CN110287747A - A kind of bar code detection method based on end-to-end depth network - Google Patents
A kind of bar code detection method based on end-to-end depth network Download PDFInfo
- Publication number
- CN110287747A CN110287747A CN201910585431.1A CN201910585431A CN110287747A CN 110287747 A CN110287747 A CN 110287747A CN 201910585431 A CN201910585431 A CN 201910585431A CN 110287747 A CN110287747 A CN 110287747A
- Authority
- CN
- China
- Prior art keywords
- network
- cbl
- bar code
- detection method
- convolution
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 34
- 238000000926 separation method Methods 0.000 claims abstract description 7
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 239000000203 mixture Substances 0.000 claims description 5
- 101100194362 Schizosaccharomyces pombe (strain 972 / ATCC 24843) res1 gene Proteins 0.000 claims description 3
- 101100194363 Schizosaccharomyces pombe (strain 972 / ATCC 24843) res2 gene Proteins 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000013138 pruning Methods 0.000 claims description 3
- 239000011248 coating agent Substances 0.000 claims description 2
- 238000000576 coating method Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 238000000034 method Methods 0.000 abstract description 10
- 230000008901 benefit Effects 0.000 abstract description 6
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 4
- 238000007689 inspection Methods 0.000 description 2
- 241000208125 Nicotiana Species 0.000 description 1
- 235000002637 Nicotiana tabacum Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000008199 coating composition Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/10544—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation by scanning of the records by radiation in the optical part of the electromagnetic spectrum
- G06K7/10821—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation by scanning of the records by radiation in the optical part of the electromagnetic spectrum further details of bar or optical code scanning devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/1408—Methods for optical code recognition the method being specifically adapted for the type of code
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Electromagnetism (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Toxicology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of bar code detection methods based on end-to-end depth network;The present invention joined multi-scale prediction using end to end network using depth e-learning characteristics of image in a network, and pointedly corresponding various sizes of anchors be added all to predict various sizes of object for different scale;There is higher detection accuracy to wisp;In terms of image characteristics extraction, network of the invention has used for reference the way of residual error network, and quick connection is provided between some layers, and using cavity separation convolution, cut operator is added, reduces convolution output channel number, reduces parameter redundancy;Compared with existing recognition methods, advantage is to detect speed faster;In warehouse automation management aspect, the present invention abandons the conventional automated method that expansion is poor, precision is low, then carries out decision with the route that is aligned of counter to lorry by camera based on deep learning, improves the flexibility ratio of Automatic Warehouse.
Description
Technical field
The present invention relates to a kind of bar code detection method, in particular to a kind of bar code inspection based on end-to-end depth network
Survey method belongs to target detection and Automatic Warehouse technical field.
Background technique
Automatic Warehouse is also automatic access warehouse, and referring to can deposit in the case where directly not carrying out artificial treatment automatically
Storage and the system for taking out material, are widely used in the industries such as machinery, household electrical appliances, automobile, food, tobacco.With China's " intelligence manufacture
The propulsion of 2025 " plans and the continuous development of deep learning, Automatic Warehouse technology constantly improve, and utilization rate is continuously improved,
The requirement of detection accuracy is also higher and higher.
Nowadays, most Automatic Warehouses still rely on the systems such as automatically guiding trolley (AGV), automatic shelf, automatic sorting.But
The guide wire of AGV electromagnetic navigation is embedded in underground, change or extended route relative difficulty;The tape of tape navigation is easy breakage, needs
Periodic maintenance is wanted, and the trolley motion track of the two is all fixed, and cannot achieve intelligent evacuation.Automatic shelf structure is complicated, at
This is higher, and has higher requirements to warehouse fineness.
Therefore, existing warehousing system expansion, precision and in terms of equal existing defects, need
It improves.
Summary of the invention
The invention proposes a kind of bar code detection method based on end-to-end depth network, solve in the prior art certainly
The problem of lorry moving track limitation in dynamicization warehouse.
