CN109598303A - A kind of rubbish detection method based on City scenarios - Google Patents
A kind of rubbish detection method based on City scenarios Download PDFInfo
- Publication number
- CN109598303A CN109598303A CN201811464458.7A CN201811464458A CN109598303A CN 109598303 A CN109598303 A CN 109598303A CN 201811464458 A CN201811464458 A CN 201811464458A CN 109598303 A CN109598303 A CN 109598303A
- Authority
- CN
- China
- Prior art keywords
- rubbish
- picture
- method based
- city
- municipal refuse
- 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 OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
A kind of rubbish detection method based on City scenarios, first collection municipal refuse sample set, recycle LabelImg tool to carry out rubbish mark to municipal refuse sample set, and the picture in municipal refuse sample set is normalized；Then the samples pictures collection after conversion is input in Faster R-CNN algorithm, establish object-class model, and the object-class model established is verified using the picture of PASCAL VOC2012 data set, until statistics target classification accuracy rate is not less than 90%, picture decoding finally is carried out to the video containing City scenarios, classification and Detection is carried out to decoded picture using the Faster R-CNN garbage classification detection model of foundation, municipal waste management is monitored in real time to realize, greatly mitigate the operating pressure of city management personnel, also the equipment investment in municipal refuse testing and management can be reduced, with good compatibility and versatility.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of rubbish detection methods based on City scenarios.
Background technique
With economical fast development and social progress, population is increasingly assembled, and city size gradually expands, people
To the living environment depended on for existence, more stringent requirements are proposed.Municipal refuse as influence one of living environment it is important because
Element, being monitored how effectively to municipal refuse is just particularly important with management.Currently, being monitored for municipal refuse
Main path with management is the inspection of department, municipal administration or environmental sanitation department personnel, during inspection, by hand-held whole
The position of rubbish and picture are uploaded to command hall by end equipment, and command hall is put on record therewith, are then issued to related single
Processing result after responsible person concerned's processing, then is uploaded to command hall, tied after command hall confirmation by position and responsible person concerned
Case processing, in this way, needs to consume a large amount of manpower and material resources, very big pressure is increased to municipal administration and fiscal administration
Power.With the implementation of the engineerings such as " day net engineering ", " bright as snow engineering ", smart city, monitor video throughout each corner in city,
Science and technology is maked rapid progress, and video image identification technology is also applied to more and more industries, how by supervision of the cities and video
Image recognition technology combines, and effectively solves municipal refuse monitoring and management has become those skilled in the art's weight urgently to be resolved
Want problem.
Summary of the invention
Technical problem solved by the invention is to provide a kind of rubbish detection method based on City scenarios, on solving
State the disadvantage in background technique.
Technical problem solved by the invention is realized using following technical scheme:
A kind of rubbish detection method based on City scenarios, the specific steps are as follows:
Step 1) collects municipal refuse sample set；
Step 2 carries out rubbish mark to municipal refuse sample set using LabelImg tool, and in municipal refuse sample set
Picture is normalized；
Samples pictures collection after converting in step 2 is input in Faster R-CNN algorithm by step 3), establishes target classification mould
Type；
Step 4) is verified using the object-class model that the picture of PASCAL VOC2012 data set establishes step 3),
And count the accuracy rate of target classification；
The sample into step 3) is added when counting target classification accuracy rate lower than 90% in step 4) in false target by step 5)
In this pictures, then it is input to Faster R-CNN algorithm training iteration, is repeated in step 4), until statistics target classification is quasi-
True rate is not less than 90%, i.e., model is Fa ster R-CNN garbage classification detection model at this time；
Step 6) carries out picture decoding, then the Faster R-CNN rubbish established by step 5) to the video containing City scenarios
Classification and Detection model carries out classification and Detection to decoded picture, if it is determined that being rubbish, picture is sent to command hall foreground,
Foreground personnel recheck, and confirmation is rubbish, then carry out processing of putting on record according to city management process；If not rubbish, then
The picture of acquisition is added in samples pictures collection, is iterated training according to step 3), and obtain new object-class model,
It applies it in the rubbish detection of City scenarios.
In the present invention, in step 1), two kinds of sides of picture are shot on the spot by network collection municipal refuse picture and personnel
Formula collects municipal refuse sample set.
