CN111619992A - Intelligent garbage classification system and method based on machine vision - Google Patents

Abstract

The invention discloses an intelligent garbage classification system and method based on machine vision, which are characterized in that a conveying device is designed and realized by applying the MobileNet deep learning and machine vision technology and adopting Raspberry pi and STM32 as main controllers, so that the full-automatic garbage throwing is realized without manpower input. The test result shows that the key performance of the product is greatly improved compared with that of the classification garbage can on the market, and the method successfully realizes the following steps: (1) the method is suitable for different temperatures and illumination environments and accurate judgment; (2) the recognition precision is improved to more than 95 percent; (3) the method has excellent processing capability, can complete detection of 128 points at most within 7 seconds at high speed, and identifies the speed block.

Description

Intelligent garbage classification system and method based on machine vision

Technical Field

The invention relates to an intelligent garbage classification system and method based on machine vision.

Background

Along with the development of economic society, the production amount of domestic garbage in China is rapidly increased, so that the environmental problems caused by the rapid increase become increasingly prominent and become a restriction factor for the development of novel urbanization. The garbage classification is an important link and a key field of ecological civilization construction. China is forcibly promoting the establishment of a household garbage classification management system, various provinces and cities respond, and Changsha cities can formally implement garbage classification in 10 months this year. Before 2019, 8 cities in China have developed garbage classified collection test points, but the popularization effect is not ideal. Reasons for the garbage classification being unable to be effectively carried out include: (1) lack of awareness of residential waste classification; (2) links such as collection, transportation, treatment and the like after the garbage is put in cannot be matched and connected.

The intelligent classification garbage bin aims at realizing the automation and the intellectualization of garbage classification, and effectively promotes the popularization of garbage classification. However, the existing few system schemes at home and abroad have poor practicability and single function, and the common problems include: (1) the low recognition accuracy causes classification errors; (2) identifying that the speed is too slow; (3) the performance is unstable due to obvious influence of temperature, illumination and other environments; (4) lack detailed data analysis, difficult high-efficient butt joint with the recovery link. The lower identification accuracy and the too low identification speed are key pain points which are difficult to use for the garbage cans.

Therefore, there is a need to design an intelligent garbage classification system and method based on machine vision.

Disclosure of Invention

The invention aims to solve the technical problem of providing an intelligent garbage classification system and method based on machine vision, which can realize automatic detection and classified delivery of garbage.

The technical solution of the invention is as follows:

an intelligent garbage classification system based on machine vision comprises a garbage can, wherein an image recognition garbage classification module, a temporary storage box and a plurality of sorting boxes are arranged at the garbage can;

the image recognition garbage classification module comprises an MCU, a trigger, a camera and a distribution device; the trigger, the camera head and the sub-projection device are all connected with the MCU;

the trigger is a pyroelectric infrared sensor and is used for sensing that a human body approaches the garbage can;

the camera is used for acquiring images of the garbage in the temporary storage box;

the MCU is used for carrying out image processing on the image and identifying the type of garbage;

the garbage sorting device is used for conveying garbage into corresponding sorting boxes from the temporary storage boxes according to the sorting instructions sent by the MCU. The MCU is provided with a garbage classification module based on MobileNet and machine vision, and can identify and classify garbage based on the acquired images.

The sub-throwing device comprises a conveying belt (2), a belt wheel (3) and a conveying belt driving mechanism for driving the conveying belt to move;

the number of the belt wheels is 2, the conveying belt is arranged on the 2 belt wheels, the temporary storage box is fixed on the conveying belt, and the bottom of the temporary storage box is provided with a movable plate and a movable plate driving mechanism for controlling the movable plate to open and close; the sorting box is positioned below the conveyor belt;

the distribution and delivery device also comprises a position detection mechanism for detecting the position of each sorting box;

when the movable plate is opened, the garbage in the temporary storage box falls into one sorting box by matching with the position detection mechanism; the conveyor belt has an aperture in the middle so that waste from the buffer bin can pass through the aperture into the sorting bin.

The temporary storage box and the sorting box are internally provided with an odor sensor and an odor deodorization device. The peculiar smell deodorization device specifically adopts an ozone generator for generating ozone to remove peculiar smell.