In order to solve the above-mentioned technical problems, the present invention provides the following technical solutions:
A kind of bar code detection method based on end-to-end depth network of the present invention, the specific inspection of the object detection method
Steps are as follows for survey:
Step 1: importation: inquiring into different inputs, i.e., input original RGB image or RGB image are transformed into single pass
Gray level image, both input the influence to network performance;
Step 2: data prediction and model training part: being enhanced by data, improve the robustness of network;Network can
It is initialized with the network of the pre-training on other data sets, then the test section of network is finely tuned and trained with target data set
Point;
Step 3: network portion: mainly by convolution block (CBL), up-sampling layer (UPS), residual unit (RES Unit) with
And residual block (RESn) composition, by the processing of network portion, various sizes of testing result Y (1), Y (2), Y can be exported
(3);
Step 4: obtaining the bounding box for corresponding to original image, Y (1), Y (2) and Y (3) according to output Y (1), Y (2) and Y (3)
Port number required with prediction it is related;
Step 5: error calculation: the global error of network is made of four parts, is done to the centre coordinate of prediction respectively
It loses, the wide height of predicted boundary frame is lost, loss is done to the classification of prediction and the confidence level of prediction is lost;
Step 6: application network: being detected, obtained small using the image that trained network obtains warehouse camera
The barcode position of vehicle and cargo is simultaneously individually identified, and the motion track of lorry is instructed according to the positional relationship both on image, with
Improve warehouse automation efficiency.
A kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In step 2, the data enhancing includes but are not limited to rotation, overturning and sectioning to picture;Other described data
Integrate as imagenet data set.
As a preferred technical solution of the present invention, the expression formula of the whole network frame following formula table in step 3
Show:
S (1)=RES8 (RES2 (RES1 (CBL (X))))
S (2)=RES8 (S (1))
S (3)=CBL*5 (RES4 (S (2)))
Y (1)=CONV (CBL (S (3)))
S (4)=CBL*5 (CONCAT (S (2), UPS (CBL (S (3)))))
Y (2)=CONV (CBL (S (4)))
Y (3)=CONV (CBL (CBL*5 (CONCAT (S (1), UPS (CBL (S (4)))))))
Wherein, CONCAT function representation channel is coupled, and input is the identical three-dimensional tensor of two spaces size, output
Bulk is also identical as input, and port number is the sum of the port number of two inputs;The input of CONCAT function is usually advanced
The up-sampling of characteristic pattern and time machine characteristic pattern, it is therefore an objective in conjunction with different resolution characteristic pattern with obtain it is more accurate, to tiny
The more sensitive network of object.
As a preferred technical solution of the present invention, in step 3, the convolution block is by convolutional layer (CONV), batch mark
Quasi- layer and active coating composition;It include two convolution blocks and a residual computations inside the residual unit;The residual block includes
The convolution block and several residual units composition that convolution step-length is 2, the length n of the residual block is of internal residual unit
Number n.
As a preferred technical solution of the present invention, all convolution operations in the convolution block are cavity separation volume
Product.
All convolution are reduced by beta pruning in the part for extracting feature as a preferred technical solution of the present invention
The port number of output, to reduce computation complexity under conditions of not reducing detection accuracy.
As a preferred technical solution of the present invention, in step 4, the network output channel number is 18, each
3 bounding boxes are predicted in the spatial position of output, the parameter of the bounding box of each prediction include the center point coordinate of bounding box (x,
Y), the object type cls of the width high (w, h) of bounding box, confidence level c and the bounding box, prediction object have 1 class, but in order to increase
The editability of network, therefore increase the parameter for representing categorical measure, therefore its output channel number is 3* (5+1)=18.
