CN111242010A - Method for judging and identifying identity of litter worker based on edge AI - Google Patents
Method for judging and identifying identity of litter worker based on edge AI Download PDFInfo
- Publication number
- CN111242010A CN111242010A CN202010026367.6A CN202010026367A CN111242010A CN 111242010 A CN111242010 A CN 111242010A CN 202010026367 A CN202010026367 A CN 202010026367A CN 111242010 A CN111242010 A CN 111242010A
- Authority
- CN
- China
- Prior art keywords
- edge
- garbage
- throwing
- identity
- identifying
- 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
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000006399 behavior Effects 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 16
- 238000002372 labelling Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 7
- OLBCVFGFOZPWHH-UHFFFAOYSA-N propofol Chemical compound CC(C)C1=CC=CC(C(C)C)=C1O OLBCVFGFOZPWHH-UHFFFAOYSA-N 0.000 claims description 7
- 229960004134 propofol Drugs 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 8
- 238000004364 calculation method Methods 0.000 abstract description 7
- 230000002349 favourable effect Effects 0.000 abstract 1
- 238000011161 development Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- GOLXNESZZPUPJE-UHFFFAOYSA-N spiromesifen Chemical compound CC1=CC(C)=CC(C)=C1C(C(O1)=O)=C(OC(=O)CC(C)(C)C)C11CCCC1 GOLXNESZZPUPJE-UHFFFAOYSA-N 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- 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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of judging litter and identifying the identity of illegal persons, in particular to a method for judging and identifying the identity of a litter person based on an edge AI (artificial intelligence) algorithm, wherein the method for judging and identifying the identity of the litter person based on the edge AI algorithm provided by the invention realizes calculation resources and storage resources for accelerating identification through an edge AI calculation method, can quickly identify whether a litter action exists, adopts a face recognition technology to identify the identity of the litter person when judging the litter action by adopting the edge AI algorithm, helps to make person identity tracing, traces back the litter person to the illegal person through the illegal action, and is favorable for educating the individual who throws the litter and establishing a responsibility system.
Description
Technical Field
The invention relates to the technical field of judging litter and identifying illegal personnel identities, in particular to a method for judging and identifying the identity of a litter personnel based on an edge AI.
Background
With the rapid development of social economy in China, the problems of further improving city grade, improving city image, beautifying city environment and the like become important subjects in front of governments, environment improvement, city management optimization, new city image building, city grade improvement and full development of activities of civilized cities are promoted.
With the continuous deepening of the urbanization process, the insufficient urban environment monitoring efficiency becomes an urgent problem to be solved, household garbage is taken as an important influence factor influencing the urban environment, a large amount of household garbage is generated along with the improvement of the urban living standard, the effective control of the garbage becomes a major problem of urban environmental protection work, a large amount of randomly discarded garbage in the past causes harm to water and soil resources of people, and the survival and development of people are directly threatened; information collection, processing and analysis of sanitation work as an important ring has not yet been mature solution, and is usually only discovered and stopped by sanitation workers, so that the efficiency is low. Under the condition, the original environment monitoring mode cannot meet the actual needs of the digital city, and the comprehensive and efficient environment detection system based on artificial intelligence becomes the inevitable appeal of urbanization development.
However, the current irreconcilable phenomenon of garbage throwing is still all the more than ever, and the garbage throwing is difficult to effectively control and depends on the independent consciousness of people. The existing artificial intelligence algorithm model is used for identifying whether a garbage throwing behavior exists or not, and the existing artificial intelligence algorithm model firstly judges whether a person enters a garbage throwing forbidden area or not, and then voice broadcasting is carried out to remind that garbage can not be thrown.
The method only plays a role in reminding, the realization effect depends on the quality of individuals, and the behavior of forbidding throwing garbage randomly cannot be really realized.
