CN105469062A - Principal component analysis network framework based vehicle type identification method - Google Patents
Principal component analysis network framework based vehicle type identification method Download PDFInfo
- Publication number
- CN105469062A CN105469062A CN201510884218.2A CN201510884218A CN105469062A CN 105469062 A CN105469062 A CN 105469062A CN 201510884218 A CN201510884218 A CN 201510884218A CN 105469062 A CN105469062 A CN 105469062A
- Authority
- CN
- China
- Prior art keywords
- image
- pca
- principal component
- component analysis
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a principal component analysis (PCA) network framework based vehicle type identification method. The method comprises: firstly performing sliding window feature selection and averaging operations on vehicle type images to obtain preliminary features of the images; forming a matrix X by the preliminary features of all the images, performing PCA on X, and re-arranging each column of first L1 eigenvectors to obtain L1 new windows; performing convolution on the vehicle type images by using the L1 windows to obtain L1 convolution images; performing the operations on each convolution image of current images again to obtain a set of L1 groups of convolution images; performing binarization and weight addition on each group of the convolution images in the set to obtain L1 images; performing block histogram operation on the L1 images to obtain PCA network framework feature representation of current images; performing SVM training on the obtained feature representation of all the images to obtain a vehicle type identification system; and extracting PCA network framework features of test images, importing the system to perform testing so as to identify vehicle types.
Description
Technical field
The invention belongs to image procossing, pattern classification and data mining technology field, particularly a kind of model recognizing method based on principal component analysis (PCA) screen frame frame.
Background of invention
The Images Classification of view-based access control model semanteme is all a full of challenges research field all the time, image not only to be identified is of a great variety, also there is all multivariates in the inside of each class image, comprise illumination variation, do not mate and do not line up, deformation factor, block factor etc.For these parameters, scholars have done various effort, propose various feature and deal with these changes.More typical example has the LBP feature used in text classification, vehicle classification, SIFT and the HOG feature used in target identification.Although the low-level image feature of these pickings can be good at the data processing task under reply particular case, the generalization ability of these features is limited, and treating new problem often needs to build new feature.
Obtain interested feature from data learning and be considered to overcome the good method of of picking characteristic limitations, a most typical example carries out feature learning by degree of depth neural network exactly.The basic thought of degree of depth study is by building multitiered network, carrying out multilayer expression, to being represented the abstract semantics information of data by the high-level feature of multilayer, obtain better feature robustness to target.In the study of the current degree of depth one crucial, relatively successfully framework be convolutional network framework.A degree of depth convolutional network structure comprises multilayer training structure and has the sorter of supervision, and every one deck all comprises three sublayers, is respectively convolutional layer, Nonlinear Processing layer and down-sampling layer.This convolutional network structure is generally all trained it by gradient descent method.But want the volume that learns to obtain a high-quality and network class structure, need various tune ginseng experiences and skills.
At present, for different visual identity tasks, there has been proposed many various convolutional neural networks structures, and achieve remarkable result.Such as comparatively successful small echo scattering model, it is by changing by convolution kernel as the step that Wavelet Kernel avoids Algorithm Learning into.But be exactly that this is simply changed, convolutional network and the deep neural network of identical level can be exceeded in Handwritten Digit Recognition and text identification etc., but it is performed poor in vehicle cab recognition, be difficult to reply illumination variation and block impact.
According to the above, the present invention proposes a kind of model recognizing method based on principal component analysis (PCA) screen frame frame, propose a very succinct degree of depth learning framework about vehicle cab recognition, this framework mainly relies on several basic data processing method: (1) principal component analysis (PCA) PCA; (2) binaryzation Hash coding; (3) blocked histogram.In this framework, first Multilayer filter core is learnt by PCA method, then binaryzation Hash coding and block histogram feature is used to carry out down-sampling and encoding operation, can solve the problem well, algorithm rapidly and efficiently, improve recognition speed, consume internal memory few, there is higher practicality and robustness.
Summary of the invention
Method described in the present invention is the shortcoming in order to overcome above-mentioned prior art, carries out feature extraction and carries out vehicle segmenting the problem identified, propose a kind of model recognizing method based on principal component analysis (PCA) screen frame frame mainly for vehicle image.Concrete technical scheme is as described below.
Based on a model recognizing method for principal component analysis (PCA) screen frame frame, comprise the following steps:
Step 1: to the vehicle image of training vehicle image library, carry out moving window selected characteristic and remove average operation, obtaining the preliminary feature of image;
Step 2: the preliminary characteristic block of all images forms matrix X, is PCA, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window to X;
Step 3: to each vehicle image, does convolution with L1 new window, all obtains L1 convolved image;
Step 4: the operation each convolved image of present image being performed again to step 1-step 3, obtains the set of L1 group convolved image, in L1, often group has L2 convolved image;
Step 5: carrying out binaryzation and weights addition process to often organizing convolved image in set, obtaining new L1 and opening image;
Step 6: new L1 is opened image and carries out blocked histogram operation, obtains the principal component analysis (PCA) screen frame frame character representation of present image after associating;
Step 7: the principal component analysis (PCA) screen frame frame character representation of all images obtained is carried out SVM training, obtains model recognition system;
Step 8: to test vehicle image, extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle.
In technique scheme, described step 1 pair vehicle image, carries out moving window selected characteristic and removes average operation, obtaining the preliminary feature of image, comprise following step:
Step 1.1: choosing size is k
1× k
2window, sliding and extracting size is the local feature of the training image of m × n, obtains a k
1× k
2oK, the matrix of m × n row, each row represents a local feature block;
Step 1.2: carry out going on average, to obtain the character representation of Current vehicle image by row to the matrix obtained in step 1.1.
In technique scheme, the preliminary characteristic block of all images of described step 2 forms matrix X, is PCA, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window, comprise following step X:
Step 2.1: open training picture to all N and carry out the operation of step 1, by side-by-side for the feature obtained, obtains a new data matrix X;
Step 2.2: do PCA dimension-reduction treatment to the matrix X obtained, gets a front L1 proper vector, as wave filter;
Step 2.3: each rearrangement of this L1 proper vector is classified as a new characteristic block, obtains L1 k
1× k
2window.
In technique scheme, described step 5 carries out binaryzation and weights addition process to often organizing convolved image in set, obtains new L1 and opens image, comprise following step:
Step 5.1: do binary conversion treatment to each convolved image obtained in step 4, in former convolved image, element is greater than 0, then the position that new image element matrix is corresponding is 1, otherwise is 0;
Step 5.2: to each vehicle image, the binary image having L1 group often to organize L2 to open, opens bianry image by the L2 often organized and carries out weights addition.Weights size is corresponding in turn to the size of step 2.2 median filter;
Step 5.3: the new L1 often being opened vehicle image opens image.
In technique scheme, new L1 is opened image and carries out blocked histogram operation by described step 6, obtains the principal component analysis (PCA) screen frame frame character representation of present image, comprise following step after associating:
Step 6.1: to the L1 width image obtained in step 5, each width figure is divided into B block, carries out statistics with histogram to each piece;
Step 6.2: coupled together by the histogram of the block of the B in step 6.1 and become a vector, then all coupled together by vector corresponding for L1 width image, obtains the principal component analysis (PCA) screen frame frame character representation of present image.
In technique scheme, the character representation of all images obtained is carried out SVM training by described step 7, obtains model recognition system, comprises following step:
Step 7.1: using the principal component analysis (PCA) screen frame frame character representation of current class vehicle image as positive sample, the principal component analysis (PCA) screen frame frame character representation of other classification vehicle images is as negative sample;
Step 7.2: use Linear SVM to train this two samples, obtain the sorter of current class vehicle image;
Step 7.3: repeat the operation of step 7.1 to step 7.2, obtain the sorter of all categories vehicle image, form model recognition system after associating.
In technique scheme, described step 8 is to test vehicle image, and extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle, comprises following step:
Step 8.1: to test vehicle image, by the operation of step 1 to step 6, obtain its principal component analysis (PCA) screen frame frame character representation;
Step 8.2: by the principal component analysis (PCA) screen frame frame character representation of test vehicle image obtained, the model recognition system in steps for importing 7 is tested, and according to maximum probability matching, identifies vehicle.
Because the present invention adopts technique scheme, therefore possess following beneficial effect:
Present invention employs principal component analysis (PCA) screen frame frame to build the character representation of vehicle image, propose a very succinct degree of depth learning framework about vehicle cab recognition, this framework mainly relies on several basic data processing method: (1) principal component analysis (PCA) PCA; (2) binaryzation Hash coding; (3) blocked histogram.In this framework, first Multilayer filter core is learnt by PCA method, then binaryzation Hash coding and block histogram feature is used to carry out down-sampling and encoding operation, vehicle cab recognition can be realized well, algorithm rapidly and efficiently, improve recognition speed, consume internal memory few, there is higher practicality and robustness.
Accompanying drawing explanation
Fig. 1 is the algorithm realization schematic diagram of the principal component analysis (PCA) net frame representation of image.
Embodiment
By describing technology contents of the present invention, structural attitude in detail, realized object and effect, accompanying drawing is coordinated to be explained in detail below in conjunction with embodiment.
The present invention proposes a kind of model recognizing method based on principal component analysis (PCA) screen frame frame, vehicle vehicle cab recognition obtains good effect.Whole algorithm realization schematic diagram as shown in Figure 1, comprises step:
Step 1: to vehicle image, carries out moving window selected characteristic and removes average operation, obtaining the preliminary feature of image;
Suppose have N to open training image sample, each sample is of a size of m × n, and the filter size arranging every layer is k
1× k
2, for each pixel, all around it, carry out a k
1× k
2block sampling, collect all sampling blocks, carry out cascade, obtain the local feature block of image, particularly, comprise following step:
Step 1.1: choosing size is k
1× k
2window, sliding and extracting size is the local feature of the training image of m × n, obtains a k
1× k
2oK, the matrix of m × n row, each row represents a local feature block;
Step 1.2: carry out going on average, to obtain the character representation of Current vehicle image by row to the matrix obtained in step 1.1.
Step 2: the preliminary characteristic block of all images forms matrix X, is PCA, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window to X;
Particularly, the preliminary characteristic block of all images forms matrix X, is PCA, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window, comprise following step X:
Step 2.1: open training picture to all N and carry out the operation of step 1, by side-by-side for the feature obtained, obtains a new data matrix X;
Step 2.2: do PCA dimension-reduction treatment to the matrix X obtained, gets a front L1 proper vector, as wave filter;
Step 2.3: each rearrangement of this L1 proper vector is classified as a new characteristic block, obtains L1 k
1× k
2window.
Step 3: to each vehicle image, does convolution with this L1 window, all obtains L1 convolved image;
Particularly, to each vehicle image, do convolution with this L1 window, all obtain L1 convolved image, comprise following step:
Step 3.1: to each vehicle image, does convolution operation with each window of obtaining in step 2 respectively;
Step 3.2: L1 the convolved image obtaining each vehicle image.
Step 4: above operation is performed again to each convolved image of present image, obtains the set of L1 group convolved image;
Particularly, above operation is performed again to each convolved image of present image, obtains the set of L1 group convolved image, comprise following step:
Step 4.1: to L1 convolved image of each vehicle image obtained in step 3, performs the operation of step 1 to step 3 again;
Step 4.2: the L1 group obtaining current vehicle image, often organizes the set of L2 convolved image.
Step 5: carrying out binaryzation and weights addition process to often organizing convolved image in set, obtaining new L1 and opening image;
Particularly, carrying out binaryzation and weights addition process to often organizing convolved image in set, obtaining new L1 and opening image, comprising following step:
Step 5.1: do binary conversion treatment to each convolved image obtained in step 4, in former convolved image, element is greater than 0, then the position that new image element matrix is corresponding is 1, otherwise is 0;
Step 5.2: to each vehicle image, the binary image having L1 group often to organize L2 to open, opens bianry image by the L2 often organized and carries out weights addition.Weights size is corresponding in turn to the size of step 2.2 median filter;
Step 5.3: the new L1 often being opened vehicle image opens image.
Step 6: new L1 is opened image and carries out blocked histogram operation, obtains the principal component analysis (PCA) screen frame frame character representation of present image after associating;
Particularly, new L1 is opened image and carries out blocked histogram operation, obtain the principal component analysis (PCA) screen frame frame character representation of present image after associating, comprise following step:
Step 6.1: to the L1 width image obtained in step 5, each width figure is divided into B block, carries out statistics with histogram to each piece;
Step 6.2: coupled together by the histogram of the block of the B in step 6.1 and become a vector, then all coupled together by vector corresponding for L1 width image, obtains the principal component analysis (PCA) screen frame frame character representation of present image.
Step 7: the character representation of all images obtained is carried out SVM training, obtains model recognition system;
Particularly, the character representation of all images obtained is carried out SVM training, obtains model recognition system, comprise following step:
Step 7.1: using the principal component analysis (PCA) screen frame frame character representation of current class vehicle image as positive sample, the principal component analysis (PCA) screen frame frame character representation of other classification vehicle images is as negative sample;
Step 7.2: use Linear SVM to train this two samples, obtain the sorter of current class vehicle image;
Step 7.3: repeat the operation of step 7.1 to step 7.2, obtain the sorter of all categories vehicle image, form model recognition system after associating.
Step 8: to test vehicle image, extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle.
Particularly, to test vehicle image, extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle, comprises following step:
Step 8.1: to test vehicle image, by the operation of step 1 to step 6, obtain its principal component analysis (PCA) screen frame frame character representation;
Step 8.2: by the principal component analysis (PCA) screen frame frame character representation of test vehicle image obtained, the model recognition system in steps for importing 7 is tested, and according to maximum probability matching, identifies vehicle.
Claims (7)
1., based on a model recognizing method for principal component analysis (PCA) screen frame frame, comprise the following steps:
Step 1: to the vehicle image of training vehicle image library, carry out moving window selected characteristic and remove average operation, obtaining the preliminary feature of image;
Step 2: the preliminary characteristic block of all images forms matrix X, is PCA, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window to X;
Step 3: to each vehicle image, does convolution with L1 new window, all obtains L1 convolved image;
Step 4: the operation each convolved image of present image being performed again to step 1-step 3, obtains the set of L1 group convolved image, in L1, often group has L2 convolved image;
Step 5: carrying out binaryzation and weights addition process to often organizing convolved image in the set of L1 group convolved image, obtaining new L1 and opening image;
Step 6: new L1 is opened image and carries out blocked histogram operation, obtains the principal component analysis (PCA) screen frame frame character representation of present image after associating;
Step 7: the principal component analysis (PCA) screen frame frame character representation of all images obtained is carried out SVM training, obtains model recognition system;
Step 8: to test vehicle image, extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle.
2. according to claim 1 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, to vehicle image in described step 1, carry out moving window selected characteristic and remove average operation, obtaining the preliminary feature of image, comprise following step:
Step 1.1: choosing size is k
1× k
2window, sliding and extracting size is the local feature of the training image of m × n, obtains a k
1× k
2oK, the matrix of m × n row, each row represents a local feature block;
Step 1.2: carry out going on average, to obtain the character representation of Current vehicle image by row to the matrix obtained in step 1.1.
3. according to claim 1 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, in described step 2, the preliminary characteristic block of all images forms matrix X, PCA is done to X, gets a front L1 proper vector, by its every rearrangement, obtain L1 new window, comprise following step:
Step 2.1: open training picture to all N and carry out the operation of step 1, by side-by-side for the feature obtained, obtains a new data matrix X;
Step 2.2: do PCA dimension-reduction treatment to the matrix X obtained, gets a front L1 proper vector, as wave filter;
Step 2.3: each rearrangement of this L1 proper vector is classified as a new characteristic block, obtains L1 k
1× k
2window.
4. according to claim 3 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, carrying out binaryzation and weights addition process to often organizing convolved image in set in described step 5, obtaining new L1 and opening image, comprising following step:
Step 5.1: do binary conversion treatment to each convolved image obtained in step 4, in former convolved image, element is greater than 0, then the position that new image element matrix is corresponding is 1, otherwise is 0;
Step 5.2: to each vehicle image, the binary image having L1 group often to organize L2 to open, the L2 often organized is opened bianry image and carries out weights addition, weights size is corresponding in turn to the size of step 2.2 median filter;
Step 5.3: the new L1 often being opened vehicle image opens image.
5. according to claim 1 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, in described step 6, new L1 is opened image and carries out blocked histogram operation, obtain the principal component analysis (PCA) screen frame frame character representation of present image after associating, comprise following step:
Step 6.1: to the L1 width image obtained in step 5, each width figure is divided into B block, carries out statistics with histogram to each piece;
Step 6.2: coupled together by the histogram of the block of the B in step 6.1 and become a vector, then all coupled together by vector corresponding for L1 width image, obtains the principal component analysis (PCA) screen frame frame character representation of present image.
6. according to claim 1 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, in described step 7, the character representation of all images obtained carried out SVM training, obtain model recognition system, comprise following step:
Step 7.1: using the principal component analysis (PCA) screen frame frame character representation of current class vehicle image as positive sample, the principal component analysis (PCA) screen frame frame character representation of other classification vehicle images is as negative sample;
Step 7.2: use Linear SVM to train this two samples, obtain the sorter of current class vehicle image;
Step 7.3: repeat the operation of step 7.1 to step 7.2, obtain the sorter of all categories vehicle image, form model recognition system after associating.
7. according to claim 1 based on the model recognizing method of principal component analysis (PCA) screen frame frame, it is characterized in that, to test vehicle image in described step 8, extract its principal component analysis (PCA) net frame feature, import system is tested, and identifies vehicle, comprises following step:
Step 8.1: to test vehicle image, by the operation of step 1 to step 6, obtain its principal component analysis (PCA) screen frame frame character representation;
Step 8.2: by the principal component analysis (PCA) screen frame frame character representation of test vehicle image obtained, the model recognition system in steps for importing 7 is tested, and according to maximum probability matching, identifies vehicle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510884218.2A CN105469062A (en) | 2015-12-04 | 2015-12-04 | Principal component analysis network framework based vehicle type identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510884218.2A CN105469062A (en) | 2015-12-04 | 2015-12-04 | Principal component analysis network framework based vehicle type identification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105469062A true CN105469062A (en) | 2016-04-06 |
Family
ID=55606733
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510884218.2A Pending CN105469062A (en) | 2015-12-04 | 2015-12-04 | Principal component analysis network framework based vehicle type identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105469062A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106384100A (en) * | 2016-09-28 | 2017-02-08 | 武汉大学 | Component-based fine vehicle model recognition method |
CN108805207A (en) * | 2018-06-13 | 2018-11-13 | 江苏电力信息技术有限公司 | A kind of Large Construction vehicle raises arm detection algorithm |
CN109086866A (en) * | 2018-07-02 | 2018-12-25 | 重庆大学 | A kind of part two-value convolution method suitable for embedded device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103020582A (en) * | 2012-09-20 | 2013-04-03 | 苏州两江科技有限公司 | Method for computer to identify vehicle type by video image |
US20150193663A1 (en) * | 2014-01-06 | 2015-07-09 | Hyundai Mobis Co., Ltd | Apparatus and method for recognizing traffic sign board |
-
2015
- 2015-12-04 CN CN201510884218.2A patent/CN105469062A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103020582A (en) * | 2012-09-20 | 2013-04-03 | 苏州两江科技有限公司 | Method for computer to identify vehicle type by video image |
US20150193663A1 (en) * | 2014-01-06 | 2015-07-09 | Hyundai Mobis Co., Ltd | Apparatus and method for recognizing traffic sign board |
Non-Patent Citations (2)
Title |
---|
TSUNG-HAN CHAN ET AL: "PCANet: A Simple Deep Learning Baseline for Image Classification?", 《JOURNAL OF LATEX CLASS FILES》 * |
马文华: "基于多特征融合的车型分类方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106384100A (en) * | 2016-09-28 | 2017-02-08 | 武汉大学 | Component-based fine vehicle model recognition method |
CN108805207A (en) * | 2018-06-13 | 2018-11-13 | 江苏电力信息技术有限公司 | A kind of Large Construction vehicle raises arm detection algorithm |
CN108805207B (en) * | 2018-06-13 | 2020-08-04 | 江苏电力信息技术有限公司 | Method for detecting large construction vehicle lifting arm |
CN109086866A (en) * | 2018-07-02 | 2018-12-25 | 重庆大学 | A kind of part two-value convolution method suitable for embedded device |
CN109086866B (en) * | 2018-07-02 | 2021-07-30 | 重庆大学 | Partial binary convolution method suitable for embedded equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20190228268A1 (en) | Method and system for cell image segmentation using multi-stage convolutional neural networks | |
CN109543548A (en) | A kind of face identification method, device and storage medium | |
CN109002755B (en) | Age estimation model construction method and estimation method based on face image | |
CN103984959A (en) | Data-driven and task-driven image classification method | |
CN107657233A (en) | Static sign language real-time identification method based on modified single multi-target detection device | |
CN107292250A (en) | A kind of gait recognition method based on deep neural network | |
CN110363290A (en) | A kind of image-recognizing method based on hybrid production style, device and equipment | |
CN103366180A (en) | Cell image segmentation method based on automatic feature learning | |
Li et al. | Pedestrian detection based on deep learning model | |
CN108108751A (en) | A kind of scene recognition method based on convolution multiple features and depth random forest | |
CN112580458B (en) | Facial expression recognition method, device, equipment and storage medium | |
Liu et al. | Visual attention in deep learning: a review | |
CN106295501A (en) | The degree of depth based on lip movement study personal identification method | |
CN104700100A (en) | Feature extraction method for high spatial resolution remote sensing big data | |
CN113920516B (en) | Calligraphy character skeleton matching method and system based on twin neural network | |
Dai et al. | Tan: Temporal aggregation network for dense multi-label action recognition | |
Sethy et al. | Off-line Odia handwritten numeral recognition using neural network: a comparative analysis | |
Lee et al. | Fast object localization using a CNN feature map based multi-scale search | |
CN105469062A (en) | Principal component analysis network framework based vehicle type identification method | |
Biglari et al. | Persian/Arabic handwritten digit recognition using local binary pattern | |
CN111310820A (en) | Foundation meteorological cloud chart classification method based on cross validation depth CNN feature integration | |
CN110728238A (en) | Personnel re-detection method of fusion type neural network | |
CN114882278A (en) | Tire pattern classification method and device based on attention mechanism and transfer learning | |
CN110688966A (en) | Semantic-guided pedestrian re-identification method | |
Pham et al. | A deep learning approach for real-time 3D human action recognition from skeletal data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into 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: 20160406 |