As a preferred technical solution of the present invention, in step 5, the centre coordinate loss are as follows:
The wide high loss are as follows:
The classification loss are as follows:
The confidence level loss are as follows:
The beneficial effects obtained by the present invention are as follows being: a kind of barcode detection side based on end-to-end depth network of the invention
Method compared with prior art, have it is below the utility model has the advantages that
1, the present invention joined multiple dimensioned using depth e-learning characteristics of image using end to end network in a network
Prediction, and pointedly corresponding various sizes of anchors is added all to predict various sizes of object for different scale;With
Existing identifying system is compared, and advantage is there is higher detection accuracy to wisp.
2, in terms of image characteristics extraction, network of the invention has used for reference the way of residual error network, sets between some layers
It has set quick connection (shortcut connections), and using cavity separation convolution, cut operator has been added, it is defeated to reduce convolution
Port number out reduces parameter redundancy;Compared with existing recognition methods, advantage is to detect speed faster.
3, in warehouse automation management aspect, the present invention abandons the conventional automated method that expansion is poor, precision is low, turns
And decision is carried out with the route that is aligned of counter to lorry by camera based on deep learning, improve the spirit of Automatic Warehouse
Activity.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the target detection in a kind of bar code detection method based on end-to-end depth network proposed by the present invention
Network structure;
Fig. 2 is in a kind of Fig. 1 structure chart of the bar code detection method based on end-to-end depth network proposed by the present invention
The structure of CBL;
Fig. 3 is in a kind of Fig. 1 structure chart of the bar code detection method based on end-to-end depth network proposed by the present invention
The structure of RESN;
Fig. 4 is in a kind of Fig. 1 structure chart of the bar code detection method based on end-to-end depth network proposed by the present invention
The structure of RES unit.
Specific embodiment
Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings, it should be understood that preferred reality described herein
Apply example only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
Embodiment 1
As shown in Figs 1-4, the present invention provides a kind of bar code detection method based on end-to-end depth network, target detection
The specific detecting step of method is as follows:
Step 1: importation: inquiring into different inputs, i.e., input original RGB image or RGB image are transformed into single pass
Gray level image, both input the influence to network performance;
Step 2: data prediction and model training part: being enhanced by data, improve the robustness of network;Network can
It is initialized with the network of the pre-training on other data sets, then the test section of network is finely tuned and trained with target data set
Point;
Step 3: network portion: mainly by convolution block (CBL), up-sampling layer (UPS), residual unit (RES Unit) with
And residual block (RESn) composition, the expression formula of whole network frame are indicated with following formula:
S (1)=RES8 (RES2 (RES1 (CBL (X))))
S (2)=RES8 (S (1))
S (3)=CBL*5 (RES4 (S (2)))
Y (1)=CONV (CBL (S (3)))
S (4)=CBL*5 (CONCAT (S (2), UPS (CBL (S (3)))))
Y (2)=CONV (CBL (S (4)))
Y (3)=CONV (CBL (CBL*5 (CONCAT (S (1), UPS (CBL (S (4)))))))
Wherein, CONCAT function representation channel is coupled, and input is the identical three-dimensional tensor of two spaces size, output
Bulk is also identical as input, and port number is the sum of the port number of two inputs;The input of CONCAT function is usually advanced
The up-sampling of characteristic pattern and time machine characteristic pattern, it is therefore an objective in conjunction with different resolution characteristic pattern with obtain it is more accurate, to tiny
The more sensitive network of object;The output of network is Y (1), Y (2), Y (3), is various sizes of testing result respectively;
There is the vector of one 18 dimension on each spatial position, 3 bounding boxes are predicted in the spatial position of each output, often
The parameter of the bounding box of a prediction includes the center point coordinate (x, y) of bounding box, the width high (w, h) of bounding box, confidence level c and is somebody's turn to do
The object type cls of bounding box, prediction object have 1 class, but in order to increase the editability of network, therefore increase by one and represent classification
The parameter of quantity, therefore its output channel number is 3* (5+1)=18;
Moreover, convolution block is made of convolutional layer (CONV), batch index bed and active coating;It include two volumes inside residual unit
Block and a residual computations;Residual block includes the convolution block that a convolution step-length is 2 and several residual units composition, residual block
Length n be internal residual unit number n;All convolution operations in convolution block are cavity separation convolution.