The overflow reminding module is also included; the overflow reminding module comprises an ultrasonic sensor which is arranged in the sorting box and used for detecting the height of the garbage. And when the height of the garbage is detected to be larger than a set value, starting an alarm.

5. The machine-vision-based intelligent garbage classification system of claim 1 further comprising a server; MCU in the dustbin passes through communication module and server communication connection, and the operator can obtain the junk information through PC or cell-phone APP access server.

6. An intelligent garbage classification method based on machine vision, which is characterized in that the intelligent garbage classification system of any one of claims 1-5 is adopted; the method comprises the following steps:

step 1: detecting that a waste input is close to the trash can based on a trigger;

step 2: starting image shooting;

after the garbage is put into the temporary storage box, starting image shooting to obtain a garbage image; it is a prior art to perform necessary preprocessing, including graying processing, image segmentation processing, etc., on a garbage image.

And step 3: performing spam recognition based on the spam image;

identifying garbage based on a Caffe deep learning framework and a neural network MobileNet model;

and 4, step 4: classifying the garbage based on the garbage recognition result;

and transferring the recognized garbage from the temporary storage box into the corresponding sorting box based on the garbage recognition result.

OpenCV is a BSD license (open source) based distributed cross-platform computer vision library that can run on Linux, Windows, Android, and Mac OS operating systems. The system is lightweight and efficient, is composed of a series of C functions and a small number of C + + classes, provides interfaces of languages such as Python, Ruby, MATLAB and the like, and realizes a plurality of general algorithms in the aspects of image processing and computer vision.

Aiming at the classification requirements and the problems faced by the current urban domestic garbage, the invention provides the following solutions on the basis of carrying out a large amount of research and analysis:

(1) the information about the garbage classification is timely issued and updated on the WeChat applet; after the intelligent classification garbage can is used for recognizing and classifying the target garbage, the garbage can informs a user of the type of the garbage through voice broadcasting, can classify the popular science garbage of the citizens, and helps to reduce the difficulty of memory and classification.

(2) And (4) completing statistics on the types and the quantity of the user garbage by using a big data platform, and generating a visual report at a webpage management end. The staff can be reminded to carry out cleaning work through big data analysis, an optimal recovery route is planned for a clean vehicle, and the operation cost is saved for a sanitation structure.

(3) Utilize intelligent classification rubbish device to realize the accurate discernment separation of wet rubbish futilely, the condition that probably avoids mixing to put in takes place, and relevant technologies such as integrated application internet, big data, thing networking can rationally plan the burning cycle in refuse dump, reduce because burn the pollution that brings the atmosphere.

(4) The invention is provided with the conveying device, and the conveying device is driven by the main control system after the identification and classification of the target garbage are completed, so that the garbage is thrown into the corresponding garbage can.

(5) By adopting the lightweight neural network MobileNet model, the recognition accuracy is ensured to be more than 95%, and the recognition and classification time is controlled within 7 seconds, so that the problems of low accuracy, low recognition speed, obvious environmental influence and the like are solved.

Has the advantages that:

the intelligent garbage classification system and method based on machine vision, provided by the invention, adopt the MobileNet deep learning and machine vision technology and adopt Raspberry pi and STM32 as main controllers, so that a conveying device is designed and realized, and full-automatic garbage throwing is realized without manpower input. The test result shows that the key performance of the invention is greatly improved compared with the classification garbage can on the market, and the following steps are successfully realized: (1) the method is suitable for different temperature and illumination environments and accurate judgment; (2) the recognition precision is improved to more than 95 percent; (3) the method has excellent processing capability, can finish detection of 128 points at most within 7 seconds at high speed, and has high recognition speed.

The WeChat applet of the system is rich in functions, and comprises the following steps: (1) uploading a background autonomous training model on the unidentified garbage picture; (2) inquiring the garbage category by one key; (3) reminding of full garbage; (4) pushing an intelligent message; and (5) an intelligent robot online chat function. The Web management end builds an intelligent supervision platform by using a big data technology, can realize environment accurate supervision based on data decision, management and service, realizes functions of garbage truck path planning, garbage yard incineration cycle planning and the like, establishes a garbage classification communication channel among governments, enterprises and residents, builds a family-community garbage classification recovery network, and is beneficial to realizing a garbage classification long-term management mechanism.

