CN108960343A - A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing - Google Patents
A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN108960343A CN108960343A CN201810872155.2A CN201810872155A CN108960343A CN 108960343 A CN108960343 A CN 108960343A CN 201810872155 A CN201810872155 A CN 201810872155A CN 108960343 A CN108960343 A CN 108960343A
- Authority
- CN
- China
- Prior art keywords
- solid waste
- target
- frame
- type
- location parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Processing Of Solid Wastes (AREA)
Abstract
The invention discloses a kind of solid waste recognition methods, comprising: obtains the target image comprising solid waste, and divides the grid of M*N in the target image, M and N are positive integer, and M and N are not less than 1;The target gridding for containing multiple target solid wastes, and the centre coordinate based on target gridding are determined in the grid marked off, are generated the corresponding frame to be selected of each target solid waste in target gridding, are obtained multiple frames to be selected;It determines the location parameter of each frame to be selected, and the location parameter of each frame to be selected and target image is inputted into solid waste identification model, export the type of the target solid waste in each frame to be selected.This method may recognize that the type of multiple solid wastes in same grid, to improve solid waste recognition efficiency and accuracy rate, also save fixed-end forces cost.Correspondingly, a kind of solid waste identifying system, device and readable storage medium storing program for executing disclosed by the invention, similarly have above-mentioned technique effect.
Description
Technical field
The present invention relates to solid waste sorting technique field, more specifically to a kind of solid waste recognition methods, system,
Device and readable storage medium storing program for executing.
Background technique
Currently, continue along with human society and develop, the various aspects in socialization production process, the Life Cycle of product
Phase process, it is (in present specification that solid waste is referred to as solid that each functional areas in soil etc. can generate solid waste
It is useless), such as: solid particle, sludge, waste products (pop can, plastic bottle, waste paper carton etc.).These solid wastes are along with the mankind's
Production activity generates, and in order to realize resource recycling, then needs to carry out classification processing to it.And classification recycling is carried out to it,
Then identified firstly the need of to it.
In the prior art, generally solid waste is identified and is classified by manpower, such mode efficiency is more low, far
Inevitably there is careless mistake not as good as the generation speed of solid waste, and when artificial work, the accuracy for causing identification to be classified decreases;If logical
Increase manpower is crossed to improve recognition efficiency, then will increase fixed-end forces cost, consumes more social resources, to be unfavorable for society
The sustainable development of meeting.
Therefore, solid waste recognition efficiency and accuracy rate how are improved, fixed-end forces cost is saved, is that those skilled in the art need
It solves the problems, such as.
Summary of the invention
The purpose of the present invention is to provide a kind of solid waste recognition methods, system, device and readable storage medium storing program for executing, solid to improve
Useless recognition efficiency and accuracy rate, save fixed-end forces cost.
To achieve the above object, the embodiment of the invention provides following technical solutions:
A kind of solid waste recognition methods, comprising:
The target image comprising solid waste is obtained, and divides the grid of M*N in the target image, M and N are positive whole
Number, and M and N are not less than 1;
The target gridding for containing multiple target solid wastes is determined in the grid marked off, and based in the target gridding
Heart coordinate generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple frames to be selected;
It determines the location parameter of each frame to be selected, and the location parameter of each frame to be selected and the target image is inputted admittedly
Useless identification model, exports the type of the target solid waste in each frame to be selected.
Wherein, the centre coordinate based on the target gridding generates each target solid waste in the target gridding
Corresponding frame to be selected obtains multiple frames to be selected, comprising:
Take the centre coordinate of the target gridding as the center of each of target gridding frame to be selected, generates each mesh
Mark the corresponding frame to be selected of solid waste, the multiple frames to be selected being evenly distributed;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched.
Wherein, the location parameter of each frame to be selected of the determination, and by the location parameter of each frame to be selected and the target
Image inputs solid waste identification model, exports the type of the target solid waste in each frame to be selected, comprising:
For each frame to be selected, the type that following step identifies the target solid waste in each frame to be selected is executed respectively, specifically
Are as follows:
Determine that the location parameter of target frame to be selected, the location parameter include at least: the center of the target frame to be selected is sat
Mark, width and height;
By the location parameter of target frame to be selected, the target image and the solid waste identification model, determine described in
The location parameter and type of prediction of target solid waste in target frame to be selected, and determine the corresponding the value of the confidence of every kind of type of prediction;
The size of the corresponding the value of the confidence of more every kind of type of prediction, determines maximum the value of the confidence, and by the maximum the value of the confidence
Corresponding type of prediction is determined as the type of the target solid waste in target frame to be selected.