Step 4: obtaining the bounding box for corresponding to original image, Y (1), Y (2) and Y (3) according to output Y (1), Y (2) and Y (3)
Port number required with prediction it is related;
Step 5: error calculation: the global error of network is made of four parts, is done to the centre coordinate of prediction respectively
It loses, the wide height of predicted boundary frame is lost, loss is done to the classification of prediction and the confidence level of prediction is lost;
Centre coordinate loses
Wide high loss are as follows:
Classification is lost
Confidence level is lost
Wherein, λ is weighting parameters, and i is cell index number, and j is bounding box index number (totally 3 bounding boxes),
Is defined as:
Step 6: application network: being detected, obtained small using the image that trained network obtains warehouse camera
The barcode position of vehicle and cargo is simultaneously individually identified, and the motion track of lorry is instructed according to the positional relationship both on image, with
Improve warehouse automation efficiency.
In step 2, data enhancing includes but are not limited to rotation, overturning and sectioning to picture;Other data
Integrate as imagenet data set.
In the part for extracting feature, the port number of all convolution outputs is reduced by beta pruning, not reduce detection essence
Computation complexity is reduced under conditions of degree.
As shown in Figure 1, being the structure chart of whole network, input as an original image, network exports 3 different scales
Testing result.
As shown in Fig. 2, being that the internal structure of convolution block uniformly replaces with traditional convolution operation to improve arithmetic speed
Cavity separation convolution, in the identical situation of convolution size, convolution kernel of the invention has bigger receptive field, and utilizes separation volume
Product further speed-raising.
As shown in Figure 3 and Figure 4, the respectively internal structure of residual error module and residual unit;It is preferably quasi- to there is network
Conjunction ability, therefore residual error module is added, in not over-fitting and gradient disappearance or gradient is avoided to increase network layer under the premise of explosion
Number.
The present invention joined multiple dimensioned pre- using depth e-learning characteristics of image using end to end network in a network
It surveys, and pointedly corresponding various sizes of anchors is added all to predict various sizes of object for different scale;With it is existing
Some identifying systems are compared, and advantage is there is higher detection accuracy to wisp;It is of the invention in terms of image characteristics extraction
Network has used for reference the way of residual error network, and quick connection (shortcut connections) is provided between some layers, and
Convolution is separated using cavity, cut operator is added, reduces convolution output channel number, reduces parameter redundancy;With existing recognition methods
It compares, advantage is to detect speed faster;In warehouse automation management aspect, the present invention abandons that expansion is poor, precision is low
Conventional automated method, then decision is carried out to the route that is aligned of lorry and counter by camera based on deep learning,
Improve the flexibility ratio of Automatic Warehouse.
Finally, it should be noted that these are only the preferred embodiment of the present invention, it is not intended to restrict the invention, although
Present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, still can be right
Technical solution documented by foregoing embodiments is modified or equivalent replacement of some of the technical features.It is all
Within the spirit and principles in the present invention, any modification, equivalent replacement, improvement and so on should be included in protection of the invention
Within the scope of.