The main characteristics and innovations of the invention are as follows:

(1) machine vision intelligent garbage classification

Aiming at the problems of low recognition accuracy, low recognition speed and the like of the existing garbage classification artifact in the market, the machine vision intelligent classification system based on artificial intelligence and big data is developed by the product. And the identification accuracy is up to more than 95% by utilizing the MobielNet and computer vision, and then the identification return value is returned to the raspberry group main control system.

(2) Intelligent automatic throwing device

At present, the existing intelligent garbage can only remind users of garbage categories, and the users still need to classify the garbage manually. This team has researched and developed an intelligent full autonomic input device, through STM32 and raspberry group serial ports communication, control steering wheel positive and negative rotation. The raspberry group utilizes the result value returned by the machine vision recognition system to drive the motor and the conveyor belt, so that garbage is thrown into the garbage can corresponding to the garbage types.

(3) Deodorizing

Most of garbage cans cannot isolate peculiar smell, and foods are easy to decay and ferment due to hot weather, so that flying insects are attracted to the foods, and toxic gas can be emitted. For solving this practical problem, the intelligent classification garbage bin of this team research and development is equipped with peculiar smell deodorizing device, when detecting peculiar smell, harmful gas through the peculiar smell sensor, starts the deodorization module that disinfects, and isolated rubbish peculiar smell beautifies family sanitation.

(4) Overfill detection

It is not uncommon for a household waste bin to be full but for a user to forget to dump the waste. Aiming at the general phenomenon, the invention designs a garbage overflow reminding function, detects the height of garbage through an ultrasonic sensor, and intelligently reminds a user through a WeChat small program when the garbage is full, thereby avoiding the garbage accumulation hazard.

(5) Real-time visual monitoring

In consideration of the difficult problems that the household garbage classification consciousness is weak and the community management is difficult to really push, the system is based on the internet of things technology, effectively senses and transmits relevant information, further relies on a big data analysis technology at a Web end, counts the number of garbage types thrown by residents and the like, realizes the garbage classification visual monitoring, gradually cultivates the household-community management network. Further realizing that: a. planning the time path of the garbage transport vehicle; and b, managing the waste incineration period of the waste incineration plant, and contributing to establishing a refined and efficient-operation waste recycling system.

Drawings

FIG. 1 is a functional block diagram;

FIG. 2 is a system topology diagram;

FIG. 3 is a diagram of a flip-flop structure;

FIG. 4 is an implementation flow diagram;

FIG. 5 is a view showing the construction of a transfer device;

FIG. 6 is a diagram of other functional blocks;

FIG. 7 is a schematic view of the transport mechanism (front view);

fig. 8 is a schematic view (top view) of the transport mechanism.

Description of reference numerals: 1-temporary storage box, 2-conveyor belt, 3-belt wheel, 4-sorting box and 11-movable plate.

Detailed Description

The invention will be described in further detail below with reference to the following figures and specific examples:

example 1: total planning of system

The system can be divided into five parts:

(1) image recognition garbage classification;

(2) a garbage automatic throwing device;

(3) a cloud server;

(4) a database;

(5) a WeChat applet;

(6) web page end management background

Implementation of garbage image recognition and classification function

The garbage image recognition and classification is a core part in the system, and the image is processed by adopting OpenCV (open source computer vision library), and the position and the characteristics of the target garbage are extracted for subsequent deep learning. And classifying the garbage images by adopting a Mobile net neural network algorithm.

Introduction of functional composition

The image recognition garbage classification device comprises:

(1) and a trigger. The main function of the trigger pyroelectric infrared sensor is to acquire human body approaching information and send a photographing request to the raspberry group, so that the operation of the whole system is triggered

(2) And identifying the garbage image. And the raspberry group, the camera and the OpenCV are used to complete the image positioning and identifying function in a matching way.

(3) And (4) garbage classification. The MobileNet model is responsible for garbage classification. And if the garbage recognition and classification fails, uploading the garbage pictures to a database through the WeChat applet.

Flip-flop

The trigger is composed of a pyroelectric infrared sensor and a raspberry pie and is used for sensing a human body approach signal. When human apart from intelligent classification garbage bin when reaching the inductive distance threshold, when pyroelectric infrared sensor will detect human infrared temperature change, the transmission trigger signal is sent for the raspberry group, and the camera module is opened to the raspberry group then to prepare opening of figure recognition device.