Wherein, the size of the corresponding the value of the confidence of every kind of type of prediction determines maximum the value of the confidence, and by described in most
The big corresponding type of prediction of the value of the confidence is determined as the type of the target solid waste in target frame to be selected, comprising:
Judge whether the corresponding the value of the confidence of every kind of type of prediction is greater than preset threshold value;
If so, the corresponding type of prediction of the value of the confidence that will be greater than the threshold value is determined as the mesh in target frame to be selected
Mark the type of solid waste.
Wherein, further includes:
Using the location parameter of target frame to be selected as reference, and by Kalman prediction next frame image
The location parameter of the frame to be selected of same position.
A kind of solid waste identifying system, comprising:
Solid waste conveyer belt, for being used cooperatively with conveyer, to transport solid waste;
Solid waste identifies equipment, for the type using the identification solid waste of solid waste recognition methods described in above-mentioned any one;
Solid waste capture apparatus, for being grabbed in predeterminated position when solid waste identification equipment identifies the type of solid waste
Solid waste, so that solid waste Classification Management by type.
A kind of solid waste identification device, comprising:
Obtain module, for obtaining the target image comprising solid waste, and in the target image division M*N grid, M
It is positive integer with N, and M and N are not less than 1;
Generation module for determining the target gridding for containing multiple target solid wastes in the grid marked off, and is based on institute
The centre coordinate for stating target gridding generates the corresponding frame to be selected of each target solid waste in the target gridding, obtain it is multiple to
Select frame;
Identification module, for determining the location parameter of each frame to be selected, and by the location parameter of each frame to be selected and described
Target image inputs solid waste identification model, exports the type of the target solid waste in each frame to be selected.
Wherein, the generation module is specifically used for:
Take the centre coordinate of the target gridding as the center of each of target gridding frame to be selected, generates each mesh
Mark the corresponding frame to be selected of solid waste, the multiple frames to be selected being evenly distributed;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched.
Wherein, further includes:
Prediction module, for and passing through Kalman prediction using the location parameter of target frame to be selected as reference
The location parameter of the frame to be selected of same position in next frame image.
A kind of readable storage medium storing program for executing is stored with computer program, the computer program quilt on the readable storage medium storing program for executing
The step of processor realizes solid waste recognition methods described in above-mentioned any one when executing.
By above scheme it is found that a kind of solid waste recognition methods provided in an embodiment of the present invention, comprising: obtaining includes solid waste
Target image, and in the target image divide M*N grid, M and N are positive integer, and M and N are not less than 1;?
The target gridding for containing multiple target solid wastes, and the centre coordinate based on the target gridding are determined in the grid marked off, it is raw
At the corresponding frame to be selected of each target solid waste in the target gridding, multiple frames to be selected are obtained;Determine the position of each frame to be selected
Parameter is set, and the location parameter of each frame to be selected and the target image are inputted into solid waste identification model, exports each frame to be selected
In target solid waste type.
As it can be seen that the method divides the grid of M*N first in the target image containing solid waste, and it is based on containing multiple mesh
The target gridding for marking solid waste, generates the corresponding frame to be selected of each target solid waste in the target gridding, and then by each of generation
The location parameter and target image of frame to be selected input solid waste identification model, to export the class of the target solid waste in each frame to be selected
Type improves solid waste recognition efficiency to may recognize that the type of multiple solid wastes in same grid;Also, due to the identification
Process is realized based on computer, and each solid waste is corresponding with frame to be selected, so as to avoid mutual dry between each frame to be selected
It disturbs, improves the accuracy rate of solid waste identification, also save fixed-end forces cost.
Correspondingly, a kind of solid waste identifying system, device and readable storage medium storing program for executing provided in an embodiment of the present invention, it is also the same to have
There is above-mentioned technical effect.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of solid waste recognition methods flow chart disclosed by the embodiments of the present invention;
Fig. 2 is grid dividing schematic diagram disclosed by the embodiments of the present invention;
Fig. 3 is another solid waste recognition methods flow chart disclosed by the embodiments of the present invention;
Fig. 4 is a kind of solid waste identifying system schematic diagram disclosed by the embodiments of the present invention;
Fig. 5 is a kind of solid waste identification device schematic diagram disclosed by the embodiments of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a kind of solid waste recognition methods, system, device and readable storage medium storing program for executing, to improve solid waste
Recognition efficiency and accuracy rate save fixed-end forces cost.
