CN110276300A - The method and apparatus of rubbish quality for identification - Google Patents
The method and apparatus of rubbish quality for identification Download PDFInfo
- Publication number
- CN110276300A CN110276300A CN201910547978.2A CN201910547978A CN110276300A CN 110276300 A CN110276300 A CN 110276300A CN 201910547978 A CN201910547978 A CN 201910547978A CN 110276300 A CN110276300 A CN 110276300A
- Authority
- CN
- China
- Prior art keywords
- rubbish
- image
- identified
- training sample
- recognition result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000010813 municipal solid waste Substances 0.000 title claims abstract description 357
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000012549 training Methods 0.000 claims abstract description 101
- 238000013136 deep learning model Methods 0.000 claims abstract description 48
- 238000001514 detection method Methods 0.000 claims description 13
- 238000003672 processing method Methods 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 230000003321 amplification Effects 0.000 claims description 5
- 230000003416 augmentation Effects 0.000 claims description 5
- 235000013399 edible fruits Nutrition 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 230000000306 recurrent effect Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 abstract description 3
- 238000012545 processing Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 230000015556 catabolic process Effects 0.000 description 6
- 230000006854 communication Effects 0.000 description 6
- 238000006731 degradation reaction Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 239000002361 compost Substances 0.000 description 4
- 244000005700 microbiome Species 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000012550 audit Methods 0.000 description 2
- 239000002551 biofuel Substances 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000029087 digestion Effects 0.000 description 2
- 238000010332 dry classification Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000013016 learning Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000005291 magnetic effect Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000010333 wet classification Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000010794 food waste Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present application discloses the method and apparatus of rubbish quality for identification.One specific embodiment of this method includes: the image sequence for obtaining rubbish to be identified;The image in the image sequence of rubbish to be identified is identified using training in advance, corresponding with the classification of rubbish to be identified deep learning model, obtain the recognition result of rubbish to be identified, wherein, the recognition result of rubbish to be identified includes the information of goal-selling object present in rubbish to be identified;The recognition result of rubbish to be identified is analyzed and counted, the quality results of rubbish to be identified are generated.The embodiment is related to field of cloud calculation, using goal-selling object present in deep learning model automatic identification rubbish, improves the recognition efficiency to rubbish quality.
Description
Technical field
The invention relates to field of computer technology, and in particular to the method and apparatus of rubbish quality for identification.
Background technique
House refuse discharge rate increasingly increases, and complicated component multiplicity has the characteristics that pollution, resource and social.
Rubbish " dry and wet classification " is, moisture content height higher for food waste in China's household garbage and pericarp class rubbish ratio, no
A kind of simple and practical garbage classification mode proposed conducive to garbage reclamation and the national conditions of final disposal." dry and wet classification " is will to occupy
The house refuse of the people is divided into wet refuse and dry rubbish.Compost, Anaerobic Digestion are carried out using microorganism after wet refuse collection
Or prepare bio-fuel.Therefrom choose available substance by staff after dry garbage collection, remaining rubbish is filled
Or incineration disposal.The percentage of the ratio between contained humidity weight and rubbish total weight is moisture content in dry rubbish, and moisture content is weighing apparatus
Measure one of the important reference indicator of dry rubbish quality.The long polybag of non-degradable or degradation time in wet refuse, bottle, easily
The content for drawing tank etc. is to measure one of the important reference indicator of wet refuse quality.
The purpose of garbage disposal is innoxious, recycling and minimizing.Therefore, difference is taken for the rubbish of different qualities
Processing method seem abnormal important.
Existing measurement dry and wet rubbish quality mainly relies on artificial sample detection method, i.e., random by carrying out to rubbish
Sampling Detection differentiates rubbish quality according to result of sampling inspection.
Summary of the invention
The embodiment of the present application proposes the method and apparatus of rubbish quality for identification.
In a first aspect, the embodiment of the present application provides a kind of method of rubbish quality for identification, comprising: obtain to be identified
The image sequence of rubbish;Using training in advance, corresponding with the classification of rubbish to be identified deep learning model to rubbish to be identified
Image in the image sequence of rubbish is identified, the recognition result of rubbish to be identified is obtained, wherein the identification knot of rubbish to be identified
Fruit includes the information of goal-selling object present in rubbish to be identified;The recognition result of rubbish to be identified is analyzed and counted,
Generate the quality results of rubbish to be identified.
In some embodiments, deep learning model includes feature extraction network, sorter network and Recurrent networks.
In some embodiments, the image sequence of rubbish to be identified be to rubbish to be identified topple over or shipment carry out
Acquire multiple obtained images.
In some embodiments, training in advance, corresponding with the classification of the rubbish to be identified deep learning is being utilized
Before model identifies the image in the image sequence of rubbish to be identified, further includes: treat knowledge using image processing method
Image in the image sequence of other rubbish is pre-processed.
In some embodiments, the image in the image sequence of rubbish to be identified is located in advance using image processing method
Reason, comprising: sky background image is sorted out from the image sequence of rubbish to be identified;For being removed in the image sequence of rubbish to be identified
Image except empty background image, using empty background image as the background of the image, using the detection of Moving target detection algorithm to
It identifies the region of rubbish in the images, and the pixel value in the image in addition to detected region is arranged to preset
Value.
In some embodiments, the image in the image sequence of rubbish to be identified is located in advance using image processing method
Reason, comprising: for the image in the image sequence of rubbish to be identified, by default screenshot box according to preset step-length along the default side of movement
To moving on this image, multiple subgraphs are intercepted;For the subgraph in multiple subgraphs, eliminated using median filter method
The isolated noise point of the subgraph.
In some embodiments, training in advance, corresponding with the classification of rubbish to be identified deep learning model is being utilized
Image in the image sequence of rubbish to be identified is identified, after obtaining the recognition result of rubbish to be identified, further includes: right
Image in the image sequence of rubbish to be identified determines the image using target circle if there are goal-selling objects in the image
In goal-selling object, as marking target;First is default before choosing the image in the image sequence of rubbish to be identified
Mesh image and rear second preset number image;Target following is carried out in the image of selected taking-up based on marking target;
The recognition result of rubbish to be identified is updated based on tracking result.