Pyroelectric infrared sensor: when a person is close to the garbage can, the follow-up circuit of the sensor can trigger the switch to act after detection and treatment, and the sensor sends a signal to the raspberry pie.

Raspberry pie: the development board model used by the system is a raspberry pi 3B, 1.2GHz and 64-bit processor chip, the supporting energy consumption is low, the operation is stable, and the image processing requirement and the speed requirement of the system can be completely met.

A camera module: the input device selected by the system is an ov5647 camera which does not need to be additionally driven, and the input device is connected with the raspberry development board through a CSI interface. The module supports image acquisition of 500w pixels, and positioning of target garbage can be completed through the module and a code.

Garbage image recognition

(1) Target refuse positioning

The method adopts OpenCV to realize a background difference algorithm, and firstly converts the acquired three-channel RGB image into a single-channel gray image through a cvtColor function. In order to reduce errors caused by background illumination in the garbage can, binarization processing can be performed on the existing gray level image through a threshold function. And then, subtracting the binary image containing the target to be identified from the binary background image to realize background difference.

(2) Target garbage feature extraction

The target garbage to be detected is clearly identifiable in the image. In order to ensure the detection accuracy, an MORPHOLOGYEX function is used for calling an MORPH _ CLOSE closed operation interface, a large-area white area in a background difference result is communicated, a small-area white noise point is removed, and the image characteristics are enhanced. And finally, acquiring a minimum bounding rectangular frame from the result image of the closed operation through a minAreaRect function, setting an interested area in the original image according to the coordinates of the rectangular frame, and cutting out the target to be detected for subsequent deep learning.

(3) Background difference algorithm

The detection of the target garbage by utilizing the background difference algorithm is mainly divided into four links of background modeling, background updating, target garbage detection, post-processing and the like. The background modeling method adopts a single Gaussian distribution model: the gray value of each pixel point in the image is regarded as a random process x, wherein x is the gray value, and the probability of the occurrence of a certain pixel gray value of the point is assumed to obey gaussian distribution, which can be expressed as:

_tis the standard deviation of x, μ_tIn the expectation of x, the number of the first,_t ²is the variance of x;

the background difference operation process comprises the following steps: firstly, establishing a background image frame B of a container above a conveyor belt by using a mathematical modeling method, recording a current image frame as fn, and respectively recording gray values of pixel points corresponding to the background frame and the current frame as B (x, y) and f_n(x, y) according to formula

Dn(x，y)＝|f_n(x，y)-B(x，y)|

Subtracting the gray values of the corresponding pixel points of the two frames of images, and taking the absolute value of the gray values to obtain a difference image D_n：

Setting threshold value Tc as

Carrying out binarization processing on the pixel points one by one to obtain a binarization image R_n'. Wherein, the point with the gray value of 255 is the foreground (moving target garbage) point, and the point with the gray value of 0 is the background point; for image R_n' conducting connectivity analysis, and finally obtaining a garbage image R containing a complete target_n；

The method not only can accurately position the target to be identified in the identification process, but also can automatically intercept the sample picture from a large amount of picture information in a short time, thereby simplifying the process of obtaining the sample through deep learning. Caffe convolution neural network framework

Caffe has Python related interfaces, and provides a deep learning framework for classification and image segmentation of the target garbage image.

Lightweight neural network model MobileNet

The MobileNet can keep higher accuracy in the tasks of identifying and classifying the garbage images.

The MobileNet model optimizes the traditional full convolution mode, and decomposes the full convolution operation into Depthwise convolution and depth separable convolution to reduce the quantity of parameters to be learned, on the basis, the MobileNet model is further provided with two super parameters, a width factor and a resolution factor to control the size of the model and the resolution of an input image, so that the model is more controllable in scale and the speed of target rubbish, when the width factor is constant, the resolution factor is reduced, or the resolution factor is constant, the width factor is reduced, the parameters and an adding multiplier of a network are relatively reduced, the introduction of the width factor α can obtain a smaller model with less calculation loss, α is used for controlling the number of channels for inputting and outputting rubbish images, the input channel is changed from M to α M, the number of output channels is changed from N to α N, α is 0.5, and the resolution factor rho can reduce the calculation quantity and parameters to rho²The model can be conveniently adjusted by the user, in the example, the value of rho is 0.25, and the input resolutions are set to 224, 192, 160 and 128.