In the present invention, in step 2, the picture in municipal refuse sample set is carried out by way of linear normalization
Conversion.
In the present invention, in step 2, the samples pictures after conversion are having a size of 300*300.
In the present invention, in step 3), the step of establishing object-class model, is as follows:
The samples pictures after normalization are input in RPN network and Fast R-CNN network respectively, convolution is done to samples pictures,
To accelerate speed, parallelization acceleration is carried out using the process that CUDA does convolution to two networks respectively, is optimized by successive ignition
Afterwards, RPN network exports multiple dimensioned rubbish candidate region, chooses optimal 300 areas using nms algorithm (non-maxima suppression)
The threshold value in domain, nms algorithm is chosen for 0.7；Fast R-CNN network exports characteristic pattern, feature output figure and RPN after iteration
300 best region frames of network output map each other, then pass through s
Oftmax algorithm is classified, to obtain the position of rubbish in characteristic pattern and return acquisition best region using frame, together
When rubbish is marked.
In the present invention, in step 4), the object-class model established to step 3) is verified, and specifically first will
The picture of VOC2012 data set is normalized, then will be after the obtained spam samples collection of step 1) and normalization
VOC2012 pictures are blended in one, are input to object-class model and classify, and count the accuracy rate of target classification.
In the present invention, in step 6), the video flowing containing City scenarios in supervision of the cities is obtained by FFMPEG, simultaneously
Acceleration decoding is carried out using CUDA.
The utility model has the advantages that the present invention uses back-end analysis mode, the effect of monitor camera can be fully played, in real time to city
City's waste management is monitored, and is greatly mitigated the operating pressure of city management personnel, is contributed for administration of city appearance, also
The equipment investment in municipal refuse testing and management can be reduced, achievees the purpose that economize on resources, is had good compatible and logical
The property used.
Detailed description of the invention
Fig. 1 is the collection municipal refuse sample set schematic diagram in presently preferred embodiments of the present invention.
Fig. 2 is to mark schematic diagram using LabelImg tool in presently preferred embodiments of the present invention.
Fig. 3 is the Faster R-CNN algorithm flow schematic diagram in presently preferred embodiments of the present invention.
Fig. 4 is the PASCAL VOC2012 data set picture schematic diagram in presently preferred embodiments of the present invention.
Fig. 5 is the operating process schematic diagram of presently preferred embodiments of the present invention.
Fig. 6 is the rubbish identification schematic diagram of presently preferred embodiments of the present invention.
Specific embodiment
In order to be easy to understand the technical means, the creative features, the aims and the efficiencies achieved by the present invention, tie below
Conjunction is specifically illustrating, and the present invention is further explained.
Referring to a kind of rubbish detection method based on City scenarios shown in Fig. 1~6, the specific steps are as follows:
Step 1) shoots picture two ways by network collection municipal refuse picture and personnel on the spot and collects municipal refuse sample
Collect, totally 2000 picture, as shown in Figure 1；
Step 2 carries out rubbish mark to municipal refuse sample set using LabelImg tool, and by way of linear normalization
Convert sample to the samples pictures of size 300*300；
The samples pictures collection that size 300*300 is converted into step 2 is input in Faster R-CNN algorithm by step 3), is built
Vertical object-class model；
Faster R-CNN algorithm training step as shown in figure 3, respectively by the samples pictures after normalization be input to RPN network with
In Fast R-CNN network, convolution is done to samples pictures, in order to accelerate speed, the present embodiment is using CUDA respectively to two networks
The process for doing convolution carries out parallelization acceleration, and after successive ignition optimizes, RPN network exports multiple dimensioned rubbish candidate regions
Optimal 300 regions are chosen using nms algorithm (non-maxima suppression) in domain, and the threshold value of nms algorithm is chosen for 0.7；Fast
R-CNN network exports characteristic pattern after iteration, and 300 best region frames that feature output figure is exported with RPN network reflect each other
It penetrates, is then classified by softmax algorithm, to obtain the position of rubbish in characteristic pattern and return acquisition most preferably using frame
Region, while rubbish is marked, i.e., Faster R-CNN algorithm parameter model is required object-class model at this time；
Step 4) builds step 3) using the picture of PASCAL VOC2012 data set (international standard data set is as shown in Figure 4)
Vertical object-class model is verified, and specifically first the picture of VOC2012 data set is normalized, then by step
1) obtained spam samples collection and normalization after VOC2012 pictures be blended in one, be input to object-class model into
Row classification, counts the accuracy rate of target classification；
The sample into step 3) is added when counting target classification accuracy rate lower than 90% in step 4) in false target by step 5)
In this pictures, then it is input to Faster R-CNN algorithm training iteration, is repeated in step 4), until statistics target classification is quasi-
True rate is not less than 90%.I.e. model is Faster R-CNN garbage classification detection model at this time；
Step 6) detects the picture containing City scenarios according to process shown in Fig. 5, firstly, obtaining city by FFMPEG
The video flowing of monitoring carries out acceleration decoding using CUDA, and decoded picture is fed through Faster R-CNN rubbish detection model
Classification and Detection is carried out, if it is determined that being rubbish (being rubbish in box as shown in Figure 6), picture is sent to command hall foreground, it is preceding
Platform staff rechecks, and confirmation is rubbish, then carries out processing of putting on record according to city management process, if not rubbish, then will
The picture of acquisition is added in samples pictures collection, is iterated training according to step 3), and obtain new object-class model, will
It is applied in the rubbish detection of City scenarios.