In some embodiments, the recognition result of rubbish to be identified is updated based on tracking result, comprising: statistics is selected to be taken out
Image in trace into marking target image number;If the number for tracing into the image of marking target is more than that third is pre-
If number, retain the information of the marking target in the recognition result of rubbish to be identified;If tracing into the image of marking target
Number be no more than third preset number, delete the information of the marking target in the recognition result of rubbish to be identified.
In some embodiments, training obtains deep learning model as follows: training sample set is obtained,
In, the training sample in training sample set includes sample rubbish image and sample rubbish mark image, sample rubbish mark figure
It seem that obtained image is marked to goal-selling object present in sample rubbish image;For in training sample set
Sample rubbish in the training sample is marked figure using the sample rubbish image in the training sample as input by training sample
As output, training obtains deep learning model.
In some embodiments, after obtaining training sample set, further includes: in training sample set without mark
Training sample is signed, which is generated by MixUp guess data amplification method;This is mixed without mark
Label training sample and corresponding there are label training sample, augmentation training sample set.
Second aspect, the embodiment of the present application provide a kind of device of rubbish quality for identification, comprising: acquiring unit,
It is configured to obtain the image sequence of rubbish to be identified;Recognition unit is configured to using training in advance and rubbish to be identified
The corresponding deep learning model of classification the image in the image sequence of rubbish to be identified is identified, obtain rubbish to be identified
Recognition result, wherein the recognition result of rubbish to be identified includes the information of goal-selling object present in rubbish to be identified;System
Unit is counted, is configured to analyze and count the recognition result of rubbish to be identified, generates the quality results of rubbish to be identified.
In some embodiments, deep learning model includes feature extraction network, sorter network and Recurrent networks.
In some embodiments, the image sequence of rubbish to be identified be to rubbish to be identified topple over or shipment carry out
Acquire multiple obtained images.
In some embodiments, described device further include: processing unit is configured to treat knowledge using image processing method
Image in the image sequence of other rubbish is pre-processed.
In some embodiments, processing unit is further configured to: being sorted out from the image sequence of rubbish to be identified
Empty background image;For the image in the image sequence of rubbish to be identified in addition to empty background image, using empty background image as
The background of the image detects the region of rubbish to be identified in the images using Moving target detection algorithm, and by the image
In pixel value in addition to detected region be arranged to preset value.
In some embodiments, processing unit is further configured to: for the figure in the image sequence of rubbish to be identified
Default screenshot box is moved according to preset step-length along moving direction is preset on this image, intercepts multiple subgraphs by picture;For more
The subgraph in subgraph is opened, the isolated noise point of the subgraph is eliminated using median filter method.
In some embodiments, device further include: marking unit is configured to the image sequence for rubbish to be identified
In image the goal-selling object in the image is determined using target circle, as label if there are goal-selling objects in the image
Object;Selection unit is configured to the first default mesh figure before choosing the image in the image sequence of rubbish to be identified
Picture and rear second preset number image;Tracking cell, be configured to based on marking target in the image of selected taking-up into
Row target following;Updating unit is configured to update the recognition result of rubbish to be identified based on tracking result.
In some embodiments, updating unit is further configured to: tracing into label in the selected image taken out of statistics
The number of the image of object;If the number for tracing into the image of marking target is more than third preset number, retain to be identified
The information of marking target in the recognition result of rubbish;If it is pre- that the number for tracing into the image of marking target is no more than third
If number, the information of the marking target in the recognition result of rubbish to be identified is deleted.
In some embodiments, training obtains deep learning model as follows: training sample set is obtained,
In, the training sample in training sample set includes sample rubbish image and sample rubbish mark image, sample rubbish mark figure
It seem that obtained image is marked to goal-selling object present in sample rubbish image;For in training sample set
Sample rubbish in the training sample is marked figure using the sample rubbish image in the training sample as input by training sample
As output, training obtains deep learning model.
In some embodiments, after obtaining training sample set, further includes: in training sample set without mark
Training sample is signed, which is generated by MixUp guess data amplification method;This is mixed without mark
Label training sample and corresponding there are label training sample, augmentation training sample set.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, which includes: one or more processing
Device;Storage device is stored thereon with one or more programs;When one or more programs are executed by one or more processors,
So that one or more processors realize the method as described in implementation any in first aspect.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, should
The method as described in implementation any in first aspect is realized when computer program is executed by processor.
The method and apparatus of the quality of rubbish for identification provided by the embodiments of the present application, obtain the figure of rubbish to be identified first
As sequence;Then image in the image sequence of rubbish to be identified is input to deep learning model trained in advance, to obtain
The recognition result of rubbish to be identified;Finally the recognition result of rubbish to be identified is analyzed and counted, to generate rubbish to be identified
Quality results improved using goal-selling object present in deep learning model automatic identification rubbish to rubbish quality
Recognition efficiency.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is that this application can be applied to exemplary system architectures therein;
Fig. 2 is the flow chart according to one embodiment of the method for the quality of rubbish for identification of the application;
Fig. 3 is the flow chart according to another embodiment of the method for the quality of rubbish for identification of the application;
Fig. 4 is the flow chart according to another embodiment of the method for the quality of rubbish for identification of the application;
Fig. 5 is the flow chart according to the further embodiment of the method for the quality of rubbish for identification of the application;
Fig. 6 is the structural schematic diagram according to one embodiment of the device of the quality of rubbish for identification of the application;
Fig. 7 is adapted for the structural schematic diagram for the computer system for realizing the electronic equipment of the embodiment of the present application.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can be using the method or the dress of rubbish quality for identification of the quality of rubbish for identification of the application
The exemplary system architecture 100 for the embodiment set.
As shown in Figure 1, may include picture pick-up device 101, network 102 and server 103 in system architecture 100.Network 102
To provide the medium of communication link between picture pick-up device 101 and server 103.Network 102 may include various connection classes
Type, such as wired, wireless communication link or fiber optic cables etc..
Picture pick-up device 101 can be interacted by network 102 with server 103, to receive or send message etc..Picture pick-up device
101 can be hardware, be also possible to software.When picture pick-up device 101 is hardware, it can be and support image or video capture function
Various electronic equipments.Including but not limited to camera, camera, smart phone and tablet computer etc..When picture pick-up device 101
When for software, it may be mounted in above-mentioned electronic equipment.Multiple softwares or software module may be implemented into it, also may be implemented into
Single software or software module.It is not specifically limited herein.
Server 103 can provide various services.Such as server 103 can be to the image of the rubbish to be identified got
The data such as sequence carry out the processing such as analyzing, and generate processing result (such as quality results of rubbish to be identified).
It should be noted that server 103 can be hardware, it is also possible to software.It, can when server 103 is hardware
To be implemented as the distributed server cluster that multiple servers form, individual server also may be implemented into.When server 103 is
When software, multiple softwares or software module (such as providing Distributed Services) may be implemented into, also may be implemented into single
Software or software module.It is not specifically limited herein.
It should be noted that provided by the embodiment of the present application for identification the method for rubbish quality generally by server
103 execute, and correspondingly, the device of rubbish quality is generally positioned in server 103 for identification.
It should be understood that the number of picture pick-up device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of picture pick-up device, network and server.
With continued reference to Fig. 2, it illustrates according to one embodiment of the method for the quality of rubbish for identification of the application
Process 200.The method of the rubbish quality for identification, comprising the following steps:
Step 201, the image sequence of rubbish to be identified is obtained.
In the present embodiment, the executing subject (such as server 103 shown in FIG. 1) of the method for rubbish quality for identification
The image sequence of rubbish to be identified can be obtained from the capture apparatus (such as capture apparatus 101 shown in FIG. 1) for communicating with connection
Column.In general, capture apparatus can support image or video capture function.In this way, capture apparatus can be by rubbish to be identified
A continuous shooting is carried out, to obtain the image sequence of rubbish to be identified.In addition, capture apparatus can also be by shooting rubbish to be identified
Video, and from video choose multi-frame video frame, to generate the image sequence of rubbish to be identified.
In some optional implementations of the present embodiment, the image sequence of rubbish to be identified be can be to rubbish to be identified
Rubbish topple over or shipment is acquired multiple obtained images.In this way, the image sequence record of rubbish to be identified is just
It is the dynamic mobile process of rubbish to be identified.
Step 202, using training in advance, corresponding with the classification of rubbish to be identified deep learning model to rubbish to be identified
Image in the image sequence of rubbish is identified, the recognition result of rubbish to be identified is obtained.
In the present embodiment, above-mentioned executing subject can use training in advance, corresponding with the classification of rubbish to be identified
Deep learning model identifies the image in the image sequence of rubbish to be identified, to obtain the identification knot of rubbish to be identified
Fruit.
In some optional implementations of the present embodiment, above-mentioned executing subject can be by the image sequence of rubbish to be identified
Image in column is directly inputted into training in advance, corresponding with the classification of rubbish to be identified deep learning model, exports wait know
The recognition result of other rubbish.
In some optional implementations of the present embodiment, above-mentioned executing subject can also utilize image processing method pair
Image in the image sequence of rubbish to be identified is pre-processed.Then, above-mentioned executing subject can be by pretreated wait know
Image in the image sequence of other rubbish is input to training in advance, corresponding with the classification of rubbish to be identified deep learning mould
Type exports the recognition result of rubbish to be identified.
Here, deep learning model can be used for identifying goal-selling object present in rubbish, characterize the image sequence of rubbish
Corresponding relationship between column and the recognition result of rubbish.In general, the recognition result of rubbish to be identified may include rubbish to be identified
Present in goal-selling object information.The information of goal-selling object can include but is not limited to pre- present in rubbish to be identified
If object, the position of goal-selling object, the classification of goal-selling object, the quantity of goal-selling object, goal-selling object area
With the content of goal-selling object etc..
In general, the house refuse of resident can be divided into wet refuse and dry rubbish.Wet refuse collect after using microorganism into
Row compost, Anaerobic Digestion prepare bio-fuel.And polybag, the modeling of non-degradable or degradation time length in wet refuse
Material bottle, pop can etc. can seriously affect the quality of wet refuse.Therefore, for wet refuse, non-degradable or degradation time are long
Polybag, plastic bottle, pop can etc. can be goal-selling object.The corresponding deep learning model of wet refuse can be used for identifying
The goal-sellings object such as non-degradable present in wet refuse or polybag, plastic bottle, the pop can of degradation time length.Dry rubbish is received
Available substance is selected by staff after collection, remaining rubbish carries out landfill and incineration disposal.And in dry rubbish
Water flow etc. can seriously affect the quality of dry rubbish.Therefore, for dry rubbish, water flow etc. can be goal-selling object.It is dry
The corresponding deep learning model of rubbish can be used for identifying the goal-sellings object such as water flow present in dry rubbish.
In the present embodiment, deep learning model can be using machine learning method and training sample to existing machine
Learning model carries out obtained from Training.In general, above-mentioned executing subject can select feature extraction network, sorter network
Model training engine training deep learning model is used with Recurrent networks.Wherein, feature extraction network can be used for extracting rubbish
Present in goal-selling object feature.Sorter network can be used for identifying the classification of goal-selling object present in rubbish.It returns
Network is returned to can be used for detecting the position of goal-selling object present in rubbish.
In some optional implementations of the present embodiment, deep learning model can be trained as follows
It arrives:
Firstly, obtaining training sample set.
Here, the training sample in training sample set may include sample rubbish image and sample rubbish mark image.
Sample rubbish mark image, which can be, is marked obtained image to goal-selling object present in sample rubbish image.
In general, to the corresponding deep learning model of training wet refuse, sample rubbish image can be to wet refuse into
Row shoots obtained image, and sample rubbish mark image can be to default mesh such as water flows present in sample rubbish image
Obtained image is marked in mark object.To the corresponding deep learning model of the dry rubbish of training, sample rubbish image can be with
It is to carry out shooting obtained image to dry rubbish, sample rubbish mark image can be to present in sample rubbish image
Obtained image is marked in the goal-sellings object such as non-degradable or polybag, plastic bottle, the pop can of degradation time length.
Then, for the training sample in training sample set, using the sample rubbish image in the training sample as defeated
Enter, using the sample rubbish mark image in the training sample as output, training obtains deep learning model.
Here, above-mentioned executing subject can initialize deep learning model first, by the parameter setting of deep learning model
For initial value;Then deep learning model is trained using training sample set.In the training process, deep learning model
Parameter can constantly be adjusted, until the recognition effect of deep learning model meets preset constraint condition.
In practice, the flow shape in wet refuse changes multiplicity.Therefore, in training sample set without label training
Sample, above-mentioned executing subject can pass through MixUp guess data amplification side first using semi-supervised learnings methods such as MixMatch
Method generates the low entropy label without label training sample;Then this is mixed without label training sample and corresponding has label training sample
This, augmentation training sample set.In this way, being greatly lowered dependence of the algorithm to a large amount of mark training sample set.
Step 203, the recognition result of rubbish to be identified is analyzed and counted, generates the quality results of rubbish to be identified.
In the present embodiment, above-mentioned executing subject can analyze and count the recognition result of rubbish to be identified, with life
At the quality results of rubbish to be identified.In general, above-mentioned executing subject can count goal-selling object present in rubbish to be identified
Content, and determine according to the content of goal-selling object the quality of rubbish to be identified.In general, the content of goal-selling object is got over
The quality of height, rubbish to be identified is lower;Conversely, the quality of rubbish to be identified is higher.
It is connect in addition, the quality results of rubbish to be identified can also be sent to rubbish quality display platform by above-mentioned executing subject
By manual examination and verification.After manual examination and verification, corresponding processing method can be taken to rubbish to be identified according to audit assessment result
It is handled.Meanwhile garbage disposal management platform can also record the image sequence, recognition result, quality knot of rubbish to be identified
At least one of in fruit, audit assessment result and processing method.
In practice, the corresponding processing method of the rubbish of different qualities is different.For example, if rubbish to be identified is that quality is higher
Wet refuse, it will usually compost directly be carried out to rubbish to be identified using microorganism.In another example if rubbish to be identified is that quality is lower
Wet refuse, it will usually the long polybag of sort out non-degradable or degradation time, plastic bottle, easily first from rubbish to be identified
Tank etc. is drawn, compost then is carried out to remaining wet refuse using microorganism.
The method of the quality of rubbish for identification provided by the embodiments of the present application, obtains the image sequence of rubbish to be identified first
Column;Then image in the image sequence of rubbish to be identified is input to deep learning model trained in advance, to obtain wait know
The recognition result of other rubbish;Finally the recognition result of rubbish to be identified is analyzed and counted, to generate the product of rubbish to be identified
Matter result.Using goal-selling object present in deep learning model automatic identification rubbish, the identification to rubbish quality is improved
Efficiency.
With further reference to Fig. 3, it illustrates another implementations according to the method for the quality of rubbish for identification of the application
The process 300 of example.The method of the rubbish quality for identification, comprising the following steps:
Step 301, the image sequence of rubbish to be identified is obtained.
In the present embodiment, the executing subject (such as server 103 shown in FIG. 1) of the method for rubbish quality for identification
The image sequence of rubbish to be identified can be obtained from the capture apparatus (such as capture apparatus 101 shown in FIG. 1) for communicating with connection
Column.Wherein, the image sequence of rubbish to be identified can be toppling over or obtained by shipment is acquired to rubbish to be identified
Multiple images.In this way, the image sequence of rubbish to be identified record be exactly rubbish to be identified dynamic mobile process.
Step 302, sky background image is sorted out from the image sequence of rubbish to be identified.
In the present embodiment, above-mentioned executing subject can sort out sky Background from the image sequence of rubbish to be identified
Picture.
In general, picture pick-up device can begin to acquisition image before rubbish to be identified is toppled over or is shipped.In this way, picture pick-up device is adopted
It just will include in the image sequence collected there is no the image of rubbish to be identified, and such image is exactly empty background image.This
In, ALOCC (Adversarially Learned One-Class Classifier for can be used in above-mentioned executing subject
Novelty Detection, the confrontation Study strategies and methods for single class abnormality detection)-CVPR2018 (Computer Vision
2018,2018 computer vision of and Pattern Recognition and pattern-recognition meeting) etc. unsupervised one-class
(one kind) sorting algorithm distinguishes the empty background image in the image sequence of rubbish to be identified.
Step 303, for the image in the image sequence of rubbish to be identified in addition to empty background image, by empty background image
As the background of the image, the region of rubbish to be identified in the images is detected using Moving target detection algorithm.
In the present embodiment, above-mentioned to hold for the image in the image sequence of rubbish to be identified in addition to empty background image
Row main body can detect rubbish to be identified at this using Moving target detection algorithm using empty background image as the background of the image
Region in image.Here, Moving target detection algorithm can include but is not limited at least one of following: frame differential method, three
Frame difference method, background subtraction and optical flow method.
Step 304, the pixel value in the image in addition to detected region is arranged to preset value.
In the present embodiment, above-mentioned executing subject can set the pixel value in the image in addition to detected region
It is set to preset value.For example, the pixel value in the image in addition to detected region can be arranged to by above-mentioned executing subject
0, to eliminate influence of the background in the image to subsequent identification.
Step 305, using training in advance, corresponding with the classification of rubbish to be identified deep learning model to treated
Image in the image sequence of rubbish to be identified is identified, the recognition result of rubbish to be identified is obtained.
Step 306, the recognition result of rubbish to be identified is analyzed and counted, generates the quality results of rubbish to be identified.
In the present embodiment, the concrete operations of step 305-306 are in the embodiment shown in Figure 2 in step 202-203
It is described in detail, details are not described herein.
From figure 3, it can be seen that compared with the corresponding embodiment of Fig. 2, the quality of rubbish for identification in the present embodiment
The process 300 of method increases image preprocessing step.The scheme of the present embodiment description eliminates the background pair in image as a result,
The influence of subsequent identification, the region occurred just for rubbish carry out image recognition, and it is changeable to solve scene in rubbish identification scene
Property caused by the problem of misidentifying, realize deep learning model to rubbish identification scene complicated and changeable without helping to change, further
Improve the accuracy of model recognition result.
With further reference to Fig. 4, it illustrates another implementations according to the method for the quality of rubbish for identification of the application
The process 400 of example.The method of the rubbish quality for identification, comprising the following steps:
Step 401, the image sequence of rubbish to be identified is obtained.
In the present embodiment, the executing subject (such as server 103 shown in FIG. 1) of the method for rubbish quality for identification
The image sequence of rubbish to be identified can be obtained from the capture apparatus (such as capture apparatus 101 shown in FIG. 1) for communicating with connection
Column.Wherein, the image sequence of rubbish to be identified can be toppling over or obtained by shipment is acquired to rubbish to be identified
Multiple images.In this way, the image sequence of rubbish to be identified record be exactly rubbish to be identified dynamic mobile process.
Step 402, for the image in the image sequence of rubbish to be identified, by default screenshot box according to preset step-length along pre-
If moving direction moves on this image, multiple subgraphs are intercepted.
In the present embodiment, for the image in the image sequence of rubbish to be identified, above-mentioned executing subject can will be preset
Screenshot box is moved according to preset step-length along moving direction is preset on this image, to intercept multiple subgraphs.For example, above-mentioned execution
Main body can move on the image from left to right, from top to bottom according to preset step-length, to intercept multiple subgraphs.Wherein, it presets
Screenshot box can be a fixed-size box.
Step 403, for the subgraph in multiple subgraphs, the isolated of the subgraph is eliminated using median filter method and is made an uproar
Sound point.
In the present embodiment, for the subgraph in multiple subgraphs, above-mentioned executing subject can use median filtering side
Method eliminates the isolated noise point of the subgraph, to eliminate influence of the noise in the subgraph to subsequent identification.
Step 404, using training in advance, corresponding with the classification of rubbish to be identified deep learning model to treated
Image in the image sequence of rubbish to be identified is identified, the recognition result of rubbish to be identified is obtained.
Step 405, the recognition result of rubbish to be identified is analyzed and counted, generates the quality results of rubbish to be identified.
In the present embodiment, the concrete operations of step 404-405 are in the embodiment shown in Figure 2 in step 202-203
It is described in detail, details are not described herein.
Figure 4, it is seen that compared with the corresponding embodiment of Fig. 2, the quality of rubbish for identification in the present embodiment
The process 400 of method increases image preprocessing step.The scheme of the present embodiment description is by dividing the image into subgraph as a result,
Picture, and goal-selling object is identified using semantic segmentation algorithm, solve the problems, such as that the goal-selling object in image is seriously blocked,
Further improve the accuracy of model recognition result.
With further reference to Fig. 5, it illustrates another implementations according to the method for the quality of rubbish for identification of the application
The process 500 of example.The method of the rubbish quality for identification, comprising the following steps:
Step 501, the image sequence of rubbish to be identified is obtained.
In the present embodiment, the executing subject (such as server 103 shown in FIG. 1) of the method for rubbish quality for identification
The image sequence of rubbish to be identified can be obtained from the capture apparatus (such as capture apparatus 101 shown in FIG. 1) for communicating with connection
Column.Wherein, the image sequence of rubbish to be identified can be toppling over or obtained by shipment is acquired to rubbish to be identified
Multiple images.In this way, the image sequence of rubbish to be identified record be exactly rubbish to be identified dynamic mobile process.
Step 502, using training in advance, corresponding with the classification of rubbish to be identified deep learning model to rubbish to be identified
Image in the image sequence of rubbish is identified, the recognition result of rubbish to be identified is obtained.
In the present embodiment, the concrete operations of step 502 have carried out in step 202 in detail in the embodiment shown in Figure 2
Thin introduction, details are not described herein.
Step 503, for the image in the image sequence of rubbish to be identified, if there are goal-selling object, benefits in the image
The goal-selling object in the image is determined with target circle, as marking target.
In the present embodiment, for the image in the image sequence of rubbish to be identified, if there are goal-sellings in the image
Object, above-mentioned executing subject can use target circle and determine goal-selling object in the image, as marking target.Wherein, institute
Stating target frame can be the minimum box for surrounding goal-selling object.
Step 504, the first default mesh image and rear the before choosing the image in the image sequence of rubbish to be identified
Two preset numbers image.
In the present embodiment, above-mentioned executing subject can be chosen from the image sequence of rubbish to be identified before the image
One default mesh image and rear second preset number image.For example, above-mentioned executing subject can be from the image of rubbish to be identified
Preceding 3 images and rear 3 images of the image are chosen in sequence.
Step 505, target following is carried out in the image of selected taking-up based on marking target.
In the present embodiment, above-mentioned executing subject can carry out target based on marking target in the image of selected taking-up
Tracking, to obtain tracking result.Here, target tracking algorism can include but is not limited to CACF (Context-Aware
Correlation Filter Tracking, context-aware correlation filtering tracking), KCF (2015PAMI, high speed nuclear phase close filter
Wave tracking), SiameseFC (Fully-Convolutional Siamese Networks for Object Tracking, base
In the target following for connecting twin network entirely), C-COT (Beyond Correlation Filters:Learning
Continuous Convolution Operators for Visual Tracking surmounts correlation filtering: the continuous volume of study
Integrating), HCF (Hierarchical Convolutional Features for Visual Tracking, be layered convolution
The vision of feature tracks) etc..Wherein, tracking result may include tracing into marking target in the image of selected taking-up
The information of image.
Step 506, the recognition result of rubbish to be identified is updated based on tracking result.
In the present embodiment, above-mentioned executing subject can update the recognition result to rubbish to be identified based on tracking result.
In general, above-mentioned executing subject can count the number for tracing into the image of marking target in the selected image taken out;If tracking
Number to the image of marking target is more than third preset number (such as 2), in the recognition result for retaining rubbish to be identified
Marking target information;If the number for tracing into the image of marking target is no more than third preset number, delete wait know
The information of marking target in the recognition result of other rubbish.
Step 507, the recognition result of rubbish to be identified is analyzed and counted, generates the quality results of rubbish to be identified.
In the present embodiment, the concrete operations of step 507 have carried out in step 203 in detail in the embodiment shown in Figure 2
Thin introduction, details are not described herein.
From figure 5 it can be seen that compared with the corresponding embodiment of Fig. 2, the quality of rubbish for identification in the present embodiment
The process 500 of method increases target tracking step.The scheme of the present embodiment description is updated based on tracking result wait know as a result,
The recognition result of other rubbish, image brightness variation caused by solving video capture in the process and weather illumination power cause
Image brightness variation the problem of leading to missing inspection and object topple over or shipment in the problem of slightly blocking,
Missing inspection and erroneous detection are reduced to a certain extent, improve the recall rate and accuracy of model recognition result.