After the traditional convolution mode is subjected to 3 x 3 convolution, the convolution is subjected to a BN layer and a Relu activation function; the depth separable convolution approach used in the MobileNet algorithm: the 3 × 3 conventional convolution approach is replaced by Depthwise convolution and 1 × 1 Pointwise convolution, then passed through the BN and ReLU activation functions, respectively, as with the conventional convolution.

The MobileNet model comprises the following layers:

(1) and (3) rolling layers: the convolutional layer is composed of a plurality of local filters and is mainly used for extracting different local features from an input target garbage feature map.

(2) Batch normalization layer: the batch normalization layer enables the whole model to be more stable, and speeds up the training and convergence speed of the deep convolutional network.

(3) Scaling the layer: the scaling layer scales and displaces the normalized neuron N, and because the method adopts a deep learning framework Caffe to train a MobileNet model and acquire parameters, actual batch normalization calculation is divided into formulas in the framework

The mean, the variance and the scalefactor sum are parameters obtained by learning, mean is a mean vector in the same dimension with the characteristic diagram, variance is a variance vector in the same dimension with the garbage characteristic diagram, and scalefactor is a dimension scaling factor which is a very small constant and is usually 0.0001;

(4) nonlinear activation function layer: in order to enable the lightweight neural network model MobileNet to have nonlinear learning and expression capability, a nonlinear activation function layer is added. In the algorithm based on the MobileNet, a nonlinear rectification function (ReLU) is adopted as a nonlinear activation function, and the calculation formula of the nonlinear activation function is shown as a formula

Detailed description of the preferred embodiments

And after a Caffe deep learning framework and a MobileNet model are installed, training sample pictures of the target garbage to train a corresponding neural network. Taking a harmful garbage lighter as an example, the direction, the position and the shape of the lighter are changed, the lighter is collected for multiple times to form a data set, and the data set is uploaded to a MySQL database.

The method for acquiring the sample data set of the banana peel, the pop can and the battery is the same as the method for acquiring the sample data set by changing the shape of the lighter. The sample data set size for training is shown in the table below.

Table 1: sample data set size

Name (R)	Quantity/sheet of picture
		Lighter	460
Banana peel	378
		Pop-top can	355
Battery with a battery cell	407

The results of the test training using the MobileNet model are shown in table 2

Table 2 training test results

Name (R)	Test collection/sheet	Accuracy of	Fl-score
				Lighter	150	97.52％	0.98
Banana peel	150	96.23％	0.96
				Pop-top can	150	98.19％	0.98
Battery with a battery cell	150	97.31％	0.97

Through testing, the ordinary intelligent garbage cans on the market cannot guarantee the recognition accuracy and the recognition rate, and even can be influenced by the light environment frequently. Due to low accuracy and low recognition rate, the problems of wrong garbage classification, low garbage classification efficiency and the like are caused. The system applies a MobileNet model to well solve the problems. The data pairs of our invention and the general trash can are shown in tables 3-5 by looking up the literature.

TABLE 3 comparison data of the classification rate of the general intelligent garbage can and the invention of the team

Name (R)	Ordinary garbage can identification and classification rate	Intelligent garbage can identification and classification rate
			Lighter	20.3s	5.1s
Banana peel	18.1s	4.5s
			Pop-top can	16.1s	5.4s
Battery with a battery cell	16.3s	4.3s

TABLE 4 comparison data of recognition accuracy of common intelligent garbage can and the invention of the team

Name (R)	Identification accuracy rate of common garbage can	Intelligent garbage can identification accuracy
			Lighter	74.33％	97.52％
Banana peel	87.69％	96.23％
			Pop-top can	89.97％	98.19％
Battery with a battery cell	80.31％	97.31％

TABLE 5 comparison data of the common intelligent garbage can and the invention of the team under the influence of light

Degree of light brightness	Identification accuracy rate of common garbage can	Intelligent garbage can identification accuracy
			Bright and bright	83.71％	98.65％
Is dim	72.94％	97.14％
			Darkness	54.36％	84.77％

It can be seen that the machine vision and MoblieNet model for classification is obviously higher than the ordinary intelligent classification garbage can on the market in terms of identification speed, identification accuracy and adaptive environment capacity. The experiment test shows that the system can be used for classifying 4 types of objects such as lighters, banana peels, pop-top cans and batteries participating in the test, the identification accuracy is higher than 95%, and the system has a higher practical application value in the aspect of garbage classification.

Function realization of garbage conveying device

In order to liberate both hands of a user, fit the living needs of the user, a full-automatic throwing mode is created, and the automatic garbage throwing device is researched and developed by the product. The conveyor configuration is shown in figures 5, 7 and 8.

After the operation of the garbage identification and classification program is finished, a corresponding value is returned to the main control system, the raspberry pi is used for communicating with the STM32 serial port, and the generated value is transmitted to the STM 32. The STM32 sends out an instruction to the motor, and the motor drives the container containing the target garbage to move. After the container arrived fixed position, STM32 continued to send the instruction to the steering wheel and opened the baffle of container bottom below and drop into corresponding kind garbage bin with rubbish. The modules used for the specific transfer device are shown in table 6.

Table 6 modular table of transport device

Other functions

Besides the identification and classification of target garbage and automatic putting, the invention is also provided with a satellite positioning module such as a GPS and the like to provide user position information for generating visual reports, deodorizing peculiar smell, detecting overflow, broadcasting voice and other functions.

TABLE 7 other function implementation

Cloud server

(1) Baidu cloud server

Operating the system: centsos7.4

Hardware platform: single core 2GHz master frequency CPU, 2GB internal memory

Supporting environment and version: python2.7, MySQL5.7

(2) Service background

The service background of the system is programmed by python and sql, and long connection is established between the service background and hardware by using a socket communication mode to realize data transmission.

Database with a plurality of databases

A database used by the delivery system is developed based on MySQL. For the present system, the information required to be stored in the database of the server includes: a login account and a password of a manager; user information; the type and quantity of the garbage thrown into the garbage can and the like. So creating a database with three data tables can suffice.

The WeChat applet adopts a cloud development mode, a database is arranged, and for a client, information in the database comprises: account passwords of management personnel and common users; network news information link; the type of refuse; a classification name; four data tables are created.

And calling the cloud function by the small program, so that the communication connection between the client and the MySQL database can be realized.

WeChat applet

In consideration of convenience of use of a user, the product adopts the WeChat applet to construct the client, does not need to be installed, is taken immediately, can be conveniently acquired and spread in the WeChat, and has excellent use experience.

The following functions can be realized through a mobile phone APP or a WeChat applet:

(1) triple real-time query mode

The user can inquire the garbage types through three ways of character search, image recognition and dialect-equipped voice recognition, and the living needs are greatly met.

(2) Intelligent classified information recommendation function

The small program can intelligently push various garbage classification messages according to the garbage types, the garbage quantity and the social trend and trend of the user every day, and the garbage classification willingness and the throwing consciousness of residents are improved.

(3) Unidentified result uploading function

The user can inquire the garbage of the garbage can without identifying the result on the small program, and can upload the garbage picture to the database without inquiring, so that the aim of correct identification next time can be fulfilled through the system self-training model.

(4) Intelligent robot online chat function

In order to achieve the purposes of intimate accompanying and entertainment, the small program client is also provided with an intelligent chatting interface, and the scene can be accurately judged according to the words of the user and a response is sent out.

Intelligent garbage management function based on image recognition

(1) Automatic garbage classification throwing function

The intelligent garbage can intelligently and automatically recognize and classify garbage input by a user, can respectively input corresponding garbage cans according to four standards of dry garbage, wet garbage, recyclable matters and harmful garbage at present, does not need manual classification of the user, immediately meets the requirements, and solves the classification trouble.

(2) Positioning function such as GPS

And determining the position of the user, and providing data for the network crawler to acquire the home address information. And analyzing the output quantity of different types of garbage in each region of the city according to the information provided by the GPS, and visually collecting information for the data of the management terminal.

(3) Deodorizing function of peculiar smell

Under the influence of the environment, the garbage can ferment in the barrel to generate peculiar smell. When the sensor in the intelligent classification garbage can detects the peculiar smell and reaches a threshold value set by the peculiar smell sensor, a signal is sent to the main controller, the main controller supplies power to start the ozone generation module to eliminate the peculiar smell, and when the smell level recovers to be below the threshold value, the power is cut off to close the ozone generator.

(4) Overfill reminding function

When the garbage reaches the specified height, the garbage is considered to be full, and at the moment, the main controller of the intelligent classification garbage can sends a full prompt signal to the small program to remind a user.

(5) Voice broadcast and key detection function

After the intelligent classification garbage can successfully identifies garbage, the garbage category is broadcasted through the sound card, and under the condition of saving manpower, the classification consciousness of the citizens is improved. If the recognition is unsuccessful, reminding the user that the garbage can not be recognized by voice broadcast, determining the garbage type by the user through a multiple query mode of the WeChat applet, if the garbage type can be determined, controlling the automatic release of the target garbage through a key, and if the garbage type can not be determined, shooting and uploading through the WeChat applet.

Background management function of webpage version

Visualization of garbage classification data: according to the invention, the garbage classified putting data in daily life of the user is collected and analyzed, the user garbage classified database is established, the types and the quantity of the garbage put by the user are counted, the garbage classified putting data are displayed according to the visual icon form, and the convenience of management is improved.

(1) Garbage truck driving time and route planning

The management terminal adopts big data, a web crawler technology and a GPS positioning function, uses user names, product numbers and the like as keywords, crawls user garbage classification data on the Internet, and classifies the collected garbage classification data. And estimating the garbage overflowing degree of the community in combination with cloud computing, and informing the garbage truck of garbage collection according to time. By adopting a GPS satellite positioning technology, the optimal running route of the garbage truck is analyzed, the garbage in a community is timely recovered, the garbage accumulation is avoided, and the convenience is brought to the life of a user.

(2) And planning the garbage incineration period of the garbage yard.

Claims (6)

1. An intelligent garbage classification system based on machine vision is characterized in that,

the garbage can is provided with an image recognition garbage classification module, a temporary storage box and a plurality of sorting boxes;

the image recognition garbage classification module comprises an MCU, a trigger, a camera and a distribution device; the trigger, the camera and the sub-projection device are all connected with the MCU;

the trigger is a pyroelectric infrared sensor and is used for sensing that a human body approaches the garbage can;

the camera is used for acquiring images of the garbage in the temporary storage box;

the MCU is used for carrying out image processing on the image and identifying the type of garbage;

the sorting device is used for conveying the garbage into the corresponding sorting boxes from the temporary storage boxes according to the sorting instructions sent by the MCU.

2. The intelligent machine-vision-based garbage classification system according to claim 1, wherein the throwing device comprises a conveyor belt (2), a belt wheel (3) and a conveyor belt driving mechanism for driving the conveyor belt to move;

the number of the belt wheels is 2, the conveying belt is arranged on the 2 belt wheels, the temporary storage box is fixed on the conveying belt, and the bottom of the temporary storage box is provided with a movable plate and a movable plate driving mechanism for controlling the movable plate to open and close; the sorting box is positioned below the conveyor belt;

the distribution and delivery device also comprises a position detection mechanism for detecting the position of each sorting box;

and when the movable plate is opened, the garbage in the temporary storage box falls into a certain sorting box by matching with the position detection mechanism.

3. The intelligent machine vision-based garbage classification system according to claim 1, wherein an odor sensor and an odor deodorization device are arranged in the temporary storage box and the sorting box.

4. The machine-vision-based intelligent garbage classification system according to claim 1, further comprising an overfill reminder module; the overflow reminding module comprises an ultrasonic sensor which is arranged in the sorting box and used for detecting the height of the garbage.

5. The machine-vision-based intelligent garbage classification system of claim 1 further comprising a server; MCU in the dustbin passes through communication module and server communication connection, and the operator can obtain the junk information through PC or cell-phone APP access server.

6. An intelligent garbage classification method based on machine vision, which is characterized in that the intelligent garbage classification system of any one of claims 1-5 is adopted; the method comprises the following steps:

step 1: detecting that a waste input is close to the trash can based on a trigger;

step 2: starting image shooting;

after the garbage is put into the temporary storage box, starting image shooting to obtain a garbage image;

and step 3: performing spam recognition based on the spam image;

identifying garbage based on a Caffe deep learning framework and a neural network MobileNet model;

and 4, step 4: classifying the garbage based on the garbage recognition result;

and transferring the recognized garbage from the temporary storage box into the corresponding sorting box based on the garbage recognition result.

Images