The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (8)
1. a kind of rubbish detection method based on City scenarios, which is characterized in that specific step is as follows:
Step 1) collects municipal refuse sample set；
Step 2 carries out rubbish mark to municipal refuse sample set using LabelImg tool, and in municipal refuse sample set
Picture is normalized；
Samples pictures collection after converting in step 2 is input in Faster R-CNN algorithm by step 3), establishes target classification mould
Type；
Step 4) is verified using the object-class model that the picture of PASCAL VOC2012 data set establishes step 3),
And count the accuracy rate of target classification；
The sample into step 3) is added when counting target classification accuracy rate lower than 90% in step 4) in false target by step 5)
In this pictures, then it is input to Faster R-CNN algorithm training iteration, is repeated in step 4), until statistics target classification is quasi-
True rate is not less than 90%, i.e., model is Faster R-CNN garbage classification detection model at this time；
Step 6) carries out picture decoding, then the Faster R-CNN rubbish established by step 5) to the video containing City scenarios
Classification and Detection model carries out classification and Detection to decoded picture, if it is determined that being rubbish, picture is sent to command hall foreground,
Foreground personnel recheck, and confirmation is rubbish, then carry out processing of putting on record according to city management process；If not rubbish, then
The picture of acquisition is added in samples pictures collection, is iterated training according to step 3), and obtain new object-class model,
It applies it in the rubbish detection of City scenarios.
2. a kind of rubbish detection method based on City scenarios according to claim 1, which is characterized in that in step 1),
Picture two ways is shot on the spot by network collection municipal refuse picture and personnel collects municipal refuse sample set.
3. a kind of rubbish detection method based on City scenarios according to claim 1, which is characterized in that in step 2,
The picture in municipal refuse sample set is converted by way of linear normalization.
4. a kind of rubbish detection method based on City scenarios according to claim 3, which is characterized in that in step 2,
Samples pictures after conversion are having a size of 300*300.
5. a kind of rubbish detection method based on City scenarios according to claim 1, which is characterized in that in step 3),
The step of establishing object-class model is as follows:
The samples pictures after normalization are input in RPN network and Fast R-CNN network respectively, convolution is done to samples pictures,
After successive ignition optimizes, RPN network exports multiple dimensioned rubbish candidate region, and optimal 300 using nms algorithm picks
The threshold value in region, nms algorithm is chosen for 0.7；Fast R-CNN network exports characteristic pattern after iteration, feature output figure with
300 best region frames of RPN network output map each other, are then classified by softmax algorithm, to obtain characteristic pattern
The position of middle rubbish simultaneously returns acquisition best region using frame, while rubbish being marked.
6. a kind of rubbish detection method based on City scenarios according to claim 5, which is characterized in that accelerate speed
Degree carries out parallelization acceleration using the process that CUDA does convolution to two networks respectively.
7. a kind of rubbish detection method based on City scenarios according to claim 1, which is characterized in that in step 4),
The object-class model established to step 3) is verified, and specifically first the picture of VOC2012 data set is normalized
Processing, then the VOC2012 pictures after the obtained spam samples collection of step 1) and normalization are blended in one, it is input to mesh
Mark disaggregated model is classified, and the accuracy rate of target classification is counted.
8. a kind of rubbish detection method based on City scenarios according to claim 1, which is characterized in that in step 6),
The video flowing containing City scenarios in supervision of the cities is obtained by FFMPEG, while carrying out acceleration decoding using CUDA.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811464458.7A CN109598303A (en) | 2018-12-03 | 2018-12-03 | A kind of rubbish detection method based on City scenarios |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811464458.7A CN109598303A (en) | 2018-12-03 | 2018-12-03 | A kind of rubbish detection method based on City scenarios |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109598303A true CN109598303A (en) | 2019-04-09 |
Family
ID=65959440
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811464458.7A Pending CN109598303A (en) | 2018-12-03 | 2018-12-03 | A kind of rubbish detection method based on City scenarios |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109598303A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598800A (en) * | 2019-09-23 | 2019-12-20 | 山东浪潮人工智能研究院有限公司 | Garbage classification and identification method based on artificial intelligence |
| CN110610201A (en) * | 2019-08-30 | 2019-12-24 | 厦门快商通科技股份有限公司 | Kitchen waste recycling and classifying method and system, mobile terminal and storage medium |
| CN111124862A (en) * | 2019-12-24 | 2020-05-08 | 北京安兔兔科技有限公司 | Intelligent equipment performance testing method and device and intelligent equipment |
| CN111160123A (en) * | 2019-12-11 | 2020-05-15 | 桂林长海发展有限责任公司 | Airplane target identification method and device and storage medium |
| CN111242010A (en) * | 2020-01-10 | 2020-06-05 | 厦门博海中天信息科技有限公司 | Method for judging and identifying identity of litter worker based on edge AI |
| CN112508103A (en) * | 2020-12-10 | 2021-03-16 | 浙江金实乐环境工程有限公司 | Perishable garbage image identification and assessment management method based on garbage collection and transportation vehicle |
| CN112949668A (en) * | 2019-12-10 | 2021-06-11 | 东北大学秦皇岛分校 | Garbage detection system based on deep learning |
| CN113051963A (en) * | 2019-12-26 | 2021-06-29 | 中移(上海)信息通信科技有限公司 | Garbage detection method and device, electronic equipment and computer storage medium |
| CN113159228A (en) * | 2021-05-19 | 2021-07-23 | 江苏奥易克斯汽车电子科技股份有限公司 | Garbage classification identification method and device based on deep learning and intelligent garbage can |
| CN116189040A (en) * | 2022-12-27 | 2023-05-30 | 中国电信股份有限公司 | Method, device, storage medium and electronic equipment for city appearance environment monitoring |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203498A (en) * | 2016-07-07 | 2016-12-07 | 中国科学院深圳先进技术研究院 | A kind of City scenarios rubbish detection method and system |
| CN106845408A (en) * | 2017-01-21 | 2017-06-13 | 浙江联运知慧科技有限公司 | A kind of street refuse recognition methods under complex environment |
| US20180307926A1 (en) * | 2017-04-21 | 2018-10-25 | Ford Global Technologies, Llc | Stain and Trash Detection Systems and Methods |
-
2018
- 2018-12-03 CN CN201811464458.7A patent/CN109598303A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203498A (en) * | 2016-07-07 | 2016-12-07 | 中国科学院深圳先进技术研究院 | A kind of City scenarios rubbish detection method and system |
| CN106845408A (en) * | 2017-01-21 | 2017-06-13 | 浙江联运知慧科技有限公司 | A kind of street refuse recognition methods under complex environment |
| US20180307926A1 (en) * | 2017-04-21 | 2018-10-25 | Ford Global Technologies, Llc | Stain and Trash Detection Systems and Methods |
Non-Patent Citations (1)
| Title |
|---|
| 魏书法等: "基于图像的城市场景垃圾自动检测", 《集成技术》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110610201A (en) * | 2019-08-30 | 2019-12-24 | 厦门快商通科技股份有限公司 | Kitchen waste recycling and classifying method and system, mobile terminal and storage medium |
| CN110610201B (en) * | 2019-08-30 | 2022-06-07 | 厦门快商通科技股份有限公司 | Kitchen waste recycling and classifying method and system, mobile terminal and storage medium |
| CN110598800A (en) * | 2019-09-23 | 2019-12-20 | 山东浪潮人工智能研究院有限公司 | Garbage classification and identification method based on artificial intelligence |
| CN112949668A (en) * | 2019-12-10 | 2021-06-11 | 东北大学秦皇岛分校 | Garbage detection system based on deep learning |
| CN111160123A (en) * | 2019-12-11 | 2020-05-15 | 桂林长海发展有限责任公司 | Airplane target identification method and device and storage medium |
| CN111160123B (en) * | 2019-12-11 | 2023-06-09 | 桂林长海发展有限责任公司 | Aircraft target identification method, device and storage medium |
| CN111124862A (en) * | 2019-12-24 | 2020-05-08 | 北京安兔兔科技有限公司 | Intelligent equipment performance testing method and device and intelligent equipment |
| CN111124862B (en) * | 2019-12-24 | 2024-01-30 | 北京安兔兔科技有限公司 | Intelligent device performance testing method and device and intelligent device |
| CN113051963A (en) * | 2019-12-26 | 2021-06-29 | 中移(上海)信息通信科技有限公司 | Garbage detection method and device, electronic equipment and computer storage medium |
| CN111242010A (en) * | 2020-01-10 | 2020-06-05 | 厦门博海中天信息科技有限公司 | Method for judging and identifying identity of litter worker based on edge AI |
| CN112508103A (en) * | 2020-12-10 | 2021-03-16 | 浙江金实乐环境工程有限公司 | Perishable garbage image identification and assessment management method based on garbage collection and transportation vehicle |
| CN112508103B (en) * | 2020-12-10 | 2024-06-04 | 浙江金实乐环境工程有限公司 | Perishable garbage image identification and assessment management method based on garbage collection and transportation vehicle |
| CN113159228A (en) * | 2021-05-19 | 2021-07-23 | 江苏奥易克斯汽车电子科技股份有限公司 | Garbage classification identification method and device based on deep learning and intelligent garbage can |
| CN116189040A (en) * | 2022-12-27 | 2023-05-30 | 中国电信股份有限公司 | Method, device, storage medium and electronic equipment for city appearance environment monitoring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109598303A (en) | A kind of rubbish detection method based on City scenarios | |
| CN110119662A (en) | A kind of rubbish category identification system based on deep learning | |
| CN110348376A (en) | A kind of pedestrian's real-time detection method neural network based | |
| CN109697424A (en) | A kind of high-speed railway impurity intrusion detection device and method based on FPGA and deep learning | |
| CN110294236A (en) | A kind of garbage classification monitoring apparatus, method and server system | |
| CN110659622A (en) | Detection method, device and system for garbage dumping | |
| CN111814750A (en) | Intelligent garbage classification method and system based on deep learning target detection and image recognition | |
| CN110503070A (en) | Traffic automation monitoring method based on aerial image target detection and processing technology | |
| CN111814742A (en) | Knife switch state recognition method based on deep learning | |
| CN101923561A (en) | Automatic document classifying method | |
| CN111462167A (en) | Intelligent terminal video analysis algorithm combining edge calculation and deep learning | |
| CN107909659A (en) | A kind of wisdom curb parking charging method based on wireless video sensor network | |
| CN111145222A (en) | Fire detection method combining smoke movement trend and textural features | |
| CN117386434A (en) | Three-dimensional GIS multi-mode green mine intelligent road dust suppression system | |
| CN110826577A (en) | High-voltage isolating switch state tracking identification method based on target tracking | |
| CN111507379A (en) | Ore automatic identification and rough sorting system based on deep learning | |
| CN110276300A (en) | Method and device for identifying waste quality | |
| CN101998115A (en) | Embedded-type network camera with passenger flow counting function and passenger flow counting method | |
| CN110694934A (en) | Intelligent dry garbage classification cloud system and working method thereof | |
| CN113947737A (en) | Method and device for detecting abnormal waste in waste incineration power plant | |
| CN117728570A (en) | A power grid video analysis system and method based on edge computing | |
| CN117576602A (en) | A hydropower station valve leakage detection and early warning method and system | |
| CN118691789A (en) | A construction waste detection method based on improved YOLOv5 model | |
| CN117475276B (en) | Model training methods, garbage detection methods and devices | |
| CN118072302A (en) | Detection method and system for multiple-ripeness citrus fruits in unstructured 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: 20190409 |