Claims (8)
1. a kind of bar code detection method based on end-to-end depth network, which is characterized in that the tool of the object detection method
Steps are as follows for physical examination survey:
Step 1: importation: inquiring into different inputs, i.e., input original RGB image or RGB image are transformed into single pass gray scale
Image, both input the influence to network performance;
Step 2: data prediction and model training part: being enhanced by data, improve the robustness of network;Network can use it
The network of pre-training on his data set initializes, then the detection part of network is finely tuned and trained with target data set;
Step 3: network portion: mainly by convolution block (CBL), up-sampling layer (UPS), residual unit (RES Unit) and residual
Poor block (RESn) composition, by the processing of network portion, can export various sizes of testing result Y (1), Y (2), Y (3);
Step 4: obtain the bounding box for corresponding to original image according to output Y (1), Y (2) and Y (3), Y (1), Y (2) and Y's (3) leads to
Road number requires with prediction related;
Step 5: error calculation: the global error of network is made of four parts, is damaged to the centre coordinate of prediction respectively
It loses, the wide height of predicted boundary frame is lost, loss is done to the classification of prediction and the confidence level of prediction is lost;
Step 6: application network: detected using the image that trained network obtains warehouse camera, obtain trolley and
The barcode position of cargo is simultaneously individually identified, and the motion track of lorry is instructed according to the positional relationship both on image, to improve
Warehouse automation efficiency.
2. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In step 2, the data enhancing includes but are not limited to rotation, overturning and sectioning to picture;Other described data sets
For imagenet data set.
3. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
The expression formula of whole network frame is indicated with following formula in step 3:
S (1)=RES8 (RES2 (RES1 (CBL (X))))
S (2)=RES8 (S (1))
S (3)=CBL*5 (RES4 (S (2)))
Y (1)=CONV (CBL (S (3)))
S (4)=CBL*5 (CONCAT (S (2), UPS (CBL (S (3)))))
Y (2)=CONV (CBL (S (4)))
Y (3)=CONV (CBL (CBL*5 (CONCAT (S (1), UPS (CBL (S (4)))))))
Wherein, CONCAT function representation channel is coupled, and input is the identical three-dimensional tensor of two spaces size, the space of output
Size is also identical as input, and port number is the sum of the port number of two inputs;The input of CONCAT function is usually advanced features
The up-sampling of figure and time machine characteristic pattern, it is therefore an objective in conjunction with different resolution characteristic pattern with obtain it is more accurate, to small objects
More sensitive network.
4. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In step 3, the convolution block is made of convolutional layer (CONV), batch index bed and active coating;It include two inside the residual unit
A convolution block and a residual computations;The residual block includes the convolution block and several residual unit groups that a convolution step-length is 2
At the length n of the residual block is the number n of internal residual unit.
5. a kind of bar code detection method based on end-to-end depth network according to claim 4, which is characterized in that institute
Stating all convolution operations in convolution block is cavity separation convolution.
6. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In the part for extracting feature, the port number of all convolution outputs is reduced by beta pruning, under conditions of not reducing detection accuracy
Reduce computation complexity.
7. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In step 4, the network output channel number is 18, and 3 bounding boxes are predicted in the spatial position of each output, each prediction
The parameter of bounding box includes width high (w, h), confidence level c and the bounding box of the center point coordinate (x, y) of bounding box, bounding box
Object type cls, prediction object have 1 class, but in order to increase the editability of network, therefore increase the ginseng for representing categorical measure
Number, therefore its output channel number is 3* (5+1)=18.
8. a kind of bar code detection method based on end-to-end depth network according to claim 1, which is characterized in that
In step 5, the centre coordinate loss are as follows:
The wide high loss are as follows:
The classification loss are as follows:
The confidence level loss are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910585431.1A CN110287747A (en) | 2019-07-01 | 2019-07-01 | A kind of bar code detection method based on end-to-end depth network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910585431.1A CN110287747A (en) | 2019-07-01 | 2019-07-01 | A kind of bar code detection method based on end-to-end depth network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110287747A true CN110287747A (en) | 2019-09-27 |
Family
ID=68021497
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910585431.1A Pending CN110287747A (en) | 2019-07-01 | 2019-07-01 | A kind of bar code detection method based on end-to-end depth network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110287747A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310509A (en) * | 2020-03-12 | 2020-06-19 | 北京大学 | Real-time bar code detection system and method based on logistics waybill |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107085428A (en) * | 2017-05-18 | 2017-08-22 | 广州视源电子科技股份有限公司 | Intelligent mobile method, device, robot and storage medium |
CN109447034A (en) * | 2018-11-14 | 2019-03-08 | 北京信息科技大学 | Traffic mark detection method in automatic Pilot based on YOLOv3 network |
CN109508710A (en) * | 2018-10-23 | 2019-03-22 | 东华大学 | Based on the unmanned vehicle night-environment cognitive method for improving YOLOv3 network |
CN109543486A (en) * | 2018-10-29 | 2019-03-29 | 华南理工大学 | Bar code localization method neural network based and system |
-
2019
- 2019-07-01 CN CN201910585431.1A patent/CN110287747A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107085428A (en) * | 2017-05-18 | 2017-08-22 | 广州视源电子科技股份有限公司 | Intelligent mobile method, device, robot and storage medium |
CN109508710A (en) * | 2018-10-23 | 2019-03-22 | 东华大学 | Based on the unmanned vehicle night-environment cognitive method for improving YOLOv3 network |
CN109543486A (en) * | 2018-10-29 | 2019-03-29 | 华南理工大学 | Bar code localization method neural network based and system |
CN109447034A (en) * | 2018-11-14 | 2019-03-08 | 北京信息科技大学 | Traffic mark detection method in automatic Pilot based on YOLOv3 network |
Non-Patent Citations (1)
Title |
---|
谭俊: "一个改进的YOLOv3目标识别算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310509A (en) * | 2020-03-12 | 2020-06-19 | 北京大学 | Real-time bar code detection system and method based on logistics waybill |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109784203B (en) | Method for inspecting contraband in weak supervision X-ray image based on layered propagation and activation | |
Lacoste et al. | Point processes for unsupervised line network extraction in remote sensing | |
Besl et al. | Invariant surface characteristics for 3D object recognition in range images | |
NL2020405B1 (en) | Method of and apparatus for analyzing images | |
CN111552754B (en) | Ship track similarity measurement method and system | |
Luo et al. | The issues and approaches of a robot multi-sensor integration | |
Younis et al. | Detection and annotation of plant organs from digitised herbarium scans using deep learning | |
CN111797836A (en) | Extraterrestrial celestial body patrolling device obstacle segmentation method based on deep learning | |
CN109785333A (en) | Object detection method and device for parallel manipulator human visual system | |
CN114581782A (en) | Fine defect detection method based on coarse-to-fine detection strategy | |
CN104463826A (en) | Novel point cloud parallel Softassign registering algorithm | |
CN114743010B (en) | Ultrahigh voltage power transmission line point cloud data semantic segmentation method based on deep learning | |
CN113936220B (en) | Image processing method, storage medium, electronic device, and image processing apparatus | |
CN110287747A (en) | A kind of bar code detection method based on end-to-end depth network | |
Chen et al. | Detecting maneuvering target accurately based on a two-phase approach from remote sensing imagery | |
CN114566052A (en) | Method for judging rotation of highway traffic flow monitoring equipment based on traffic flow direction | |
CN116805157B (en) | Unmanned cluster autonomous dynamic evaluation method and device | |
CN113899361B (en) | SLAM rapid loop detection method based on space division | |
CN112634447B (en) | Outcrop stratum layering method, device, equipment and storage medium | |
CN115374880A (en) | Offshore target identification-oriented multistage incremental data fusion system | |
CN113112471A (en) | Target detection method based on RI-HOG characteristics and quick pyramid | |
Feng et al. | Target recognition and grabbing positioning method based on convolutional neural network | |
CN113963280B (en) | Identification method and device for intelligent detection and judgment of material and part and storage medium | |
CN109034695B (en) | Intelligent storage logistics state detection method | |
Panimayam Fernando | Prioritized 2D-3D matching for visual localization revisited |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190927 |
|
RJ01 | Rejection of invention patent application after publication |