Disclosure of Invention
In order to solve the problem that the prior art mentioned in the background art can only remind people who forbid garbage throwing and cannot confirm the identity of garbage throwing personnel, the invention provides a method for judging and identifying the identity of garbage throwing personnel based on edge AI, which comprises the following steps:
s100, performing original data algorithm model learning by using a fastercnnn framework of a deep learning target detection algorithm based on Region Proposal by a cloud algorithm center;
s200, distributing an original data example model obtained by the famtercnn framework learning to each edge computing node by the cloud example center through a cloud unloading mode of cloud computing;
s300, the field camera transmits the collected video and pictures of the personnel throwing garbage to an edge computing node, the edge computing node is combined with an existing example model to carry out computing, whether the personnel throwing garbage has illegal behaviors of throwing garbage or not is judged according to a computing result, if yes, an instruction is sent to a field terminal device, and the field terminal device gives a prompt according to the instruction;
and S400, the edge computing node transmits the judgment result and the collected video and picture of the person throwing garbage to a cloud computing center for storage or viewing.
On the basis of the scheme, further, the raw data example model is obtained through the following steps:
s110, constructing an image database for identifying people and garbage throwing behaviors, and performing model depth model training and labeling on the image to form source data;
and S120, performing AI modeling based on the source data and the fastercnnn algorithm network to generate an original data example model.
On the basis of the above scheme, further, the step S110 includes the following steps:
s111, obtaining pictures containing people and garbage throwing behaviors as data sources;
and S112, carrying out model depth model training and labeling on the data source and forming source data, wherein the labeling comprises the position of the image, the name of the image, the width and the height of the image, the dimension of the image, the name of a labeled object and the xy coordinate value of the Bbox.
On the basis of the above scheme, further, the step S120 includes the following steps:
s121, extracting feature maps of the candidate images from the original images through a multilayer convolutional neural network, wherein the feature maps can be shared for use by a subsequent RPN network and a full connection layer;
s122, generating region prosages through RPN layers by the feature maps, firstly generating a stack of anchors, then cutting and filtering, judging whether the anchors belong to the foreground or the background through softmax, and meanwhile correcting the anchors by utilizing bounding box regression to form more accurate prosages;
s123, obtaining a propofol feature map with a fixed size on a Roi Pooling layer by the propofol;
s124, the generic feature map performs full-connection operation, specific category classification is performed on the full-connection layer and softmax, and a bounding box regression operation is completed by utilizing L1 Loss to obtain the accurate position of the object, so that an original data example model is generated.
On the basis of the above scheme, further, the S122 includes the following steps:
the fastercnnn adopts the SPP layer in the SPP-net, and introduces a multitask loss function for calculating loss of bbox regression and classification.
On the basis of the above scheme, further, the multitask loss function is:
wherein, the p isiThe probability of being the target is predicted for the anchor,Nclsis the total number of anchors, ti=(tx,ty,tw,th) Is a vector representing the anchor, the offset predicted by the RPN training stage,is and tiVectors with the same dimension represent anchors, the offset of the RPN training stage relative to gt is actual, gt is a labeled correct target, lambda is a labeled number, and L isregTo mark the correct number of targets, LclsTo classify the number, NregThe number of positive anchors.
On the basis of the above scheme, further, the calculation of the loss of bbox regression and classification is expressed by using a thought vector (x, y, w, h), and the calculation steps are as follows:
1) and (Δ x, Δ y), Δ x ═ PWdx(P),Δy=Phdy(P);
2) Scaling (S)W,Sh),SW=PWdW(P),Sh=Phdh(P);
Wherein x, y, w, h respectively represent the central point coordinate and width and height of the image, PWIs the target position, PhIs a target height, dx(P) is the target abscissa, dy(P) is the object ordinate, dw(P) is the zoom size, dh(P) is the zoom height, SwTo target zoom size, ShThe target zoom height is obtained.
On the basis of the scheme, the equipment model adopted by the EDGE computing node is VA-TU-DL, and preferably VA-TU-EDGE-B01.
Compared with the prior art, the method for judging and identifying the identity of the litter thrower based on the edge AI has the following technical effects:
the invention can quickly and accurately identify whether the garbage throwing behavior exists in the garbage throwing forbidden area or not, and identify the identity of the garbage throwing personnel, thereby helping to establish a set of complete garbage throwing forbidden system.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a block diagram of a method for determining and identifying the identity of a person who discards garbage according to an edge AI of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it should be noted that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplicity of description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
The invention provides a method for judging and identifying the identity of a litter worker based on an edge AI, which comprises the following steps:
s100, performing original data algorithm model learning by using a fastercnnn framework of a deep learning target detection algorithm based on Region Proposal by a cloud algorithm center;
s200, distributing an original data example model obtained by the famtercnn framework learning to each edge computing node by the cloud example center through a cloud unloading mode of cloud computing;
s300, the field camera transmits the collected video and pictures of the personnel throwing garbage to an edge computing node, the edge computing node is combined with an existing example model to carry out computing, whether the personnel throwing garbage has illegal behaviors of throwing garbage or not is judged according to a computing result, if yes, an instruction is sent to a field terminal device, and the field terminal device gives a prompt according to the instruction;
and S400, the edge computing node transmits the judgment result and the collected video and picture of the person throwing garbage to a cloud computing center for storage or viewing.
In specific implementation, as shown in fig. 1, the method comprises the following steps:
s100, performing original data algorithm model learning by using a fastercnnn framework of a deep learning target detection algorithm based on Region Proposal by a cloud algorithm center;
s200, distributing an original data example model obtained by the famtercnn framework learning to each edge computing node by the cloud example center through a cloud unloading mode of cloud computing;
s300, the field camera transmits the collected video and pictures of the personnel throwing garbage to an edge computing node, the edge computing node is combined with an existing example model to carry out computing, whether the personnel throwing garbage has illegal behaviors of throwing garbage or not is judged according to a computing result, if yes, an instruction is sent to the field terminal device, and the field terminal device gives a prompt according to the instruction.
The edge computing device detects that a person holding garbage enters a monitoring site and starts tracking the person holding garbage and the garbage on hands until the garbage is thrown out, the edge computing device calculates a calculation result value by combining a process video of the garbage thrown out with an example model, judges whether the person throwing the garbage has illegal behaviors of throwing the garbage according to the calculation result, specifically judges whether the illegal behaviors of throwing the garbage exist or not by comparing a numerical value of the result with a set threshold value through the example model, and if the illegal behaviors of throwing the garbage are lower than the threshold value, the garbage is thrown; if the instruction exists, the instruction is sent to the field terminal equipment, and the field terminal equipment gives a prompt according to the instruction, wherein the prompt can be used for playing voice or other related operations, such as broadcasting through a sound-producing device or an alarm bell;
it should be noted that, not only can the field video data be acquired by the camera, but also a person skilled in the art can implement the method by other devices for acquiring image or video data, and redundant description is not repeated here;
and S400, the edge computing node transmits the judgment result and the collected video and picture of the person throwing garbage to a cloud computing center for storage or viewing.
The edge computing node transmits the judgment result and the collected video and picture of the personnel throwing garbage to the cloud computing center for storage or checking, so that the identity of the illegal personnel can be identified conveniently, the illegal personnel can be traced back by illegal behaviors, and education and responsibility system establishment for the individual throwing garbage are facilitated.
The method for judging and identifying the identity of the garbage thrower based on the edge AI can quickly and accurately identify whether the garbage thrower exists in the garbage throwing forbidden area or not, identify the identity of the garbage thrower and help establish a complete garbage throwing forbidden system.
Preferably, the raw data example model is obtained by the following steps:
s110, constructing an image database for identifying people and garbage throwing behaviors, and performing model depth model training and labeling on the image to form source data;
and S120, performing AI modeling based on the source data and the fastercnnn algorithm network to generate an original data example model.
Preferably, the raw data example model is obtained by the following steps:
s110, constructing an image database for identifying people and garbage throwing behaviors, and performing model depth model training and labeling on the image to form source data;
and S120, performing AI modeling based on the source data and the fastercnnn algorithm network to generate an original data example model.
In specific implementation, as a CNN network target detection method, fastercnnn extracts feature maps of an original image by using a convolutional layer + pooling layer mode, and provides support for a subsequent RPN layer and a full connection layer; the RPN network is mainly used for generating region explosals;
firstly, after the picture with any size passes Conv layers, the original image is cut into 1/16, then the Feature Map enters into Intemediate layer filter to generate Anchor box finally, then judge that anchors belong to the foreground or the background, namely judge that the object is or is not, namely generate a matrix with MxN size on the image, firstly, performing convolution with 1x1 to output the image with WxHx18 size,
then whether the object is finally obtained through a softmax function and a resurcope layer, and meanwhile, a bounding box regression correction anchor box forms accurate propofol;
secondly, utilizing the prosals generated by the RPN and the feature map obtained by the last layer of the VGG16 to finally obtain the prosal feature map with fixed size, and entering the back to perform target identification and positioning by utilizing full connection operation;
finally, forming a feature map with a fixed size on a Roi Pooling layer for full connection operation, and classifying specific categories by using Softmax, wherein the feature map is mapped into values (0,1) through a Softmax function, and the cumulative sum of the values is 1 to meet the property of probability, so that the value with the largest probability can be selected as a prediction target, classifying according to the values, then completing bounding box regression operation by using L1 Loss to obtain the accurate position of the object, namely rpn _ Loss _ bbox uses a smooth L1 calculation function, performing prediction marking on the anchor box in rpn-data, and calculating the offset between the anchor box and gt _ boxes; meanwhile, the bounding box regression obtains the position offset bbox _ pred of each region proxy, and regression obtains a more accurate target detection position and finally obtains the accurate position of the object;
finally, an example model of the raw data is generated in the above manner.
Preferably, the step S110 includes the steps of:
s111, obtaining pictures containing people and garbage throwing behaviors as data sources;
and S112, carrying out model depth model training and labeling on the data source and forming source data, wherein the labeling comprises the position of the image, the name of the image, the width and the height of the image, the dimension of the image, the name of a labeled object and the xy coordinate value of the Bbox.
In specific implementation, model depth model training and labeling are carried out by the following method:
preferably, the step S120 includes the steps of:
s121, extracting feature maps of the candidate images from the original images through a multilayer convolutional neural network, wherein the feature maps can be shared for use by a subsequent RPN network and a full connection layer;
s122, generating region prosages through RPN layers by the feature maps, firstly generating a stack of anchors, then cutting and filtering, judging whether the anchors belong to the foreground or the background through softmax, and meanwhile correcting the anchors by utilizing bounding box regression to form more accurate prosages;
s123, obtaining a propofol feature map with a fixed size on a Roi Pooling layer by the propofol;
s124, the generic feature map performs full-connection operation, specific category classification is performed on the full-connection layer and softmax, and a bounding box regression operation is completed by utilizing L1 Loss to obtain the accurate position of the object, so that an original data example model is generated.
Preferably, the S122 includes the steps of:
the fastercnnn adopts the SPP layer in the SPP-net, and introduces a multitask loss function for calculating loss of bbox regression and classification.
Preferably, the multitask penalty function is:
wherein, the p isiThe probability of being the target is predicted for the anchor,Nclsis the total number of anchors, ti=(tx,ty,tw,th) Is a vector representing the anchor, the offset predicted by the RPN training stage,is and tiVectors with the same dimension represent anchors, the offset of the RPN training stage relative to gt is actual, gt is a labeled correct target, lambda is a labeled number, and L isregTo mark the correct number of targets, LclsTo classify the number, NregThe number of positive anchors.
Preferably, the computation of the loss of bbox regression and classification, expressed using a thought vector (x, y, w, h), is performed as follows:
1) and (Δ x, Δ y), Δ x ═ PWdx(P),Δy=Phdy(P);
2) Scaling (S)W,Sh),SW=PWdW(P),Sh=Phdh(P);
Wherein x, y, w, h respectively represent the central point coordinate and width and height of the image, PWIs the target position, PhIs a target height, dx(P) is the target abscissa, dy(P) is the object ordinate, dw(P) is the zoom size, dh(P) is the zoom height, SwTo target zoom size, ShThe target zoom height is obtained.
Preferably, the equipment model adopted by the edge computing node is VA-TU-DL.
In particular, the EDGE equipment model is preferably VA-TU-EDGE-B01.
In order to test the application of the method for judging and identifying the identity of the people who throw garbage based on the edge AI in practice, the following tests are provided:
100000 human face pictures and 100000 garbage throwing pictures are collected by the example center for marking training, and a human face recognition model and a garbage throwing recognition model are established. The edge equipment is provided with a face recognition model and a garbage throwing recognition model, 1500 face pictures and 1000 garbage throwing pictures are collected by a camera of the edge equipment every day, and the model studies and judges the pictures collected by the camera and outputs the studied and judged results. The pictures collected on site are returned to the calculation center for continuous marking training, the accuracy of the face recognition model reaches 98%, and the accuracy of the garbage throwing recognition model reaches 90%.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. A method for judging and identifying the identity of a litter worker based on an edge AI is characterized by comprising the following steps:
s100, performing original data algorithm model learning by using a fastercnnn framework of a deep learning target detection algorithm based on Region Proposal by a cloud algorithm center;
s200, distributing an original data example model obtained by the famtercnn framework learning to each edge computing node by the cloud example center through a cloud unloading mode of cloud computing;
s300, the field camera transmits the collected video and pictures of the personnel throwing garbage to an edge computing node, the edge computing node is combined with an existing example model to carry out computing, whether the personnel throwing garbage has illegal behaviors of throwing garbage or not is judged according to a computing result, if yes, an instruction is sent to a field terminal device, and the field terminal device gives a prompt according to the instruction;
and S400, the edge computing node transmits the judgment result and the collected video and picture of the person throwing garbage to a cloud computing center for storage or viewing.
2. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 1, wherein the raw data example model is obtained by:
s110, constructing an image database for identifying people and garbage throwing behaviors, and performing model depth model training and labeling on the image to form source data;
and S120, performing AI modeling based on the source data and the fastercnnn algorithm network to generate an original data example model.
3. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 2, wherein the step S110 comprises the steps of:
s111, obtaining pictures containing people and garbage throwing behaviors as data sources;
and S112, carrying out model depth model training and labeling on the data source and forming source data, wherein the labeling comprises the position of the image, the name of the image, the width and the height of the image, the dimension of the image, the name of a labeled object and the xy coordinate value of the Bbox.
4. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 2, wherein the step S120 comprises the steps of:
s121, extracting feature maps of the candidate images from the original images through a multilayer convolutional neural network, wherein the feature maps can be shared for use by a subsequent RPN network and a full connection layer;
s122, generating region prosages through RPN layers by the feature maps, firstly generating a stack of anchor boxes, then cutting and filtering, judging whether the anchors belong to the foreground or the background through softmax, and meanwhile correcting the anchor boxes by utilizing bounding box regression to form more accurate prosages;
s123, obtaining a propofol feature map with a fixed size on a Roi Pooling layer by the propofol;
s124, the generic feature map performs full-connection operation, specific category classification is performed on the full-connection layer and softmax, and a bounding box regression operation is completed by utilizing L1 Loss to obtain the accurate position of the object, so that an original data example model is generated.
5. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 4, wherein the step S122 comprises the steps of:
the fastercnnn adopts the SPP layer in the SPP-net, and introduces a multitask loss function for calculating loss of bbox regression and classification.
6. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 5, wherein the multitasking loss function is:
wherein, the p isiThe probability of being the target is predicted for the anchor,Nclsis the total number of anchors, ti=(tx,ty,tw,th) Is a vector representing the anchor, the offset predicted by the RPN training stage,is and tiVectors with the same dimension represent anchors, the offset of the RPN training stage relative to gt is actual, gt is a labeled correct target, lambda is a labeled number, and L isregTo mark the correct number of targets, LclsTo classify the number, NregThe number of positive anchors.
7. The method for determining and identifying identity of people throwing garbage based on edge AI of claim 5, wherein: the computation of the loss of bbox regression and classification, expressed using thought vectors (x, y, w, h), is performed as follows:
1) and (Δ x, Δ y), Δ x ═ PWdx(P),Δy=Phdy(P);
2) Scaling (S)W,Sh),SW=PWdW(P),Sh=Phdh(P);
Wherein x, y, w, h respectively represent the central point coordinate and width and height of the image, PWIs the target position, PhIs a target height, dx(P) is the target abscissa, dy(P) is the object ordinate, dw(P) is the zoom size, dh(P) is the zoom height, SwTo target zoom size, ShThe target zoom height is obtained.
8. The method for determining and identifying identity of people throwing garbage based on edge AI according to claim 1, wherein: the equipment model adopted by the edge computing node is VA-TU-DL.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010026367.6A CN111242010A (en) | 2020-01-10 | 2020-01-10 | Method for judging and identifying identity of litter worker based on edge AI |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010026367.6A CN111242010A (en) | 2020-01-10 | 2020-01-10 | Method for judging and identifying identity of litter worker based on edge AI |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111242010A true CN111242010A (en) | 2020-06-05 |
Family
ID=70864096
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010026367.6A Pending CN111242010A (en) | 2020-01-10 | 2020-01-10 | Method for judging and identifying identity of litter worker based on edge AI |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111242010A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112115846A (en) * | 2020-09-15 | 2020-12-22 | 上海迥灵信息技术有限公司 | Identification method and identification device for littering garbage behavior and readable storage medium |
CN115394065A (en) * | 2022-10-31 | 2022-11-25 | 之江实验室 | AI-based automatic identification packet loss behavior alarm method and device |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107666594A (en) * | 2017-09-18 | 2018-02-06 | 广东电网有限责任公司东莞供电局 | Method for monitoring illegal operation in real time through video monitoring |
CN107730904A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Multitask vehicle driving in reverse vision detection system based on depth convolutional neural networks |
CN107730906A (en) * | 2017-07-11 | 2018-02-23 | 银江股份有限公司 | Zebra stripes vehicle does not give precedence to the vision detection system of pedestrian behavior |
CN107766889A (en) * | 2017-10-26 | 2018-03-06 | 济南浪潮高新科技投资发展有限公司 | A kind of the deep learning computing system and method for the fusion of high in the clouds edge calculations |
CN109167963A (en) * | 2018-09-28 | 2019-01-08 | 广东马上到网络科技有限公司 | A kind of monitoring method and system of public civilization |
CN109165582A (en) * | 2018-08-09 | 2019-01-08 | 河海大学 | A kind of detection of avenue rubbish and cleannes appraisal procedure |
CN109598303A (en) * | 2018-12-03 | 2019-04-09 | 江西洪都航空工业集团有限责任公司 | A kind of rubbish detection method based on City scenarios |
WO2019144575A1 (en) * | 2018-01-24 | 2019-08-01 | 中山大学 | Fast pedestrian detection method and device |
CN110189292A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | A kind of cancer cell detection method based on Faster R-CNN and density estimation |
CN110378935A (en) * | 2019-07-22 | 2019-10-25 | 四创科技有限公司 | Parabolic recognition methods based on image, semantic information |
CN110659622A (en) * | 2019-09-27 | 2020-01-07 | 北京文安智能技术股份有限公司 | Detection method, device and system for garbage dumping |
-
2020
- 2020-01-10 CN CN202010026367.6A patent/CN111242010A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107730904A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Multitask vehicle driving in reverse vision detection system based on depth convolutional neural networks |
CN107730906A (en) * | 2017-07-11 | 2018-02-23 | 银江股份有限公司 | Zebra stripes vehicle does not give precedence to the vision detection system of pedestrian behavior |
CN107666594A (en) * | 2017-09-18 | 2018-02-06 | 广东电网有限责任公司东莞供电局 | Method for monitoring illegal operation in real time through video monitoring |
CN107766889A (en) * | 2017-10-26 | 2018-03-06 | 济南浪潮高新科技投资发展有限公司 | A kind of the deep learning computing system and method for the fusion of high in the clouds edge calculations |
WO2019144575A1 (en) * | 2018-01-24 | 2019-08-01 | 中山大学 | Fast pedestrian detection method and device |
CN109165582A (en) * | 2018-08-09 | 2019-01-08 | 河海大学 | A kind of detection of avenue rubbish and cleannes appraisal procedure |
CN109167963A (en) * | 2018-09-28 | 2019-01-08 | 广东马上到网络科技有限公司 | A kind of monitoring method and system of public civilization |
CN109598303A (en) * | 2018-12-03 | 2019-04-09 | 江西洪都航空工业集团有限责任公司 | A kind of rubbish detection method based on City scenarios |
CN110189292A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | A kind of cancer cell detection method based on Faster R-CNN and density estimation |
CN110378935A (en) * | 2019-07-22 | 2019-10-25 | 四创科技有限公司 | Parabolic recognition methods based on image, semantic information |
CN110659622A (en) * | 2019-09-27 | 2020-01-07 | 北京文安智能技术股份有限公司 | Detection method, device and system for garbage dumping |
Non-Patent Citations (4)
Title |
---|
I_LINDA: "Mask R-CNN—Faster RCNN学习笔记", Retrieved from the Internet <URL:《https://blog.csdn.net/qq_34713831/article/details/85122631》> * |
Y.Z.Y.: "目标检测——Faster R-CNN(三)", Retrieved from the Internet <URL:《https://blog.csdn.net/qq_42823043/article/details/90744280》> * |
做计算机视觉的小硕妹子: "论文阅读之Siamese RPN以及一些其他内容的补充", Retrieved from the Internet <URL:《https://blog.csdn.net/weixin_44287997/article/details/100782196》> * |
大哲子: "图像处理—对象的定位与识别", Retrieved from the Internet <URL:《https://blog.csdn.net/qq_36183496/article/details/98581895》> * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112115846A (en) * | 2020-09-15 | 2020-12-22 | 上海迥灵信息技术有限公司 | Identification method and identification device for littering garbage behavior and readable storage medium |
CN112115846B (en) * | 2020-09-15 | 2024-03-01 | 上海迥灵信息技术有限公司 | Method and device for identifying random garbage behavior and readable storage medium |
CN115394065A (en) * | 2022-10-31 | 2022-11-25 | 之江实验室 | AI-based automatic identification packet loss behavior alarm method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108596277B (en) | Vehicle identity recognition method and device and storage medium | |
CN108062349A (en) | Video frequency monitoring method and system based on video structural data and deep learning | |
CN113822247B (en) | Method and system for identifying illegal building based on aerial image | |
CN111353338B (en) | Energy efficiency improvement method based on business hall video monitoring | |
CN112668375B (en) | Tourist distribution analysis system and method in scenic spot | |
CN111626162A (en) | Overwater rescue system based on space-time big data analysis and drowning warning situation prediction method | |
CN112183472A (en) | Method for detecting whether test field personnel wear work clothes or not based on improved RetinaNet | |
CN111242010A (en) | Method for judging and identifying identity of litter worker based on edge AI | |
CN114267082B (en) | Bridge side falling behavior identification method based on depth understanding | |
CN111935450A (en) | Intelligent suspect tracking method and system and computer readable storage medium | |
CN112819068A (en) | Deep learning-based real-time detection method for ship operation violation behaviors | |
CN112637550B (en) | PTZ moving target tracking method for multi-path 4K quasi-real-time spliced video | |
CN110490150A (en) | A kind of automatic auditing system of picture violating the regulations and method based on vehicle retrieval | |
CN111125290B (en) | Intelligent river patrol method and device based on river growth system and storage medium | |
CN115294519A (en) | Abnormal event detection and early warning method based on lightweight network | |
CN112395967A (en) | Mask wearing monitoring method, electronic device and readable storage medium | |
CN113095301A (en) | Road occupation operation monitoring method, system and server | |
CN115223204A (en) | Method, device, equipment and storage medium for detecting illegal wearing of personnel | |
CN113536997B (en) | Intelligent security system and method based on image recognition and behavior analysis | |
CN112836590B (en) | Flood disaster monitoring method and device, electronic equipment and storage medium | |
CN114821486B (en) | Personnel identification method in power operation scene | |
CN115830381A (en) | Improved YOLOv 5-based detection method for mask not worn by staff and related components | |
CN114170677A (en) | Network model training method and equipment for detecting smoking behavior | |
CN113963310A (en) | People flow detection method and device for bus station and electronic equipment | |
CN110443197A (en) | Intelligent understanding method and system for visual scene |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200605 |