Referring to Fig. 1, a kind of solid waste recognition methods provided in an embodiment of the present invention, comprising:
S101, the target image comprising solid waste is obtained, and divides the grid of M*N in the target image, M and N are
Positive integer, and M and N are not less than 1;
S102, the target gridding for containing multiple target solid wastes is determined in the grid marked off, and be based on the target network
Center of a lattice coordinate generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple frames to be selected;
Fig. 2 is referred to, Fig. 2 is grid dividing schematic diagram disclosed in the present embodiment.If Fig. 2 be target image, therein three
It is angular and it is square represent solid waste therein, then Fig. 2 to be divided into the grid of 4*4, i.e. M=N=4.Wherein, triangle and just
It is rectangular to be located at that tertial grid of the third line, then the grid is target gridding.
If the side length of each grid is 2 using the lower left corner Fig. 2 as coordinate origin, then the centre coordinate of the target gridding is
For (5,3), then triangle is with square, the center of corresponding frame to be selected is (5,3).
S103, the location parameter for determining each frame to be selected, and by the location parameter and the target image of each frame to be selected
Solid waste identification model is inputted, the type of the target solid waste in each frame to be selected is exported.
Wherein, the location parameter of frame to be selected includes: shape, side length and center point coordinate.Each frame to be selected is one corresponding
Target solid waste, so if may recognize that two solid wastes by taking Fig. 2 as an example.
Specifically, the solid waste identification model is made of convolutional layer and full articulamentum, it is deep vision model.The model melts
Convolutional Neural and deep neural network have been closed, can extract multi-level images feature, has realized multi-level images feature to identification prediction result
Complicated function mapping, the weight of mapping can be obtained by training, thus its precision is higher, the scope of application is wider, applicability
Preferably.
It should be noted that the present embodiment is illustrated by taking recyclable solid waste as an example, i.e. the type of solid waste can include: easily
Draw tank, plastic bottle and cardboard etc..
As it can be seen that present embodiments providing a kind of solid waste recognition methods, the method is first in the target image containing solid waste
The middle grid for dividing M*N, and based on the target gridding for containing multiple target solid wastes, each target generated in the target gridding is solid
Give up corresponding frame to be selected, and then the location parameter of each of generation frame to be selected and target image are inputted solid waste identification model, with
The type of the target solid waste in each frame to be selected is exported, to may recognize that the type of multiple solid wastes in same grid, is improved
Solid waste recognition efficiency;Also, since the identification process is realized based on computer, and each solid waste is corresponding with frame to be selected, from
And interfering with each other between each frame to be selected is avoided, the accuracy rate of solid waste identification is improved, fixed-end forces cost is also saved.
The embodiment of the invention discloses another solid waste recognition methods, and relative to a upper embodiment, the present embodiment is to technology
Scheme has made further instruction and optimization.
Referring to Fig. 3, another kind solid waste recognition methods provided in an embodiment of the present invention, comprising:
S301, the target image comprising solid waste is obtained, and divides the grid of M*N in the target image, M and N are
Positive integer, and M and N are not less than 1;
S302, the target gridding for containing multiple target solid wastes is determined in the grid marked off, and with the target gridding
Centre coordinate be each of target gridding frame to be selected center, generate the corresponding frame to be selected of each target solid waste, obtain
To equally distributed multiple frames to be selected;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched;
Specifically, in the present embodiment, the frame to be selected center having the same in same grid, and be uniformly distributed, therefore
It can indistinguishably identify all solid wastes in the grid, and will not be interfered with each other between frame to be selected, to improve identification
Efficiency and accuracy rate.
Meanwhile frame to be selected and solid waste correspond, and are mutually matched, and provide good prerequisite item for subsequent identification step
Part is conducive to improve recognition efficiency and accuracy rate.Such as: the side view of pop can is generally rectangle, and capacity is generally
355ml, 310ml etc. therefore, can be based on the shapes of existing pop can on the market and big so its size not has no rule
It is small to preset matched frame to be selected, to improve recognition efficiency.Certainly, other kinds of solid waste can also be pre- according to above-mentioned thought
If corresponding frame to be selected.
S303, the location parameter for determining each frame to be selected, and by the location parameter and the target image of each frame to be selected
Solid waste identification model is inputted, the type of the target solid waste in each frame to be selected is exported.
Preferably, in the present embodiment, each to be selected while exporting the type of the target solid waste in each frame to be selected
The location parameter of frame can also export therewith, in order to accurately grab target solid waste, improve the classification effectiveness of solid waste.
As it can be seen that present embodiments providing another solid waste recognition methods, the method is first in the target figure containing solid waste
The grid of M*N is divided as in, and based on the target gridding for containing multiple target solid wastes, generates each target in the target gridding
The corresponding frame to be selected of solid waste, and then the location parameter of each of generation frame to be selected and target image are inputted into solid waste identification model,
To export the type of the target solid waste in each frame to be selected, to may recognize that the type of multiple solid wastes in same grid, mention
High solid waste recognition efficiency;Also, since the identification process is realized based on computer, and each solid waste is corresponding with frame to be selected,
So as to avoid interfering with each other between each frame to be selected, improve the accuracy rate of solid waste identification, also save fixed-end forces at
This.
Based on above-mentioned any embodiment, it should be noted that the location parameter of each frame to be selected of determination, and will be each
The location parameter of frame to be selected and the target image input solid waste identification model, export the class of the target solid waste in each frame to be selected
Type, comprising:
For each frame to be selected, the type that following step identifies the target solid waste in each frame to be selected is executed respectively, specifically
Are as follows:
Determine that the location parameter of target frame to be selected, the location parameter include at least: the center of the target frame to be selected is sat
Mark, width and height;
By the location parameter of target frame to be selected, the target image and the solid waste identification model, determine described in
The location parameter and type of prediction of target solid waste in target frame to be selected, and determine the corresponding the value of the confidence of every kind of type of prediction;
The size of the corresponding the value of the confidence of more every kind of type of prediction, determines maximum the value of the confidence, and by the maximum the value of the confidence
Corresponding type of prediction is determined as the type of the target solid waste in target frame to be selected.
It should be noted that each frame to be selected can identify the type of wherein solid waste according to above-mentioned steps.For the ease of retouching
It states, is explained by taking target frame to be selected (its quantity is one) as an example below.
Specifically, the location parameter of target frame to be selected can substantially determine the location parameter of target solid waste therein, through knowing
After not, which may be multiple.Such as: if the prediction class of the target solid waste in target frame to be selected
Type are as follows: plastic bottle, pop can and cardboard, corresponding the value of the confidence are as follows: 0.08,0.1 and 0.4, then the value of the confidence by contrast
Size it is found that the corresponding the value of the confidence of cardboard is maximum, then the type of the target solid waste in target frame to be selected is determined as cardboard.
Wherein, the size of the corresponding the value of the confidence of every kind of type of prediction determines maximum the value of the confidence, and by described in most
The big corresponding type of prediction of the value of the confidence is determined as the type of the target solid waste in target frame to be selected, comprising:
Judge whether the corresponding the value of the confidence of every kind of type of prediction is greater than preset threshold value;
If so, the corresponding type of prediction of the value of the confidence that will be greater than the threshold value is determined as the mesh in target frame to be selected
Mark the type of solid waste.
Specifically, when determining the type of the target solid waste in target frame to be selected, can also in such a way that threshold value is set into
Row.Still for above-mentioned, if the type of prediction of the target solid waste in target frame to be selected are as follows: plastic bottle, pop can and cardboard, point
Not corresponding the value of the confidence are as follows: 0.08,0.1 and 0.4, preset threshold is 0.2 at this time, preset rules are as follows: greater than the value of the confidence of threshold value
Corresponding type of prediction is the type of the target solid waste in target frame to be selected.So the corresponding the value of the confidence 0.4 of cardboard is greater than threshold value
0.2, then the type of the target solid waste in target frame to be selected is determined as cardboard.
Based on above-mentioned any embodiment, it should be noted that multiple in same grid when determining in the target image
When the frame to be selected of target solid waste, the location parameter of each frame to be selected and target image are inputted into solid waste identification model, at this time if
All frames to be selected have preset same threshold value, then the frame quantity to be selected exported is then possibly less than the quantity of the frame to be selected of input, still
The accuracy of output can be improved in such cases.
Such as: if 3 frames to be selected have been determined based on target gridding A, this 3 frame and target image input solid wastes to be selected are known
Other model, the confidence Distribution value of this 3 frames to be selected is respectively (assuming that the type of prediction of the target solid waste in all frames to be selected at this time
Are as follows: plastic bottle, pop can and cardboard):
Frame 1 to be selected: the corresponding the value of the confidence of plastic bottle is 0.1, the corresponding the value of the confidence of pop can is 0.2, and cardboard is corresponding to be set
Letter value is 0.2;
Frame 2 to be selected: the corresponding the value of the confidence of plastic bottle is 0.05, the corresponding the value of the confidence of pop can is 0.31, and cardboard is corresponding
The value of the confidence is 0.18;
Frame 3 to be selected: the corresponding the value of the confidence of plastic bottle is 0.06, the corresponding the value of the confidence of pop can is 0.15, and cardboard is corresponding
The value of the confidence is 0.35;
If preset threshold value is 0.3 at this time, what it is greater than threshold value is to export, then the output of solid waste identification model are as follows: to
Selecting the target solid waste in frame 2 is pop can, and the target solid waste in frame 3 to be selected is cardboard.Wherein, due in frame 1 to be selected every kind it is pre-
It surveys the corresponding the value of the confidence of type and is not greater than 0.3, therefore there is no outputs for frame to be selected 1.Therefore, when inputting 3 frames to be selected, may be used also
It can solid waste type of the output less than 3.
Based on above-mentioned any embodiment, it should be noted that further include:
Using the location parameter of target frame to be selected as reference, and by Kalman prediction next frame image
The location parameter of the frame to be selected of same position.
Specifically, the same position in Kalman prediction next frame image can be passed through in solid waste identification process
The location parameter of frame to be selected.Wherein, Kalman filtering can be added to the solid waste identification model of this disclosure, it is pre- to obtain
Survey type solid waste identification model.The location parameter of the frame to be selected of the same position in next frame image predicted, can be used for adjusting
The position of mechanical arm in solid waste assorting process, to realize accurate crawl and sorting.
Based on above-mentioned any embodiment, it should be noted that solid waste recognition methods disclosed in the present embodiment without weighting operations,
To which solid waste recognition efficiency and accuracy rate can be improved.
The weighting operations of the prior art are as follows: carry out multiple neighbouring frames to be selected certain solid waste probability according to present in it
Merge, the frame to be selected that the pop can probability such as center in (5,5), long 4, high 2 is 0.5, with center (7,5), long 3, high 2 it is easy
In (6,5), long 5, high 2 pop can frame to be selected centered on drawing the frame to be selected that tank probability is 0.2 to merge.The present invention is due to being omitted
The operation, and the shapes and sizes of frame to be selected and solid waste are mutually matched, so improving the recognition efficiency and accuracy rate of solid waste.
A kind of solid waste identifying system provided in an embodiment of the present invention is introduced below, a kind of solid waste described below is known
Other system can be cross-referenced with a kind of above-described solid waste recognition methods.
Referring to fig. 4, a kind of solid waste identifying system provided in an embodiment of the present invention, comprising:
Solid waste conveyer belt 401, for being used cooperatively with conveyer, to transport solid waste;
Solid waste identifies equipment 402, for the class using the identification solid waste of solid waste recognition methods described in above-mentioned any embodiment
Type;
Solid waste capture apparatus 403, for being grabbed in predeterminated position when solid waste identification equipment identifies the type of solid waste
Solid waste is taken, so that solid waste Classification Management by type.
The solid waste recognition methods provided based on above-mentioned any embodiment can make solid waste identifying system provided in this embodiment,
Wherein, solid waste is delivered to solid waste by conveyer belt and identifies that equipment, solid waste identification equipment know the solid waste on conveyer belt in real time
Do not classify, and then solid waste is picked by solid waste capture apparatus, processing is centrally placed in the solid waste of same type.
Wherein, solid waste identification equipment described in solid waste identification equipment include: image collecting device, image processing apparatus and
The components such as memory, image processing apparatus therein are mainly used for knowing by solid waste recognition methods described in above-mentioned any embodiment
The type of other solid waste.Solid waste capture apparatus can be mechanical arm etc..
Simultaneously as solid waste is shipped on solid waste conveyer belt, in order to obtain convenient for the target image of identification, can be improved
The shooting performance of image collecting device, such as: its resolution ratio is improved, it is reduced and exposes delay time etc., it so can be to a certain degree
The upper accuracy rate for improving solid waste identification.
Certainly, solid waste identifying system may also include that the devices such as the container for placing different solid wastes.
As it can be seen that present embodiments providing a kind of solid waste identifying system, which is mainly set by solid waste conveyer belt, solid waste identification
Standby and solid waste capture apparatus is constituted, and each equipment room cooperates, and improves solid waste recognition efficiency and accuracy rate, is also improved solid
The efficiency and accuracy rate of useless classification sorting.
A kind of solid waste identification device provided in an embodiment of the present invention is introduced below, a kind of solid waste described below is known
Other device can be cross-referenced with a kind of above-described solid waste recognition methods and system.
Referring to Fig. 5, a kind of solid waste identification device provided in an embodiment of the present invention, comprising:
Obtain module 501, for obtaining the target image comprising solid waste, and in the target image division M*N net
Lattice, M and N are positive integer, and M and N are not less than 1;
Generation module 502 for determining the target gridding for containing multiple target solid wastes in the grid marked off, and is based on
The centre coordinate of the target gridding generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple
Frame to be selected;
Identification module 503, for determining the location parameter of each frame to be selected, and by the location parameter of each frame to be selected and institute
Target image input solid waste identification model is stated, the type of the target solid waste in each frame to be selected is exported.
Wherein, the generation module is specifically used for:
Take the centre coordinate of the target gridding as the center of each of target gridding frame to be selected, generates each mesh
Mark the corresponding frame to be selected of solid waste, the multiple frames to be selected being evenly distributed;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched.
Wherein, the identification module is directed to each frame to be selected, executes following step respectively and identifies the mesh in each frame to be selected
The type for marking solid waste, specifically includes:
Determination unit, for determining that the location parameter of target frame to be selected, the location parameter include at least: the target waits for
Select the centre coordinate, width and height of frame;
Execution unit, for location parameter, the target image and the solid waste identification by target frame to be selected
Model determines the location parameter and type of prediction of the target solid waste in target frame to be selected, and determines every kind of type of prediction pair
The value of the confidence answered;
Recognition unit determines maximum the value of the confidence for the size of the corresponding the value of the confidence of more every kind of type of prediction, and by institute
State the type for the target solid waste that the corresponding type of prediction of maximum the value of the confidence is determined as in target frame to be selected.
Wherein, the recognition unit includes:
Judgment sub-unit, for judging whether the corresponding the value of the confidence of every kind of type of prediction is greater than preset threshold value;
Subelement is executed, it is to be selected that the corresponding type of prediction of the value of the confidence for will be greater than the threshold value is determined as the target
The type of target solid waste in frame.
Wherein, further includes:
Prediction module, for and passing through Kalman prediction using the location parameter of target frame to be selected as reference
The location parameter of the frame to be selected of same position in next frame image.
A kind of readable storage medium storing program for executing provided in an embodiment of the present invention is introduced below, one kind described below is readable to deposit
Storage media can be cross-referenced with a kind of above-described solid waste recognition methods, system and device.
A kind of readable storage medium storing program for executing is stored with computer program, the computer program quilt on the readable storage medium storing program for executing
The step of solid waste recognition methods as described in above-mentioned any embodiment is realized when processor executes.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (10)
1. a kind of solid waste recognition methods characterized by comprising
Obtain include solid waste target image, and divide in the target image grid of M*N, M and N are positive integer, and M
It is not less than 1 with N;
The target gridding for containing multiple target solid wastes is determined in the grid marked off, and is sat based on the target network center of a lattice
Mark, generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple frames to be selected;
It determines the location parameter of each frame to be selected, and the location parameter of each frame to be selected and target image input solid waste is known
Other model exports the type of the target solid waste in each frame to be selected.
2. solid waste recognition methods according to claim 1, which is characterized in that described to be sat based on the target network center of a lattice
Mark, generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple frames to be selected, comprising:
Take the centre coordinate of the target gridding as the center of each of target gridding frame to be selected, it is solid to generate each target
Give up corresponding frame to be selected, the multiple frames to be selected being evenly distributed;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched.
3. solid waste recognition methods according to claim 1, which is characterized in that join the position of each frame to be selected of determination
Number, and the location parameter of each frame to be selected and the target image are inputted into solid waste identification model, it exports in each frame to be selected
The type of target solid waste, comprising:
For each frame to be selected, the type that following step identifies the target solid waste in each frame to be selected is executed respectively, specifically:
Determine that the location parameter of target frame to be selected, the location parameter include at least: centre coordinate, the width of the target frame to be selected
Degree and height;
By the location parameter of target frame to be selected, the target image and the solid waste identification model, the target is determined
The location parameter and type of prediction of target solid waste in frame to be selected, and determine the corresponding the value of the confidence of every kind of type of prediction;
The size of the corresponding the value of the confidence of more every kind of type of prediction determines maximum the value of the confidence, and the maximum the value of the confidence is corresponding
Type of prediction be determined as the type of the target solid waste in target frame to be selected.
4. solid waste recognition methods according to claim 3, which is characterized in that every kind of type of prediction is corresponding to set
The size of letter value, determines maximum the value of the confidence, and it is to be selected that the corresponding type of prediction of the maximum the value of the confidence is determined as the target
The type of target solid waste in frame, comprising:
Judge whether the corresponding the value of the confidence of every kind of type of prediction is greater than preset threshold value;
Consolidate if so, the corresponding type of prediction of the value of the confidence that will be greater than the threshold value is determined as the target in target frame to be selected
Useless type.
5. solid waste recognition methods according to claim 3, which is characterized in that further include:
Using the location parameter of target frame to be selected as reference, and by same in Kalman prediction next frame image
The location parameter of the frame to be selected of position.
6. a kind of solid waste identifying system characterized by comprising
Solid waste conveyer belt, for being used cooperatively with conveyer, to transport solid waste;
Solid waste identifies equipment, for the class using the solid waste recognition methods identification solid waste as described in claim 1-5 any one
Type;
Solid waste capture apparatus, for grabbing solid waste in predeterminated position when solid waste identification equipment identifies the type of solid waste,
So that solid waste Classification Management by type.
7. a kind of solid waste identification device characterized by comprising
Obtain module, for obtaining the target image comprising solid waste, and in the target image division M*N grid, M and N
It is positive integer, and M and N are not less than 1;
Generation module for determining the target gridding for containing multiple target solid wastes in the grid marked off, and is based on the mesh
The centre coordinate for marking grid, generates the corresponding frame to be selected of each target solid waste in the target gridding, obtains multiple frames to be selected;
Identification module, for determining the location parameter of each frame to be selected, and by the location parameter of each frame to be selected and the target
Image inputs solid waste identification model, exports the type of the target solid waste in each frame to be selected.
8. solid waste identification device according to claim 7, which is characterized in that the generation module is specifically used for:
Take the centre coordinate of the target gridding as the center of each of target gridding frame to be selected, it is solid to generate each target
Give up corresponding frame to be selected, the multiple frames to be selected being evenly distributed;
Wherein, the frame to be selected and the target solid waste correspond, and size is mutually matched.
9. solid waste identification device according to claim 7, which is characterized in that further include:
Prediction module for using the location parameter of target frame to be selected as reference, and passes through under Kalman prediction one
The location parameter of the frame to be selected of same position in frame image.
10. a kind of readable storage medium storing program for executing, which is characterized in that be stored with computer program, the meter on the readable storage medium storing program for executing
The step of solid waste recognition methods as described in claim 1-5 any one is realized when calculation machine program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810872155.2A CN108960343A (en) | 2018-08-02 | 2018-08-02 | A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810872155.2A CN108960343A (en) | 2018-08-02 | 2018-08-02 | A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108960343A true CN108960343A (en) | 2018-12-07 |
Family
ID=64467205
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810872155.2A Pending CN108960343A (en) | 2018-08-02 | 2018-08-02 | A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108960343A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113191176A (en) * | 2020-06-08 | 2021-07-30 | 阿里巴巴集团控股有限公司 | Data processing method and device |
CN115331129A (en) * | 2022-10-14 | 2022-11-11 | 彼图科技(青岛)有限公司 | Junk data identification method based on unmanned aerial vehicle and artificial intelligence |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050119916A1 (en) * | 2003-09-19 | 2005-06-02 | Mallett Scott R. | Waste sensor for a disposable container |
CN105772407A (en) * | 2016-01-26 | 2016-07-20 | 耿春茂 | Waste classification robot based on image recognition technology |
CN206126013U (en) * | 2016-08-31 | 2017-04-26 | 苏州德品医疗科技股份有限公司 | Automatic medical waste classification system of discernment rubbish kind |
CN106780612A (en) * | 2016-12-29 | 2017-05-31 | 浙江大华技术股份有限公司 | Object detecting method and device in a kind of image |
CN107092914A (en) * | 2017-03-23 | 2017-08-25 | 广东数相智能科技有限公司 | Refuse classification method, device and system based on image recognition |
CN107451602A (en) * | 2017-07-06 | 2017-12-08 | 浙江工业大学 | A kind of fruits and vegetables detection method based on deep learning |
CN107844794A (en) * | 2016-09-21 | 2018-03-27 | 北京旷视科技有限公司 | Image-recognizing method and device |
CN108212807A (en) * | 2017-12-29 | 2018-06-29 | 深圳市越疆科技有限公司 | A kind of method, apparatus and mechanical arm of robot segregating articles |
CN108297084A (en) * | 2018-03-16 | 2018-07-20 | 五邑大学 | A kind of intelligent conveyor type mechanical arm system based on image recognition |
-
2018
- 2018-08-02 CN CN201810872155.2A patent/CN108960343A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050119916A1 (en) * | 2003-09-19 | 2005-06-02 | Mallett Scott R. | Waste sensor for a disposable container |
CN105772407A (en) * | 2016-01-26 | 2016-07-20 | 耿春茂 | Waste classification robot based on image recognition technology |
CN206126013U (en) * | 2016-08-31 | 2017-04-26 | 苏州德品医疗科技股份有限公司 | Automatic medical waste classification system of discernment rubbish kind |
CN107844794A (en) * | 2016-09-21 | 2018-03-27 | 北京旷视科技有限公司 | Image-recognizing method and device |
CN106780612A (en) * | 2016-12-29 | 2017-05-31 | 浙江大华技术股份有限公司 | Object detecting method and device in a kind of image |
CN107092914A (en) * | 2017-03-23 | 2017-08-25 | 广东数相智能科技有限公司 | Refuse classification method, device and system based on image recognition |
CN107451602A (en) * | 2017-07-06 | 2017-12-08 | 浙江工业大学 | A kind of fruits and vegetables detection method based on deep learning |
CN108212807A (en) * | 2017-12-29 | 2018-06-29 | 深圳市越疆科技有限公司 | A kind of method, apparatus and mechanical arm of robot segregating articles |
CN108297084A (en) * | 2018-03-16 | 2018-07-20 | 五邑大学 | A kind of intelligent conveyor type mechanical arm system based on image recognition |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113191176A (en) * | 2020-06-08 | 2021-07-30 | 阿里巴巴集团控股有限公司 | Data processing method and device |
CN115331129A (en) * | 2022-10-14 | 2022-11-11 | 彼图科技(青岛)有限公司 | Junk data identification method based on unmanned aerial vehicle and artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Automatic detection and classification system of domestic waste via multimodel cascaded convolutional neural network | |
CN109948506A (en) | A kind of multi-angle garbage classification cloud platform based on deep learning | |
CN109389161A (en) | Rubbish identification evolutionary learning method, apparatus, system and medium based on deep learning | |
CN107909107A (en) | Fiber check and measure method, apparatus and electronic equipment | |
CN107368614A (en) | Image search method and device based on deep learning | |
CN104809187A (en) | Indoor scene semantic annotation method based on RGB-D data | |
CN107480643A (en) | A kind of robot of Intelligent refuse classification processing | |
CN103426172B (en) | The method for tracking target of view-based access control model and device | |
CN109740752A (en) | Depth model training method and device, electronic equipment and storage medium | |
CN110110752A (en) | A kind of identification of rubbish and classification method, device and terminal device | |
CN110516625A (en) | A kind of method, system, terminal and the storage medium of rubbish identification classification | |
CN108960343A (en) | A kind of solid waste recognition methods, system, device and readable storage medium storing program for executing | |
CN110458096A (en) | A kind of extensive commodity recognition method based on deep learning | |
Kshirsagar et al. | Artificial Intelligence‐Based Robotic Technique for Reusable Waste Materials | |
Nafiz et al. | Convowaste: An automatic waste segregation machine using deep learning | |
Wu et al. | A garbage detection and classification method based on visual scene understanding in the home environment | |
CN109857878A (en) | Article mask method and device, electronic equipment and storage medium | |
CN109492699A (en) | Passway for transmitting electricity method for three-dimensional measurement and device | |
CN108241662A (en) | The optimization method and device of data mark | |
CN108508844A (en) | Transferred product system and transfer method based on plane secondary light source and machine vision | |
Bhandari | Automatic waste sorting in industrial environments via machine learning approaches | |
CN112862136B (en) | Sorting method, sorting device, sorting system, server and storage medium | |
Han et al. | A deep learning model for automatic plastic waste monitoring using unmanned aerial vehicle (UAV) data | |
CN106778872A (en) | The clustering method and device of the connected graph based on density | |
CN113610100A (en) | Image recognition-based garbage processing method and device 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |