CN113859803A

CN113859803A - Intelligent identification trash can and intelligent identification method thereof

Info

Publication number: CN113859803A
Application number: CN202111152579.XA
Authority: CN
Inventors: 黄琦均; 祝天培; 吴苏栗; 孙宇豪
Original assignee: Jiaxing Dixing Technology Co ltd
Current assignee: Jiaxing Dixing Technology Co ltd
Priority date: 2021-09-29
Filing date: 2021-09-29
Publication date: 2021-12-31

Abstract

The invention discloses an intelligent identification garbage can and an intelligent identification method thereof, and the intelligent identification garbage can comprises a garbage can body, wherein a garbage throwing area is arranged on the garbage can body, a feed opening is arranged at the bottom of the garbage throwing area, a turnover plate capable of turning over is plugged at the feed opening and driven by a turnover motor, the garbage throwing area is provided with a sensing module and a visual identification module, a plurality of garbage classification storage areas are arranged in the garbage can body, and the garbage classification storage areas correspond to the turnover plate. The intelligent recognition garbage can deal with the situation that more than 90% of domestic garbage is thrown in and classified, the aim of automatic garbage classification is fulfilled, the automatic process of garbage sensing, recognition, classification and containing is fulfilled, human factors are reduced to the maximum extent, and automatic garbage classification is achieved.

Description

Intelligent identification trash can and intelligent identification method thereof

The technical field is as follows:

the invention relates to the field of garbage putting, in particular to an intelligent recognition garbage can and an intelligent recognition method thereof.

Background art:

along with the development and industrialization process of society, social substances are greatly enriched, and various kinds of garbage are generated; in order to better protect the environment and achieve sustainable development, the classification of garbage is greatly promoted in China and even in the world; but the method depends on individuals to classify the garbage, is limited by the influence of factors such as the quality, the education degree, the personal habits and the like of people, and has relatively slow promotion degree of garbage classification and no plain sailing. Particularly, in the household life, the office and the public places, only the most basic garbage cans of different types need to be manually put into the corresponding garbage cans, and the garbage can is passively classified by completely depending on personal behaviors during putting, so that the purpose of garbage classification is basically not achieved.

The invention content is as follows:

the invention provides an intelligent recognition garbage can which can really achieve the purpose of automatic garbage classification, and the intelligent recognition garbage can solve the problem that people only need to lose garbage in the process of classifying the garbage by manpower, so that the whole structure system can complete the subsequent processes of induction, recognition, classification and throwing, and the purpose of automatic garbage classification is really achieved; therefore, the garbage classification accuracy is greatly improved, the social public health and the garbage classification efficiency are improved, and the aim of garbage classification from the source is really achieved.

The invention provides an intelligent identification garbage can which comprises a garbage can body, wherein a garbage throwing area is arranged on the garbage can body, a feed opening is formed in the bottom of the garbage throwing area, a turnover plate capable of turning over is sealed at the feed opening and driven by a turnover motor, the garbage throwing area is provided with a sensing module and a visual identification module, a plurality of garbage classification storage areas are arranged in the garbage can body and correspond to the turnover plate, and the intelligent identification garbage can also comprises a server which is in wireless connection with the sensing module, the visual identification module and the turnover motor.

As preferred technical scheme, garbage bin body top undercut forms rubbish and puts in the district, and response module and visual identification module set up respectively on the lateral wall in rubbish input district.

As the preferred technical scheme, the sensing module is arranged below the visual recognition module.

As a preferred technical scheme, the sensing module is an infrared sensor or a vibration sensor, and the visual identification module is a camera.

As preferred technical scheme, the profile and the size adaptation of returning face plate and feed opening, the upset motor is fixed on garbage bin body inner wall, and the drive shaft of upset motor is inserted and is located in the returning face plate, is connected with the axis of rotation for inserting the opposite side of the returning face plate that is equipped with the drive shaft, and the axis of rotation is rotated and is connected on the garbage bin body, and is corresponding, and the rubbish classification deposits the district and be two, is located the both sides of drive shaft and axis of rotation connecting wire respectively.

As a preferred technical scheme, the device also comprises a power supply module which is electrically connected with the overturning motor, the induction module and the visual identification module.

The invention also provides an intelligent identification method for intelligently identifying the garbage can, which comprises the following steps,

step one, a data collection processing stage, firstly collecting shot garbage pictures; manually screening and marking the garbage pictures, and outputting a marking file in a TXT format; then dividing the formed data set into a training set, a verification set and a test set according to a preset proportion;

step two, in the model training phase, the garbage classification type is set firstly; model parameters are set, a deep neural network model is built by utilizing PyTorch, a channel attention mechanism and a space attention mechanism are added to a feature extraction part of the network, and the feature extraction capability of the model is enhanced, so that the garbage picture features can be better utilized, and the identification precision of the model is improved; then training the model by adopting a sorted image training set and a verification set based on the deep neural network model until the preliminary identification precision of the model reaches more than 85 percent;

step three, performing a garbage classification stage by using the trained model, and receiving a client request by a server to obtain garbage pictures acquired by a camera; secondly, identifying the category of the garbage in the garbage picture by using the deep learning model in the second step, and classifying the garbage corresponding to the garbage picture into recoverable garbage or non-recoverable garbage according to the identification result; then storing the garbage pictures and the corresponding classification data into a database; and finally, returning the garbage category information to the client for further processing.

As a preferable technical proposal, the method also comprises

Step four, a model updating stage, wherein in the actual use process, after the garbage pictures collected by the camera are identified by the deep learning model of the server, the garbage pictures and the corresponding categories thereof are stored in a database built by MySQL; after the newly added garbage pictures are collected to a preset number, carrying out manual inspection on the added garbage pictures and the automatic labels of the deep learning models; and performing data enhancement on the newly added garbage pictures, including operations of randomly rotating the newly added garbage pictures, changing image color channels, adjusting image brightness and the like, and expanding the number of images in the data set by using the garbage pictures after data enhancement, thereby being beneficial to subsequently improving the recognition capability of the deep learning model on the garbage pictures at different angles and under various brightness conditions and further improving the robustness of the deep learning model.

Furthermore, in the fourth step, the deep learning model is finely adjusted by using the newly added garbage pictures, that is, on the basis of the previous model training, the weight in the model is finely adjusted according to the new data set, so that the identification precision of the model is improved to over 90%, and more articles can be effectively and accurately identified.

Compared with the prior art, the invention has the following advantages after adopting the scheme: the garbage can be classified by more than 90 percent of domestic garbage, the aim of automatic garbage classification is fulfilled, the automatic process of garbage sensing, recognition, classification and containing is fulfilled, human factors are reduced to the maximum extent, and automatic garbage classification is fulfilled; the subsequent work of garbage classification is reduced, and the social efficiency is improved.

Description of the drawings:

fig. 1 is a schematic structural diagram of an intelligent trash can of the present invention.

Fig. 2 is a sectional view a-a of fig. 1.

The specific implementation mode is as follows:

the invention will be further described with respect to specific embodiments in conjunction with the following drawings:

as shown in fig. 1, 2, an intelligent recognition garbage bin, including the garbage bin body, it puts in district 3 to be equipped with rubbish on the garbage bin body, 3 bottoms in district are equipped with the feed opening to rubbish, the feed opening shutoff has the returning face plate 4 that can overturn, returning face plate 4 is driven by upset motor 5, rubbish is put in district 3 and is equipped with induction module (not shown) and visual identification module (not shown), it deposits the district to be equipped with a plurality of waste classification in the garbage bin body, waste classification deposits the district and corresponds with returning face plate 4, still include the server, the server is remote server, server and induction module, visual identification module and 5 wireless connection of upset motor. The garbage throwing area can be arranged to be a part of the barrel cover 1, the garbage classification storage area is arranged to be a part of the barrel body 2, and the barrel cover 1 and the barrel body 2 are connected in the prior art in a connection mode such as buckling and sleeving. Wireless connectivity includes communication connections, such as the internet, Wi-Fi networks, cellular networks, Local Area Networks (LANs), Wide Area Networks (WANs), and so forth. And the solar energy turning device also comprises a power supply module, wherein the power supply module is electrically connected with the turning motor, the induction module and the visual identification module, and the power supply module can select a battery type, a plug-in type, a charging type or an outdoor solar energy conventional technical means. Of course, the server has a storage function and a control function, and the control function may be implemented by a conventional processor, which may include one or more microprocessors, microcontrollers, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Programmable Logic Devices (PLDs), Programmable Logic Arrays (PLAs), programmable array logic devices (PALs), or Digital Signal Processors (DSPs), etc.

In this embodiment, the middle region undercut of bung 1 forms rubbish and puts in district 3, and response module and visual identification module are fixed respectively on the lateral wall in rubbish and put in the district, and fixed mode can adopt modes such as bonding, hot melt adhesive to fix and realize. Wherein, the response module is infrared sensor or vibration sensor, and the vision recognition module is the camera to, the response module is located vision recognition module below and sets up.

Upset board 4 and feed opening's profile and size adaptation, upset motor 5 is fixed on 1 inner wall of bung, upset motor 5's drive shaft is inserted and is located in the upset board, opposite side for inserting the upset board that is equipped with the drive shaft is connected with axis of rotation 6, axis of rotation 6 rotates and connects on the bung wall, the axis of rotation is round pin axle spare, it is corresponding, the categorised district of depositing of rubbish is two, set up two categorised districts of throwing in of rubbish in the ladle body promptly, be located the both sides of drive shaft and axis of rotation connecting wire respectively, like this, can realize the downward sloping of upset board above two categorised districts of throwing in rubbish of difference through the corotation and the reversal of upset motor, thereby make the rubbish that is located on the upset board enter into corresponding categorised district of throwing in rubbish, realize waste classification. The structure of this embodiment is adaptable to two classifications of recyclable waste and non-recyclable waste.

Of course, also can adjust the upset mode of returning face plate to the rubbish classification that adapts to more kinds, if adopt the returning face plate that the upset motor will hold rubbish, with three to four kinds of angle slopes: the upper left, the lower left, the upper right and the lower right realize classifying the garbage into four different garbage cans.

According to the intelligent trash can, the operation flow of trash identification and classification putting is as follows, induction trash putting → picture shooting → picture transmission to an identification system → identification and classification of the identification system → operation signal sending → trash can receiving signal → trash can classification putting is completed according to the signal. In particular, the method comprises the following steps of,

the intelligent identification method for intelligently identifying the garbage can comprises the following steps,

Further, also comprises

Step four, a model updating stage, wherein in the actual use process, after the garbage pictures collected by the camera are identified by the deep learning model of the server, the garbage pictures and the corresponding categories thereof are stored in a database built by MySQL; after the newly added garbage pictures are collected to a preset number, carrying out manual inspection on the added garbage pictures and the automatic labels of the deep learning models; and performing data enhancement on the newly added garbage pictures, including operations of randomly rotating the newly added garbage pictures, changing image color channels, adjusting image brightness and the like, and expanding the number of images in the data set by using the garbage pictures after data enhancement, thereby being beneficial to subsequently improving the recognition capability of the deep learning model on the garbage pictures at different angles and under various brightness conditions and further improving the robustness of the deep learning model. Meanwhile, the deep learning model is finely adjusted by using the newly added garbage pictures, namely, on the basis of the last model training, the weight in the model is adjusted in a small range according to the new data set, so that the identification precision of the model is improved to over 90 percent, and more articles can be effectively and accurately identified.

Through the improvement, the garbage automatic sorting system can deal with more than 90% of the domestic garbage throwing and sorting conditions, complete the automatic garbage sorting target, complete the automatic garbage sensing, recognizing, sorting and containing processes, maximally reduce human factors and complete automatic garbage sorting; the subsequent work of garbage classification is reduced, the social efficiency is improved, the existing garbage can be practically applied only by adding a can cover or properly modifying the existing garbage can, and the cost is low; the classification is finished at the source, the subsequent work of garbage classification is reduced, and the social efficiency is improved.

The foregoing is illustrative of the preferred embodiments of the present invention only and is not to be construed as limiting the claims. All the equivalent structures or equivalent process changes made by the description of the invention are included in the scope of the patent protection of the invention.

Claims

1. The utility model provides an intelligent recognition garbage bin, includes the garbage bin body, its characterized in that: be equipped with rubbish on the garbage bin body and put in the district, rubbish is put in district bottom and is equipped with the feed opening, and the feed opening shutoff has the returning face plate that can overturn, and the returning face plate is by upset motor drive, and rubbish is put in the district and is equipped with response module and visual identification module, is equipped with a plurality of waste classification in the garbage bin body and deposits the district, and waste classification deposits the district and corresponds with the returning face plate, still includes the server, server and response module, visual identification module and upset motor wireless connection.

2. The intelligent recognition trash can of claim 1, wherein: the garbage can body is sunken downwards at the top to form a garbage throwing area, and the sensing module and the visual identification module are arranged on the side wall of the garbage throwing area respectively.

3. The intelligent recognition trash can of claim 2, wherein: the sensing module is arranged below the visual recognition module.

4. The intelligent recognition trash can of claim 1, wherein: the sensing module is an infrared sensor or a vibration sensor, and the visual identification module is a camera.

5. The intelligent recognition trash can of claim 1, wherein: the turnover plate is matched with the profile and the size of the feed opening, the turnover motor is fixed on the inner wall of the garbage can body, the driving shaft of the turnover motor is inserted into the turnover plate, the opposite side of the turnover plate, which is provided with the driving shaft, is connected with a rotating shaft, the rotating shaft is rotatably connected onto the garbage can body, correspondingly, two garbage classification storage areas are correspondingly arranged, and the two garbage classification storage areas are respectively positioned on two sides of the driving shaft and a connecting line of the rotating shaft.

6. The intelligent recognition trash can of claim 1, wherein: the turnover motor is electrically connected with the power module, the induction module and the visual recognition module.

7. An intelligent identification method for intelligently identifying a garbage can is characterized in that: comprises the following steps of (a) carrying out,

step two, in the model training phase, the garbage classification type is set firstly; setting model parameters, building a deep neural network model by using PyTorch, adding a channel attention mechanism and a space attention mechanism into a feature extraction part of the network, and enhancing the feature extraction capability of the model so as to better utilize the features of the junk pictures; then training the model by adopting a sorted image training set and a verification set based on the deep neural network model until the preliminary identification precision of the model reaches more than 85 percent;

8. The intelligent identification method for intelligently identifying the trash can according to claim 1, characterized in that: also comprises

Step four, in the model updating stage, after the garbage pictures collected by the camera are identified by the deep learning model of the server, the garbage pictures and the corresponding types thereof are stored in a database built by MySQL; after the newly added garbage pictures are collected to a preset number, carrying out manual inspection on the added garbage pictures and the automatic labels of the deep learning models; and performing data enhancement on the newly added garbage pictures, and expanding the number of the images in the data set by using the garbage pictures after the data enhancement so as to improve the robustness of the deep learning model.

9. The intelligent identification method for intelligently identifying the trash can according to claim 8, wherein: and in the fourth step, the deep learning model is finely adjusted by utilizing the newly added garbage pictures, and the weight in the deep learning model is adjusted according to the new data set, so that the identification precision of the deep learning model is improved to over 90 percent.