CN113435514A - Construction waste fine classification method and device based on meta-deep learning - Google Patents
Construction waste fine classification method and device based on meta-deep learning Download PDFInfo
- Publication number
- CN113435514A CN113435514A CN202110723467.9A CN202110723467A CN113435514A CN 113435514 A CN113435514 A CN 113435514A CN 202110723467 A CN202110723467 A CN 202110723467A CN 113435514 A CN113435514 A CN 113435514A
- Authority
- CN
- China
- Prior art keywords
- construction waste
- classification
- image
- meta
- deep learning
- 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
- 239000002699 waste material Substances 0.000 title claims abstract description 125
- 238000010276 construction Methods 0.000 title claims abstract description 124
- 238000013135 deep learning Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 21
- 230000003595 spectral effect Effects 0.000 claims abstract description 16
- 239000010813 municipal solid waste Substances 0.000 claims description 17
- 238000001914 filtration Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000002146 bilateral effect Effects 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 101100455978 Arabidopsis thaliana MAM1 gene Proteins 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 description 9
- 239000000463 material Substances 0.000 description 6
- 239000002910 solid waste Substances 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 239000004568 cement Substances 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 239000011449 brick Substances 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000004579 marble Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a construction waste fine classification method and device based on meta-deep learning, wherein the method comprises the following steps: constructing a first classification network, a second classification network and a third classification network which are respectively used for classifying the construction waste in the construction waste 2D image, the construction waste 3D image and the construction waste spectral image; acquiring images to construct a training set, and training the first classification network, the second classification network and the third classification network based on a meta-depth learning algorithm; collecting 2D images and 3D images of construction wastes, performing first-stage prediction by using a first classification network and a second classification network, outputting construction waste categories when the confidence coefficient of a prediction result of the first stage is greater than a set threshold, and otherwise, collecting spectral images of the construction wastes, inputting the spectral images of the construction wastes into a third classification network, performing second-stage prediction and outputting the construction waste categories. Compared with the prior art, the method has the advantage of accurate classification result.
Description
Technical Field
The invention relates to the technical field of building garbage classification, in particular to a building garbage fine classification method and device based on meta-deep learning.
Background
With the continuous improvement of the consciousness of garbage classification and the high-speed development of artificial intelligence, the garbage classification by adopting the artificial intelligence technology becomes an accurate and convenient mode.
Due to the fact that the scene of a construction site is complex, and different construction wastes are difficult to distinguish. For example, bricks and cement have similar colors and textures, and deep learning networks have difficulty in well classifying the bricks and the cement. In addition, the intelligent manipulator needs to pick up construction wastes of different categories according to the obtained waste classification information. Therefore, the requirements on the classification accuracy and the real-time performance of the construction waste are high.
A paper A combination model based on transfer learning for waste classification in 2019 proposes a method based on a transfer learning network to solve the problem of garbage classification. The pre-trained VGG19, DenseNet169 and NASNNetLarge models are adopted as base classifiers in the network, and the classifier with the highest training accuracy is finally selected. However, this approach is through pre-training the images in the public dataset, and the classes of images in the test set are also similar to the classes of the dataset, assuming that the garbage classes in the actual scene are consistent or similar to the classes in the public dataset. However, for the construction site field, images of cement, stone, woven bags, etc. cannot be found in public data sets, i.e. effective pre-training cannot be performed.
Chinese patent CN 109629795A: an intelligent building garbage classification device is provided with a multilayer particle size separation channel. Considering that different solid waste materials are different in size, the solid waste materials can be divided into different piles through the electric baffle and the corner motor, and therefore classification of the solid waste materials is achieved. However, the classification device disclosed in this patent classifies the construction waste according to the size thereof, and only the size information of the objects is considered, and different objects with the same size cannot be effectively distinguished. The construction waste classification method needs to classify the solid waste of different materials, which requires that the characteristics of different construction wastes, not only the object size information, must be captured.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a construction waste fine classification method and device based on meta-deep learning.
The purpose of the invention can be realized by the following technical scheme:
a construction waste fine classification method based on meta-deep learning comprises the following steps:
building a construction waste classification model, wherein the construction waste classification model comprises a first classification network for classifying construction waste in a construction waste 2D image, a second classification network for classifying construction waste in a construction waste 3D image and a third classification network for classifying construction waste in a construction waste spectral image;
collecting a construction waste 2D image, a construction waste 3D image and a construction waste spectral image to construct a training set, and training the first classification network, the second classification network and the third classification network based on a meta-deep learning algorithm;
collecting 2D images and 3D images of construction wastes, performing first-stage prediction by using a first classification network and a second classification network, outputting construction waste categories when the confidence coefficient of a prediction result of the first stage is greater than a set threshold, and otherwise, collecting spectral images of the construction wastes, inputting the spectral images of the construction wastes into a third classification network, performing second-stage prediction and outputting the construction waste categories.
Preferably, the first, second and third classification networks comprise RetinaNet networks.
Preferably, the meta deep learning algorithm comprises a MAML algorithm.
Preferably, the training phase and the prediction phase both comprise preprocessing of the construction waste 2D images and the construction waste 3D images, and then input to the corresponding network for training or prediction.
Preferably, the preprocessing of the construction waste 2D image includes filtering the 2D image with a bilateral filter.
Preferably, the preprocessing of the construction waste 3D image includes performing median filtering processing on the 3D image.
Preferably, the first stage prediction mode is as follows: and classifying the construction waste in the construction waste 2D image by adopting a first classification network, directly outputting the construction waste classification if the confidence coefficient is greater than a set threshold, otherwise, classifying the construction waste in the construction waste 3D image by adopting a second classification network, directly outputting the construction waste classification if the confidence coefficient is greater than the set threshold, and otherwise, entering a second stage of prediction.
The utility model provides a building rubbish fine classification device based on meta-deep learning, includes:
an image acquisition device: the system comprises a 2D camera, a 3D camera and a high-resolution spectrometer, wherein the 2D camera, the 3D camera and the high-resolution spectrometer are used for correspondingly acquiring 2D images of construction waste, 3D images of the construction waste and spectral images of the construction waste;
a memory: for storing a computer program;
a processor: when the computer program is executed, the construction waste fine classification method based on the meta-deep learning is realized.
Preferably, the image acquisition equipment is erected above a conveyor belt for conveying construction waste.
Preferably, the 2D camera and the 3D camera are located at the same position, and the hyperspectral meter is arranged in front of the 2D camera along the running direction of the conveyor belt.
Compared with the prior art, the invention has the following advantages:
(1) according to the invention, through the fusion of multi-dimensional information, 2-dimensional information such as color and texture of solid waste materials, 3-dimensional depth information and material spectrum information are fused, so that the method is used for identifying the type of construction waste. In the prior art, objects are distinguished only through common 2D images, and only a small number of types of construction waste can be classified, so that the defect that construction waste with similar characteristics is difficult to distinguish through a common 2D optical vision method is overcome.
(2) The invention carries out network training through a meta-deep learning strategy, takes the construction waste of each category as a task, thereby being capable of comprehensively processing the classification problem of the construction waste of multiple categories; meanwhile, the meta-deep learning technology solves the problem of small samples, namely, a deep learning model capable of accurately predicting the category of the construction waste can be trained under the condition that only a small number of construction waste images exist. The prior art usually needs a large number of samples to train a relatively accurate network model, cannot train a model with a relatively high accuracy when the sample amount is small or a sample picture is relatively single, and has relatively weak generalization capability.
(3) The hyperspectral image scanning method based on the multi-spectral image recognition is only used for the training phase and the second phase of the network in the scheme considering that the hyperspectral image scanning time is high in cost. The hyperspectral image plays a role when the model cannot obtain high accuracy rate well only through 2D and 3D images, when the model can obtain confidence coefficient larger than a set threshold value by only identifying the 2D and 3D images, the final result is directly output, when the model outputs the confidence coefficient smaller than the set threshold value, the object is subjected to hyperspectral scanning, and a corresponding spectrogram is input into a network for prediction to obtain a final prediction result.
Drawings
Fig. 1 is a flow chart of a construction waste fine classification method based on meta-deep learning according to the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. Note that the following description of the embodiments is merely a substantial example, and the present invention is not intended to be limited to the application or the use thereof, and is not limited to the following embodiments.
Example 1
The embodiment provides a construction waste fine classification method based on meta-deep learning, which comprises the following steps of;
step 1: the 2D camera, the depth (3D) camera, the high-speed spectrometer and other equipment are deployed above the conveyor belt, and physical connection between the camera equipment and the display equipment is built. The 2D camera and the 3D camera are located at the same position, and the hyperspectral meter is placed at the position, 1 meter in front of the 2D camera in the running direction of the conveyor belt. Note that the time stamp of the continuous shooting is set T, T ═ T1, T2, T3, ·.
Step 2:
and acquiring the information of the construction waste images shot by the 2D camera in real time, and connecting the construction waste images to a digital image acquisition unit to convert the acquired images into digital images. And connecting the 2D equipment with an intelligent computer, and dynamically acquiring the construction waste 2D digital image sample by using the form of control flow. Recording the set of 2D image samples which are continuously shot as D, D ═ Dt1,Dt2,···]；
And acquiring the image information of the construction waste shot by the depth camera in real time, and connecting the image processor to acquire a 3D image corresponding to the target construction waste. The set of 3D images continuously captured is recorded as M, M ═ Mt1,Mt2,···]；
And (4) collecting the building waste classification information by using a high-resolution spectrometer to obtain original spectrum data. The original spectrum image obtained continuously is recorded as N, N ═ Nt1,Nt2,···]。
And step 3:
recording 2D image samples DtnHas a pixel size of Xtn×YtnThe current pixel point is marked as (x)tn,ytn) Then the current pixel point (x)tn,ytn) Respectively, the red, green and blue luminance values in the RGB color space are denoted as IR(xtn,ytn)、IG(xtn,ytn) And IB(xtn,ytn) The brightness difference between the current pixel and the adjacent pixel is recorded as
Computing
Filtering the image by adopting a bilateral filter, wherein smoothing parameters in the filter are respectively recorded as sigmac、σdThe length and width of the filter window are denoted as m and n, respectively. And recording the filtered brightness data of the current pixel point as F (x)tn,ytn) And then:
the unbounded data set of the 2D image after recording bilateral filtering is W, W ═ Wt1,Wt2,···]。
And 4, step 4: similarly, the 3D image is subjected to median filtering to reduce data overflow, and the size of a filtering window is set to be 3 multiplied by 3. The smoothed 3D image data set is denoted V, V ═ Vt1,Vt2,···]。
And 5: from raw spectral data NtnObtaining multiple sets of background spectral vectors EtnCarrying out differential rearrangement on the effective spectrum set F, and comprehensively selecting l effective spectrum bands to form an effective spectrum band set Ftn. Let the effective spectrum set be F, F ═ Ft1,Ft2,···]。
Step 6: note that the pixel coordinate of a certain point in the image is set (x, y), and its neighboring pixels are (x +1, y + 1). The gray scale difference between the x direction and the y direction is | f (x, y) -f (x +1, y) |, | f (x, y) -f (x, y +1) |, respectively. The gray-scale variance product is | f (x, y) -f (x +1, y) | f (x, y) -f (x, y +1) |. Gray scale variance product function SMD2That is, after multiplying two gray level differences in each pixel field, accumulating the multiplied differences one by one, namely:
SMD2＝∑y∑x|f(x,y)-f(x+1,y)|*|f(x,y)-f(x,y+1)|
and selecting the picture with the largest calculated value as the image with the highest definition.
And 7: and (4) manually designing labels for the collected 2D, 3D and hyperspectral images respectively. And for 2D, 3D and hyperspectral images from different acquisition modes, classifying the same category of construction waste into the same label. All marble labels as in 2D are 2D-marble. According to the method, the category of the construction waste is acquired, and the coordinate of the target object is fed back to the grabbing manipulator, so that the problems of target classification and detection are solved. For 2D and 3D images, corresponding detection frames are marked on all objects, so that the regression frames can be well predicted by a later model.
And 8: and adopting an MAML algorithm, and taking RetinaNet as a basic target detection model as a model used by the scheme. And (5) inputting the three types of images marked in the step (6) as a training set into the network for network training. Remembering that each type of construction waste is a task, which is t1,t2,t3...tn]. Each kind of task is trained respectively, and in the training process, each kind only has 10 images, and 20 kinds of construction waste are shared, and the total number is 200 pictures. For 20 categories of construction waste, 12 categories of the construction waste are taken as training tasks, 4 categories of the construction waste are taken as verification tasks, and the last 4 categories of the construction waste are taken as prediction tasks. And for the task of the first category, testing once after the training is finished, and updating the model parameters through back propagation of the test result. And repeating the steps until the nth task is finished and the updating is finished. The final parameters have the ability to achieve higher accuracy in the first n tasks. The network configuration needs a CUDA deep learning acceleration module and a cudnn tool module, a deep learning framework Tensorflow is built and adopted, and the deep learning framework Tensionflow is trained and deployed in an Ubuntu system of Linux.
And step 9: and in the prediction stage, 2D and 3D images obtained by shooting above the conveyor belt are conveyed to the trained model to perform first-stage prediction. And taking the class with the prediction confidence coefficient more than 0.6 given by the model as a prediction class. And (4) performing second-stage prediction on the image with the confidence coefficient smaller than 0.6, scanning the object in the image by using a high-resolution spectrometer to obtain an atlas, and conveying the atlas to the model in the step 8 to obtain the object class. The first-stage prediction mode is as follows: and classifying the construction waste in the construction waste 2D image by adopting a first classification network, directly outputting the construction waste classification if the confidence coefficient is greater than a set threshold value of 0.6, otherwise, classifying the construction waste in the construction waste 3D image by adopting a second classification network, directly outputting the construction waste classification if the confidence coefficient is greater than the set threshold value of 0.6, and otherwise, entering a second stage of prediction.
The invention takes the construction waste of each category as a task through the meta-learning strategy, thereby comprehensively treating the problem of classifying the construction waste of multiple categories. In the prior art, only the conventional construction waste classification problems which are few in categories and easy to distinguish are concerned with by adopting a single classifier;
the method disclosed by the invention integrates multi-dimensional image information to train the model, and not only takes the 2D information of the image into consideration, but also takes the depth information and the spectrum information of the image into consideration. The fusion of multi-dimensional information enables the model to utilize more useful information in prediction;
the invention solves the problem of small samples by using a meta-learning technology, namely, a deep learning model capable of accurately predicting the category of the construction waste can be trained under the condition that only a small number of construction waste images exist. The prior art usually needs a large number of samples to train a relatively accurate network model, cannot train a model with higher accuracy when the sample amount is small or a sample picture is relatively single, and has relatively weak generalization capability;
the invention adopts a second stage of prediction, which plays a double verification role. In the testing process, the hyperspectral meter is only used when the model cannot obtain high accuracy rate well only through 2D and 3D images. When the construction waste runs to the position of the hyperspectral instrument, the instrument scans the spectrum and transmits the spectrogram to the network model to obtain a prediction result. The method well combines 2D, 3D and hyperspectral information.
Example 2
This embodiment provides a building rubbish fine classification device based on meta-deep learning, includes:
an image acquisition device: the system comprises a 2D camera, a 3D camera and a hyperspectral meter, wherein the 2D camera, the 3D camera and the hyperspectral meter are used for correspondingly acquiring 2D images of construction wastes, 3D images of the construction wastes and spectral images of the construction wastes;
a memory: for storing a computer program;
a processor: the method for building garbage fine classification based on meta-deep learning is the same as that in embodiment 1, and includes image acquisition, image processing, network training and prediction, and the specific method thereof is specifically described in embodiment 1, and is not described in detail in this embodiment.
The above embodiments are merely examples and do not limit the scope of the present invention. These embodiments may be implemented in other various manners, and various omissions, substitutions, and changes may be made without departing from the technical spirit of the present invention.
Claims (10)
1. A construction waste fine classification method based on meta-deep learning is characterized by comprising the following steps:
building a construction waste classification model, wherein the construction waste classification model comprises a first classification network for classifying construction waste in a construction waste 2D image, a second classification network for classifying construction waste in a construction waste 3D image and a third classification network for classifying construction waste in a construction waste spectral image;
collecting a construction waste 2D image, a construction waste 3D image and a construction waste spectral image to construct a training set, and training the first classification network, the second classification network and the third classification network based on a meta-deep learning algorithm;
collecting 2D images and 3D images of construction wastes, performing first-stage prediction by using a first classification network and a second classification network, outputting construction waste categories when the confidence coefficient of a prediction result of the first stage is greater than a set threshold, and otherwise, collecting spectral images of the construction wastes, inputting the spectral images of the construction wastes into a third classification network, performing second-stage prediction and outputting the construction waste categories.
2. The method for building rubbish fine classification based on meta deep learning as claimed in claim 1, wherein the first classification network, the second classification network and the third classification network comprise RetinaNet network.
3. The method for building rubbish fine classification based on meta-deep learning as claimed in claim 1, wherein the meta-deep learning algorithm comprises a MAML algorithm.
4. The construction waste fine classification method based on meta-deep learning as claimed in claim 1, wherein the training phase and the prediction phase both comprise preprocessing of construction waste 2D images and construction waste 3D images, and then input to corresponding networks for training or prediction.
5. The method for building rubbish fine classification based on meta-deep learning as claimed in claim 4, wherein the pre-processing of the building rubbish 2D image includes filtering the 2D image by a bilateral filter.
6. The method for building rubbish fine classification based on meta-deep learning as claimed in claim 4, wherein the pre-processing of the building rubbish 3D image comprises performing median filtering processing on the 3D image.
7. The construction waste fine classification method based on the meta-deep learning as claimed in claim 1, wherein the prediction mode of the first stage is as follows: and classifying the construction waste in the construction waste 2D image by adopting a first classification network, directly outputting the construction waste classification if the confidence coefficient is greater than a set threshold, otherwise, classifying the construction waste in the construction waste 3D image by adopting a second classification network, directly outputting the construction waste classification if the confidence coefficient is greater than the set threshold, and otherwise, entering a second stage of prediction.
8. The utility model provides a building rubbish fine classification device based on meta-deep learning which characterized in that includes:
an image acquisition device: the system comprises a 2D camera, a 3D camera and a high-resolution spectrometer, wherein the 2D camera, the 3D camera and the high-resolution spectrometer are used for correspondingly acquiring 2D images of construction waste, 3D images of the construction waste and spectral images of the construction waste;
a memory: for storing a computer program;
a processor: when the computer program is executed, the construction waste fine classification method based on the meta-deep learning is realized according to any one of claims 1 to 7.
9. The construction waste fine classification device based on meta-deep learning as claimed in claim 8, wherein the image acquisition equipment is erected above a conveyor belt for conveying construction waste.
10. The construction waste fine classification device based on meta-deep learning as claimed in claim 9, wherein the 2D camera and the 3D camera are located at the same position, and the hyperspectral meter is arranged in front of the 2D camera along the running direction of the conveyor belt.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110723467.9A CN113435514A (en) | 2021-06-29 | 2021-06-29 | Construction waste fine classification method and device based on meta-deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110723467.9A CN113435514A (en) | 2021-06-29 | 2021-06-29 | Construction waste fine classification method and device based on meta-deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113435514A true CN113435514A (en) | 2021-09-24 |
Family
ID=77757420
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110723467.9A Pending CN113435514A (en) | 2021-06-29 | 2021-06-29 | Construction waste fine classification method and device based on meta-deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113435514A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113971746A (en) * | 2021-12-24 | 2022-01-25 | 季华实验室 | Garbage classification method and device based on single hand teaching and intelligent sorting system |
| CN117150361A (en) * | 2023-09-06 | 2023-12-01 | 浙江立创信息技术服务有限公司 | Garbage classification method, system and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109155002A (en) * | 2016-02-05 | 2019-01-04 | 渊慧科技有限公司 | Strength neural network with external memory |
| CN109389161A (en) * | 2018-09-28 | 2019-02-26 | 广州大学 | Rubbish identification evolutionary learning method, apparatus, system and medium based on deep learning |
| CN110458193A (en) * | 2019-07-08 | 2019-11-15 | 北京百度网讯科技有限公司 | Garbage classification processing method, apparatus and system, computer equipment and readable medium |
| CN110516570A (en) * | 2019-08-15 | 2019-11-29 | 东莞弓叶互联科技有限公司 | A kind of garbage classification recognition methods of view-based access control model and device |
| CN110921146A (en) * | 2019-11-28 | 2020-03-27 | 蚌埠学院 | Household garbage classification method and system based on internet big data and image processing technology |
| CN111582033A (en) * | 2020-04-07 | 2020-08-25 | 苏宁云计算有限公司 | Garbage classification identification method and system and computer readable storage medium |
| CN112058695A (en) * | 2020-08-14 | 2020-12-11 | 光大环境科技（中国）有限公司 | Garbage classification and identification device |
-
2021
- 2021-06-29 CN CN202110723467.9A patent/CN113435514A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109155002A (en) * | 2016-02-05 | 2019-01-04 | 渊慧科技有限公司 | Strength neural network with external memory |
| CN109389161A (en) * | 2018-09-28 | 2019-02-26 | 广州大学 | Rubbish identification evolutionary learning method, apparatus, system and medium based on deep learning |
| CN110458193A (en) * | 2019-07-08 | 2019-11-15 | 北京百度网讯科技有限公司 | Garbage classification processing method, apparatus and system, computer equipment and readable medium |
| CN110516570A (en) * | 2019-08-15 | 2019-11-29 | 东莞弓叶互联科技有限公司 | A kind of garbage classification recognition methods of view-based access control model and device |
| CN110921146A (en) * | 2019-11-28 | 2020-03-27 | 蚌埠学院 | Household garbage classification method and system based on internet big data and image processing technology |
| CN111582033A (en) * | 2020-04-07 | 2020-08-25 | 苏宁云计算有限公司 | Garbage classification identification method and system and computer readable storage medium |
| CN112058695A (en) * | 2020-08-14 | 2020-12-11 | 光大环境科技（中国）有限公司 | Garbage classification and identification device |
Non-Patent Citations (3)
| Title |
|---|
| CANCUI 等: "A recommendation system for meta-modeling: A meta-learning based approach", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
| DAWEI CHEN 等: "Edge Computing Resources Reservation in Vehicular Networks: A Meta-Learning Approach", 《IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY》 * |
| 陈遒正: "少样本下的细粒度视觉识别算法的研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113971746A (en) * | 2021-12-24 | 2022-01-25 | 季华实验室 | Garbage classification method and device based on single hand teaching and intelligent sorting system |
| CN113971746B (en) * | 2021-12-24 | 2022-03-11 | 季华实验室 | Garbage classification method and device based on single hand teaching and intelligent sorting system |
| CN117150361A (en) * | 2023-09-06 | 2023-12-01 | 浙江立创信息技术服务有限公司 | Garbage classification method, system and storage medium |
| CN117150361B (en) * | 2023-09-06 | 2024-04-26 | 浙江立创信息技术服务有限公司 | Garbage classification method, system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101863196B1 (en) | An Apparatus and A Method For Detecting A Defect On Surfaces Based On Deep Learning | |
| RU2694021C1 (en) | Method and apparatus for identifying portions of fragmented material within an image | |
| CN108229561B (en) | Particle product defect detection method based on deep learning | |
| CN112270280B (en) | Open-pit mine detection method in remote sensing image based on deep learning | |
| CN112733950A (en) | Power equipment fault diagnosis method based on combination of image fusion and target detection | |
| US8922674B2 (en) | Method and system for facilitating color balance synchronization between a plurality of video cameras and for obtaining object tracking between two or more video cameras | |
| CN111160249A (en) | Multi-class target detection method of optical remote sensing image based on cross-scale feature fusion | |
| US8488878B2 (en) | Sky detection system used in image extraction device and method using sky detection system | |
| CN113435514A (en) | Construction waste fine classification method and device based on meta-deep learning | |
| CN103530638A (en) | Method for matching pedestrians under multiple cameras | |
| Nguwi et al. | Automatic road sign recognition using neural networks | |
| CN111738338B (en) | Defect detection method applied to motor coil based on cascaded expansion FCN network | |
| Li et al. | Identifying photorealistic computer graphics using second-order difference statistics | |
| CN108108678B (en) | Tungsten ore identification and separation method | |
| CN110569716A (en) | Goods shelf image copying detection method | |
| Kheder et al. | Transfer learning based traffic light detection and recognition using CNN inception-V3 model | |
| CN112488165A (en) | Infrared pedestrian identification method and system based on deep learning model | |
| Culibrk | Neural network approach to Bayesian background modeling for video object segmentation | |
| Qaisar et al. | Scene to text conversion and pronunciation for visually impaired people | |
| CN113763356A (en) | Target detection method based on visible light and infrared image fusion | |
| EP3973476A1 (en) | Systems and methods to train a cell object detector | |
| WO2005024726A1 (en) | Method for detecting object traveling direction | |
| CN113420724B (en) | Unmanned aerial vehicle remote sensing river inlet and outlet recognition method and device | |
| JP2003271929A (en) | Method for recognizing article by robot | |
| Sakila et al. | Image enhancement using morphological operations |
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: 20210924 |