With further reference to Fig. 6, as the realization to method shown in above-mentioned each figure, this application provides a kind of rubbish for identification
One embodiment of the device of rubbish quality, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which specifically may be used
To be applied in various electronic equipments.
As shown in fig. 6, the device 600 of the quality of rubbish for identification of the present embodiment may include: acquiring unit 601, know
Other unit 602 and statistic unit 603.Wherein, acquiring unit 601 are configured to obtain the image sequence of rubbish to be identified;Identification
Unit 602 is configured to deep learning model using training in advance, corresponding with the classification of rubbish to be identified to rubbish to be identified
Image in the image sequence of rubbish is identified, the recognition result of rubbish to be identified is obtained, wherein the identification knot of rubbish to be identified
Fruit includes the information of goal-selling object present in rubbish to be identified;Statistic unit 603 is configured to the knowledge to rubbish to be identified
Other result analyzes and counts, and generates the quality results of rubbish to be identified.
In the present embodiment, for identification in the device 600 of rubbish quality: acquiring unit 601, recognition unit 602 and system
The specific processing and its brought technical effect for counting unit 603 can be respectively with reference to step 201, the steps in Fig. 2 corresponding embodiment
Rapid 202 and step 203 related description, details are not described herein.
In some optional implementations of the present embodiment, deep learning model includes feature extraction network, classification net
Network and Recurrent networks.
In some optional implementations of the present embodiment, the image sequence of rubbish to be identified is to rubbish to be identified
Topple over or shipment is acquired multiple obtained images.
In some optional implementations of the present embodiment, the device 600 of rubbish quality for identification further include: processing
Unit (not shown) is configured to carry out the image in the image sequence of rubbish to be identified using image processing method pre-
Processing.
In some optional implementations of the present embodiment, processing unit is further configured to: from rubbish to be identified
Image sequence in sort out sky background image;For the figure in the image sequence of rubbish to be identified in addition to empty background image
Picture detects rubbish to be identified in the images using Moving target detection algorithm using empty background image as the background of the image
Region, and the pixel value in the image in addition to detected region is arranged to preset value.
In some optional implementations of the present embodiment, processing unit is further configured to: for rubbish to be identified
Default screenshot box is moved according to preset step-length along moving direction is preset on this image, is cut by the image in the image sequence of rubbish
Take multiple subgraphs;For the subgraph in multiple subgraphs, the isolated noise of the subgraph is eliminated using median filter method
Point.
In some optional implementations of the present embodiment, the device 600 of rubbish quality for identification further include: label
Unit (not shown), the image being configured in the image sequence for rubbish to be identified, if existing in the image default
Object determines the goal-selling object in the image using target circle, as marking target;Selection unit (is not shown in figure
Out), it is configured to the first default mesh before choosing the image in the image sequence of rubbish to be identified and opens image and rear second in advance
If number image;Tracking cell (not shown), be configured to based on marking target in the image of selected taking-up into
Row target following;Updating unit (not shown) is configured to update the recognition result of rubbish to be identified based on tracking result.
In some optional implementations of the present embodiment, updating unit is further configured to: statistics is selected to be taken out
Image in trace into marking target image number;If the number for tracing into the image of marking target is more than that third is pre-
If number, retain the information of the marking target in the recognition result of rubbish to be identified;If tracing into the image of marking target
Number be no more than third preset number, delete the information of the marking target in the recognition result of rubbish to be identified.
In some optional implementations of the present embodiment, training obtains deep learning model as follows: obtaining
Take training sample set, wherein the training sample in training sample set includes sample rubbish image and sample rubbish mark figure
Picture, it is that obtained image is marked to goal-selling object present in sample rubbish image that sample rubbish, which marks image,;It is right
Training sample in training sample set, using the sample rubbish image in the training sample as input, by the training sample
In sample rubbish mark image as output, training obtain deep learning model.
In some optional implementations of the present embodiment, after obtaining training sample set, further includes: for instruction
Practice sample set in without label training sample, this is generated without label training sample by MixUp guess data amplification method
Low entropy label;This is mixed without label training sample and corresponding has label training sample, augmentation training sample set.
Below with reference to Fig. 7, it is (such as shown in FIG. 1 that it illustrates the electronic equipments for being suitable for being used to realize the embodiment of the present application
Server 103) computer system 700 structural schematic diagram.Electronic equipment shown in Fig. 7 is only an example, should not be right
The function and use scope of the embodiment of the present application bring any restrictions.
As shown in fig. 7, computer system 700 includes central processing unit (CPU) 701, it can be read-only according to being stored in
Program in memory (ROM) 702 or be loaded into the program in random access storage device (RAM) 703 from storage section 708 and
Execute various movements appropriate and processing.In RAM 703, also it is stored with system 700 and operates required various programs and data.
CPU 701, ROM 702 and RAM 703 are connected with each other by bus 704.Input/output (I/O) interface 705 is also connected to always
Line 704.
I/O interface 705 is connected to lower component: the importation 706 including keyboard, mouse etc.;It is penetrated including such as cathode
The output par, c 707 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section 708 including hard disk etc.;
And the communications portion 709 of the network interface card including LAN card, modem etc..Communications portion 709 via such as because
The network of spy's net executes communication process.Driver 710 is also connected to I/O interface 705 as needed.Detachable media 711, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 710, in order to read from thereon
Computer program be mounted into storage section 708 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 709, and/or from detachable media
711 are mounted.When the computer program is executed by central processing unit (CPU) 701, limited in execution the present processes
Above-mentioned function.
It should be noted that computer-readable medium described herein can be computer-readable signal media or meter
Calculation machine readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but not
Be limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or any above combination.Meter
The more specific example of calculation machine readable storage medium storing program for executing can include but is not limited to: have the electrical connection, just of one or more conducting wires
Taking formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In this application, computer readable storage medium can be it is any include or storage journey
The tangible medium of sequence, the program can be commanded execution system, device or device use or in connection.And at this
In application, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.Include on computer-readable medium
Program code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc. are above-mentioned
Any appropriate combination.
The calculating of the operation for executing the application can be write with one or more programming languages or combinations thereof
Machine program code, described program design language include object-oriented programming language-such as Java, Smalltalk, C+
+, further include conventional procedural programming language-such as " C " language or similar programming language.Program code can
Fully to execute, partly execute on the user computer on the user computer, be executed as an independent software package,
Part executes on the remote computer or holds on remote computer or electronic equipment completely on the user computer for part
Row.In situations involving remote computers, remote computer can pass through the network of any kind --- including local area network
(LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as utilize internet
Service provider is connected by internet).
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor packet
Include acquiring unit, recognition unit and statistic unit.Wherein, the title of these units is not constituted in this case to the unit sheet
The restriction of body, for example, acquiring unit is also described as " obtaining the unit of the image sequence of rubbish to be identified ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in electronic equipment described in above-described embodiment;It is also possible to individualism, and without in the supplying electronic equipment.
Above-mentioned computer-readable medium carries one or more program, when said one or multiple programs are held by the electronic equipment
When row, so that the electronic equipment: obtaining the image sequence of rubbish to be identified;Utilize training in advance and rubbish to be identified class
Not corresponding deep learning model identifies the image in the image sequence of rubbish to be identified, obtains the knowledge of rubbish to be identified
Other result, wherein the recognition result of rubbish to be identified includes the information of goal-selling object present in rubbish to be identified;Treat knowledge
The recognition result of other rubbish analyzes and counts, and generates the quality results of rubbish to be identified.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (13)
1. a kind of method of rubbish quality for identification, comprising:
Obtain the image sequence of rubbish to be identified;
Using training in advance, corresponding with the classification of the rubbish to be identified deep learning model to the rubbish to be identified
Image in image sequence is identified, the recognition result of the rubbish to be identified is obtained, wherein the knowledge of the rubbish to be identified
Other result includes the information of goal-selling object present in the rubbish to be identified;
The recognition result of the rubbish to be identified is analyzed and counted, the quality results of the rubbish to be identified are generated.
2. according to the method described in claim 1, wherein, the deep learning model includes feature extraction network, sorter network
And Recurrent networks.
3. according to the method described in claim 1, wherein, the image sequence of the rubbish to be identified is to the rubbish to be identified
Topple over or shipment is acquired multiple obtained images.
4. according to the method described in claim 3, wherein, utilizing training in advance and the rubbish to be identified class described
Before not corresponding deep learning model identifies the image in the image sequence of the rubbish to be identified, further includes:
The image in the image sequence of the rubbish to be identified is pre-processed using image processing method.
5. according to the method described in claim 4, wherein, it is described using image processing method to the image of the rubbish to be identified
Image in sequence is pre-processed, comprising:
Sky background image is sorted out from the image sequence of the rubbish to be identified;
For the image in the image sequence of the rubbish to be identified in addition to the empty background image, by the empty background image
As the background of the image, the region of the rubbish to be identified in the images is detected using Moving target detection algorithm, and
Pixel value in the image in addition to detected region is arranged to preset value.
6. according to the method described in claim 4, wherein, it is described using image processing method to the image of the rubbish to be identified
Image in sequence is pre-processed, comprising:
For the image in the image sequence of the rubbish to be identified, by default screenshot box according to preset step-length along the default side of movement
To moving on this image, multiple subgraphs are intercepted;
For the subgraph in multiple described subgraphs, the isolated noise point of the subgraph is eliminated using median filter method.
7. according to the method described in claim 3, wherein, utilizing training in advance and the rubbish to be identified class described
Not corresponding deep learning model identifies the image in the image sequence of the rubbish to be identified, obtains described to be identified
After the recognition result of rubbish, further includes:
Target frame is utilized if there are goal-selling objects in the image for the image in the image sequence of the rubbish to be identified
The goal-selling object in the image is confined, as marking target;
First default mesh image and rear second present count before choosing the image in the image sequence of the rubbish to be identified
Mesh image;
Target following is carried out in the image of selected taking-up based on the marking target;
The recognition result of the rubbish to be identified is updated based on tracking result.
8. according to the method described in claim 7, wherein, the identification knot that the rubbish to be identified is updated based on tracking result
Fruit, comprising:
The number of the image of the marking target is traced into the selected image taken out of statistics;
If the number for tracing into the image of the marking target is more than third preset number, retain the knowledge of the rubbish to be identified
The information of the marking target in other result;
If the number for tracing into the image of the marking target is no more than the third preset number, the rubbish to be identified is deleted
The information of the marking target in the recognition result of rubbish.
9. method described in one of -8 according to claim 1, wherein the deep learning model is trained as follows
It arrives:
Obtain training sample set, wherein the training sample in the training sample set includes sample rubbish image and sample
Rubbish marks image, and sample rubbish mark image is obtained by goal-selling object present in sample rubbish image is marked
Image;
It will using the sample rubbish image in the training sample as input for the training sample in the training sample set
For sample rubbish mark image in the training sample as output, training obtains the deep learning model.
10. according to the method described in claim 9, wherein, after the acquisition training sample set, further includes:
For, without label training sample, generating the nothing by MixUp guess data amplification method in the training sample set
The low entropy label of label training sample;
Mix this without label training sample and it is corresponding have label training sample, training sample set described in augmentation.
11. a kind of device of rubbish quality for identification, comprising:
Acquiring unit is configured to obtain the image sequence of rubbish to be identified;
Recognition unit is configured to deep learning model using training in advance, corresponding with the classification of the rubbish to be identified
Image in the image sequence of the rubbish to be identified is identified, the recognition result of the rubbish to be identified is obtained, wherein
The recognition result of the rubbish to be identified includes the information of goal-selling object present in the rubbish to be identified;
Statistic unit is configured to analyze and count the recognition result of the rubbish to be identified, generates the rubbish to be identified
The quality results of rubbish.
12. a kind of electronic equipment, comprising:
One or more processors;
Storage device is stored thereon with one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now method as described in any in claim 1-10.
13. a kind of computer-readable medium, is stored thereon with computer program, wherein the computer program is held by processor
The method as described in any in claim 1-10 is realized when row.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910547978.2A CN110276300B (en) | 2019-06-24 | 2019-06-24 | Method and device for identifying quality of garbage |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910547978.2A CN110276300B (en) | 2019-06-24 | 2019-06-24 | Method and device for identifying quality of garbage |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110276300A true CN110276300A (en) | 2019-09-24 |
CN110276300B CN110276300B (en) | 2021-12-28 |
Family
ID=67961724
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910547978.2A Active CN110276300B (en) | 2019-06-24 | 2019-06-24 | Method and device for identifying quality of garbage |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110276300B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111017429A (en) * | 2019-11-20 | 2020-04-17 | 重庆特斯联智慧科技股份有限公司 | Community garbage classification method and system based on multi-factor fusion |
CN111753661A (en) * | 2020-05-25 | 2020-10-09 | 济南浪潮高新科技投资发展有限公司 | Target identification method, device and medium based on neural network |
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 |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104036490A (en) * | 2014-05-13 | 2014-09-10 | 重庆大学 | Foreground segmentation method applied to mobile communication network transmission |
CN106000904A (en) * | 2016-05-26 | 2016-10-12 | 北京新长征天高智机科技有限公司 | Automatic sorting system for household refuse |
CN106423913A (en) * | 2016-09-09 | 2017-02-22 | 华侨大学 | Construction waste sorting method and system |
CN106494789A (en) * | 2016-11-14 | 2017-03-15 | 上海理工大学 | Refuse classification statistic device, equipment and system |
CN107054936A (en) * | 2017-03-23 | 2017-08-18 | 广东数相智能科技有限公司 | A kind of refuse classification prompting dustbin and system based on image recognition |
CN108639601A (en) * | 2018-05-18 | 2018-10-12 | 赵欣 | A kind of intelligent garbage classification storage device |
CN108861183A (en) * | 2018-03-26 | 2018-11-23 | 厦门快商通信息技术有限公司 | A kind of intelligent garbage classification method based on machine learning |
CN109201514A (en) * | 2017-06-30 | 2019-01-15 | 京东方科技集团股份有限公司 | Waste sorting recycle method, garbage classification device and classified-refuse recovery system |
CN109472200A (en) * | 2018-09-29 | 2019-03-15 | 深圳市锦润防务科技有限公司 | A kind of intelligent sea rubbish detection method, system and storage medium |
-
2019
- 2019-06-24 CN CN201910547978.2A patent/CN110276300B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104036490A (en) * | 2014-05-13 | 2014-09-10 | 重庆大学 | Foreground segmentation method applied to mobile communication network transmission |
CN106000904A (en) * | 2016-05-26 | 2016-10-12 | 北京新长征天高智机科技有限公司 | Automatic sorting system for household refuse |
CN106423913A (en) * | 2016-09-09 | 2017-02-22 | 华侨大学 | Construction waste sorting method and system |
CN106494789A (en) * | 2016-11-14 | 2017-03-15 | 上海理工大学 | Refuse classification statistic device, equipment and system |
CN107054936A (en) * | 2017-03-23 | 2017-08-18 | 广东数相智能科技有限公司 | A kind of refuse classification prompting dustbin and system based on image recognition |
CN109201514A (en) * | 2017-06-30 | 2019-01-15 | 京东方科技集团股份有限公司 | Waste sorting recycle method, garbage classification device and classified-refuse recovery system |
CN108861183A (en) * | 2018-03-26 | 2018-11-23 | 厦门快商通信息技术有限公司 | A kind of intelligent garbage classification method based on machine learning |
CN108639601A (en) * | 2018-05-18 | 2018-10-12 | 赵欣 | A kind of intelligent garbage classification storage device |
CN109472200A (en) * | 2018-09-29 | 2019-03-15 | 深圳市锦润防务科技有限公司 | A kind of intelligent sea rubbish detection method, system and storage medium |
Non-Patent Citations (1)
Title |
---|
DAVID BERTHELOT等: "MixMatch: A Holistic Approach to Semi-Supervised Learning", 《ARXIV》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111017429A (en) * | 2019-11-20 | 2020-04-17 | 重庆特斯联智慧科技股份有限公司 | Community garbage classification method and system based on multi-factor fusion |
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 |
CN111753661A (en) * | 2020-05-25 | 2020-10-09 | 济南浪潮高新科技投资发展有限公司 | Target identification method, device and medium based on neural network |
CN111753661B (en) * | 2020-05-25 | 2022-07-12 | 山东浪潮科学研究院有限公司 | Target identification method, device and medium based on neural network |
Also Published As
Publication number | Publication date |
---|---|
CN110276300B (en) | 2021-12-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | An effective data augmentation strategy for CNN-based pest localization and recognition in the field | |
CN111178197B (en) | Mass R-CNN and Soft-NMS fusion based group-fed adherent pig example segmentation method | |
CN110276300A (en) | The method and apparatus of rubbish quality for identification | |
CN112926405A (en) | Method, system, equipment and storage medium for detecting wearing of safety helmet | |
CN107657175A (en) | A kind of homologous detection method of malice sample based on image feature descriptor | |
CN109145828A (en) | Method and apparatus for generating video classification detection model | |
CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
CN113838034B (en) | Quick detection method for surface defects of candy package based on machine vision | |
CN112070135A (en) | Power equipment image detection method and device, power equipment and storage medium | |
CN103632247A (en) | Meter information identification system and method based on intelligent storage system | |
CN110458193A (en) | Garbage classification processing method, apparatus and system, computer equipment and readable medium | |
CN112613438A (en) | Portable online citrus yield measuring instrument | |
CN109086827A (en) | Method and apparatus for detecting monocrystaline silicon solar cell defect | |
CN115239209B (en) | Method, equipment and system for screening heavy metal-containing humus soil in refuse landfill | |
CN113792578A (en) | Method, device and system for detecting abnormity of transformer substation | |
CN109002810A (en) | Model evaluation method, Radar Signal Recognition method and corresponding intrument | |
CN113850562A (en) | Intelligent side station supervision method and system | |
CN114170597A (en) | Algae detection equipment and detection method | |
CN202815869U (en) | Vehicle microcomputer image and video data extraction apparatus | |
Bian et al. | Medical waste classification system based on opencv and ssd-mobilenet for 5g | |
CN117115722A (en) | Construction scene detection method and device, storage medium and electronic equipment | |
CN112231391A (en) | Big data information analysis system based on cloud computing | |
Wang et al. | Automatic identification and location of tunnel lining cracks | |
CN117576632A (en) | Multi-mode AI large model-based power grid monitoring fire early warning system and method | |
CN112529836A (en) | High-voltage line defect detection method and device, storage medium and electronic equipment |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |