle model recognition system according to claim 5 based on global and local Fusion Features, characterized in that Video image acquisition module is 4 high-definition cameras being arranged on 4 diagonal lines.
CN201810463134.5A 2018-05-15 2018-05-15 Wide-angle model recognizing method and system based on global and local Fusion Features Pending CN108681707A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810463134.5A CN108681707A (en) 2018-05-15 2018-05-15 Wide-angle model recognizing method and system based on global and local Fusion Features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810463134.5A CN108681707A (en) 2018-05-15 2018-05-15 Wide-angle model recognizing method and system based on global and local Fusion Features

Publications (1)

Publication Number Publication Date
CN108681707A true CN108681707A (en) 2018-10-19

Family

ID=63805307

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810463134.5A Pending CN108681707A (en) 2018-05-15 2018-05-15 Wide-angle model recognizing method and system based on global and local Fusion Features

Country Status (1)

Country Link
CN (1) CN108681707A (en)

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109359696A (en) * 2018-10-29 2019-02-19 重庆中科云丛科技有限公司 A kind of vehicle money recognition methods, system and storage medium
CN109886312A (en) * 2019-01-28 2019-06-14 同济大学 A kind of bridge wheel of vehicle detection method based on multilayer feature fused neural network model
CN109902563A (en) * 2019-01-17 2019-06-18 桂林远望智能通信科技有限公司 A kind of multi-angle model recognizing method and system
CN109919072A (en) * 2019-02-28 2019-06-21 桂林电子科技大学 Fine vehicle type recognition and flow statistics method based on deep learning and trajectory tracking
CN109948610A (en) * 2019-03-14 2019-06-28 西南交通大学 A kind of vehicle fine grit classification method in the video based on deep learning
CN110084230A (en) * 2019-04-11 2019-08-02 北京百度网讯科技有限公司 Vehicle body direction detection method and device based on image
CN110211039A (en) * 2019-04-29 2019-09-06 西安电子科技大学 A kind of image processing method and its device
CN110348463A (en) * 2019-07-16 2019-10-18 北京百度网讯科技有限公司 The method and apparatus of vehicle for identification
CN110472503A (en) * 2019-07-11 2019-11-19 桂林电子科技大学 A kind of road target detection method, device and storage medium
CN110689481A (en) * 2019-01-17 2020-01-14 成都通甲优博科技有限责任公司 Vehicle type identification method and device
CN110765954A (en) * 2019-10-24 2020-02-07 浙江大华技术股份有限公司 Vehicle weight recognition method, equipment and storage device
CN110991255A (en) * 2019-11-11 2020-04-10 智慧互通科技有限公司 Method for detecting fake-licensed vehicle based on deep learning algorithm
CN111340026A (en) * 2020-03-05 2020-06-26 苏州科达科技股份有限公司 Training method of vehicle annual payment identification model and vehicle annual payment identification method
CN111382787A (en) * 2020-03-06 2020-07-07 芯薇(上海)智能科技有限公司 Target detection method based on deep learning
CN109447095B (en) * 2018-10-30 2020-09-29 上海七牛信息技术有限公司 Visual attribute identification method, device 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
CN112233071A (en) * 2020-09-28 2021-01-15 国网浙江省电力有限公司杭州供电公司 Multi-granularity hidden danger detection method and system based on power transmission network picture in complex environment
CN112598646A (en) * 2020-12-23 2021-04-02 山东产研鲲云人工智能研究院有限公司 Capacitance defect detection method and device, electronic equipment and storage medium
CN113177133A (en) * 2021-04-23 2021-07-27 深圳依时货拉拉科技有限公司 Image retrieval method, device, equipment and storage medium
CN113221604A (en) * 2020-02-05 2021-08-06 北京京东乾石科技有限公司 Target identification method and device, storage medium and electronic equipment
CN113657462A (en) * 2021-07-28 2021-11-16 讯飞智元信息科技有限公司 Method for training vehicle recognition model, vehicle recognition method and computing device
WO2022027873A1 (en) * 2020-08-05 2022-02-10 智慧互通科技有限公司 Vehicle reidentification method and device based on multimodal information fusion
CN115100509A (en) * 2022-07-15 2022-09-23 山东建筑大学 Image identification method and system based on multi-branch block-level attention enhancement network
CN115761659A (en) * 2023-01-09 2023-03-07 南京隼眼电子科技有限公司 Recognition model construction method, vehicle type recognition method, electronic device, and storage medium

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104680795A (en) * 2015-02-28 2015-06-03 武汉烽火众智数字技术有限责任公司 Vehicle type recognition method and device based on partial area characteristic
US20150254531A1 (en) * 2014-03-07 2015-09-10 Tata Consultancy Services Limited Multi range object detection device and method
CN105488517A (en) * 2015-11-30 2016-04-13 杭州全实鹰科技有限公司 Vehicle brand model identification method based on deep learning
CN106127107A (en) * 2016-06-14 2016-11-16 宁波熵联信息技术有限公司 The model recognizing method that multi-channel video information based on license board information and vehicle's contour merges
CN106529446A (en) * 2016-10-27 2017-03-22 桂林电子科技大学 Vehicle type identification method and system based on multi-block deep convolutional neural network
CN106875693A (en) * 2017-03-29 2017-06-20 广西信路威科技发展有限公司 A kind of method and system of vehicle feature recognition
CN107330463A (en) * 2017-06-29 2017-11-07 南京信息工程大学 Model recognizing method based on CNN multiple features combinings and many nuclear sparse expressions
CN107729818A (en) * 2017-09-21 2018-02-23 北京航空航天大学 A kind of multiple features fusion vehicle recognition methods again based on deep learning

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150254531A1 (en) * 2014-03-07 2015-09-10 Tata Consultancy Services Limited Multi range object detection device and method
CN104680795A (en) * 2015-02-28 2015-06-03 武汉烽火众智数字技术有限责任公司 Vehicle type recognition method and device based on partial area characteristic
CN105488517A (en) * 2015-11-30 2016-04-13 杭州全实鹰科技有限公司 Vehicle brand model identification method based on deep learning
CN106127107A (en) * 2016-06-14 2016-11-16 宁波熵联信息技术有限公司 The model recognizing method that multi-channel video information based on license board information and vehicle's contour merges
CN106529446A (en) * 2016-10-27 2017-03-22 桂林电子科技大学 Vehicle type identification method and system based on multi-block deep convolutional neural network
CN106875693A (en) * 2017-03-29 2017-06-20 广西信路威科技发展有限公司 A kind of method and system of vehicle feature recognition
CN107330463A (en) * 2017-06-29 2017-11-07 南京信息工程大学 Model recognizing method based on CNN multiple features combinings and many nuclear sparse expressions
CN107729818A (en) * 2017-09-21 2018-02-23 北京航空航天大学 A kind of multiple features fusion vehicle recognition methods again based on deep learning

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHEN CHAOCUN, ET AL: "《Vehicle type recognition based on multi-branch and multi-layer features》", 《2017 IEEE 2ND ADVANCED INFORMATION TECHNOLOGY, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (IAEAC). IEEE》 *
甘凯今 等: "《融合整体与局部特征的车辆型号识别方法》", 《现代电子技术》 *
蔡晓东: "《基于多分支卷积神经网络的车辆图像比对方法》", 《电视技术》 *

Cited By (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109359696A (en) * 2018-10-29 2019-02-19 重庆中科云丛科技有限公司 A kind of vehicle money recognition methods, system and storage medium
CN109359696B (en) * 2018-10-29 2021-04-02 重庆中科云从科技有限公司 Vehicle money identification method, system and storage medium
CN109447095B (en) * 2018-10-30 2020-09-29 上海七牛信息技术有限公司 Visual attribute identification method, device and storage medium
CN110689481A (en) * 2019-01-17 2020-01-14 成都通甲优博科技有限责任公司 Vehicle type identification method and device
CN109902563A (en) * 2019-01-17 2019-06-18 桂林远望智能通信科技有限公司 A kind of multi-angle model recognizing method and system
CN109886312A (en) * 2019-01-28 2019-06-14 同济大学 A kind of bridge wheel of vehicle detection method based on multilayer feature fused neural network model
CN109886312B (en) * 2019-01-28 2023-06-06 同济大学 Bridge vehicle wheel detection method based on multilayer feature fusion neural network model
CN109919072A (en) * 2019-02-28 2019-06-21 桂林电子科技大学 Fine vehicle type recognition and flow statistics method based on deep learning and trajectory tracking
CN109948610A (en) * 2019-03-14 2019-06-28 西南交通大学 A kind of vehicle fine grit classification method in the video based on deep learning
CN110084230B (en) * 2019-04-11 2021-05-28 北京百度网讯科技有限公司 Image-based vehicle body direction detection method and device
CN110084230A (en) * 2019-04-11 2019-08-02 北京百度网讯科技有限公司 Vehicle body direction detection method and device based on image
CN110211039A (en) * 2019-04-29 2019-09-06 西安电子科技大学 A kind of image processing method and its device
CN110472503A (en) * 2019-07-11 2019-11-19 桂林电子科技大学 A kind of road target detection method, device and storage medium
CN110348463A (en) * 2019-07-16 2019-10-18 北京百度网讯科技有限公司 The method and apparatus of vehicle for identification
CN110765954A (en) * 2019-10-24 2020-02-07 浙江大华技术股份有限公司 Vehicle weight recognition method, equipment and storage device
CN110991255A (en) * 2019-11-11 2020-04-10 智慧互通科技有限公司 Method for detecting fake-licensed vehicle based on deep learning algorithm
CN110991255B (en) * 2019-11-11 2023-09-08 智慧互通科技股份有限公司 Method for detecting fake-licensed car based on deep learning algorithm
CN113221604A (en) * 2020-02-05 2021-08-06 北京京东乾石科技有限公司 Target identification method and device, storage medium and electronic equipment
CN111340026A (en) * 2020-03-05 2020-06-26 苏州科达科技股份有限公司 Training method of vehicle annual payment identification model and vehicle annual payment identification method
CN111382787A (en) * 2020-03-06 2020-07-07 芯薇(上海)智能科技有限公司 Target detection method based on deep learning
WO2022027873A1 (en) * 2020-08-05 2022-02-10 智慧互通科技有限公司 Vehicle reidentification method and device based on multimodal information fusion
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
CN112233071A (en) * 2020-09-28 2021-01-15 国网浙江省电力有限公司杭州供电公司 Multi-granularity hidden danger detection method and system based on power transmission network picture in complex environment
CN112598646A (en) * 2020-12-23 2021-04-02 山东产研鲲云人工智能研究院有限公司 Capacitance defect detection method and device, electronic equipment and storage medium
CN112598646B (en) * 2020-12-23 2024-06-11 山东产研鲲云人工智能研究院有限公司 Capacitance defect detection method and device, electronic equipment and storage medium
CN113177133A (en) * 2021-04-23 2021-07-27 深圳依时货拉拉科技有限公司 Image retrieval method, device, equipment and storage medium
CN113177133B (en) * 2021-04-23 2024-03-29 深圳依时货拉拉科技有限公司 Image retrieval method, device, equipment and storage medium
CN113657462A (en) * 2021-07-28 2021-11-16 讯飞智元信息科技有限公司 Method for training vehicle recognition model, vehicle recognition method and computing device
CN115100509A (en) * 2022-07-15 2022-09-23 山东建筑大学 Image identification method and system based on multi-branch block-level attention enhancement network
CN115100509B (en) * 2022-07-15 2022-11-29 山东建筑大学 Image identification method and system based on multi-branch block-level attention enhancement network
CN115761659A (en) * 2023-01-09 2023-03-07 南京隼眼电子科技有限公司 Recognition model construction method, vehicle type recognition method, electronic device, and storage medium

Similar Documents

Publication Publication Date Title
CN108681707A (en) Wide-angle model recognizing method and system based on global and local Fusion Features
Li et al. TransUNetCD: A hybrid transformer network for change detection in optical remote-sensing images
Zou et al. Robust lane detection from continuous driving scenes using deep neural networks
CN112818903B (en) Small sample remote sensing image target detection method based on meta-learning and cooperative attention
Liu et al. Multiscale U-shaped CNN building instance extraction framework with edge constraint for high-spatial-resolution remote sensing imagery
CN103679674B (en) Method and system for splicing images of unmanned aircrafts in real time
CN107578091B (en) Pedestrian and vehicle real-time detection method based on lightweight deep network
CN117422971A (en) Bimodal target detection method and system based on cross-modal attention mechanism fusion
CN115861619A (en) Airborne LiDAR (light detection and ranging) urban point cloud semantic segmentation method and system of recursive residual double-attention kernel point convolution network
Zang et al. Traffic lane detection using fully convolutional neural network
CN115019043A (en) Image point cloud fusion three-dimensional target detection method based on cross attention mechanism
CN117157679A (en) Perception network, training method of perception network, object recognition method and device
Wu et al. Vehicle detection based on adaptive multi-modal feature fusion and cross-modal vehicle index using RGB-T images
CN118015490A (en) Unmanned aerial vehicle aerial image small target detection method, system and electronic equipment
CN116343095A (en) Vehicle track extraction method based on video stitching and related equipment
CN116597411A (en) Method and system for identifying traffic sign by unmanned vehicle in extreme weather
CN116740516A (en) Target detection method and system based on multi-scale fusion feature extraction
CN110826449A (en) Non-motor vehicle re-identification target retrieval method based on light convolutional neural network
Algamdi et al. Instant Counting & Vehicle Detection during Hajj Using Drones
Jiang et al. Knowledge distillation from 3D to bird’s-eye-view for LiDAR semantic segmentation
CN115205568B (en) Road traffic multi-element detection method based on multi-scale feature fusion
CN116994164A (en) Multi-mode aerial image fusion and target detection combined learning method
CN116935175A (en) Parking space detection method based on multi-scale detection head feature fusion
CN116485894A (en) Video scene mapping and positioning method and device, electronic equipment and storage medium
CN112733934B (en) Multi-mode feature fusion road scene semantic segmentation method in complex environment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20181019