Automatic classification and recognition and recycling device for waste beverage bottles

Abstract

The application discloses an automatic classification identification and recycling device for waste beverage bottles. In order to solve the problems of high power consumption and identification accuracy bottleneck in the existing intelligent recycling system, the image acquisition unit of the application includes a low-power event capture unit and a camera; the low-power event capture unit includes an event camera and an SNN processor, and the SNN processor performs reasoning according to the perception of the environment by the event camera; after detecting the delivery of a beverage bottle in the field of view of the event camera, the SNN processor generates a trigger signal according to the pulse event output by the event camera; the control unit is configured to control the camera to capture the current image after a first delay after receiving the trigger signal, and perform identification based on the captured image. The application not only overcomes the defect of high system power consumption in the traditional scheme, but also solves the problem of identification accuracy bottleneck, and achieves the technical effect of low power consumption and high accuracy. The application is suitable for the fields of artificial intelligence and Internet of Things.

Description

Technical Field

[0001] This invention relates to an intelligent waste recycling device, and more particularly to a high-precision automatic sorting, identification, and recycling device for waste beverage bottles with ultra-low power consumption. Technical Background

[0002] In recent years, the beverage industry has experienced rapid development, and various beverages have become an indispensable part of people's lives. However, this rapid growth and soaring sales have also led to the problem of difficult-to-recycle discarded beverage bottles, causing significant environmental pollution. Current methods for recycling discarded beverage bottles have proven largely ineffective, and the environmental pollution caused by these bottles remains unresolved.

[0003] Traditionally, waste beverage bottles are recycled through household collection. People typically collect the empty bottles after finishing their drinks and then take them to a recycling center to exchange for cash. However, this method is not only cumbersome, but many households also lack the habit of collecting waste beverage bottles, often simply discarding them directly. Although the government has promoted waste sorting in recent years and set up recycling bins in many places, many of these are ineffective, as many people do not place their waste beverage bottles in the designated recycling bins.

[0004] Traditional methods for recycling waste beverage bottles have significant room for improvement. Simplifying the recycling process is the most effective way to increase recycling efficiency, while raising public awareness and motivation to recycle waste beverage bottles is the fundamental approach. However, accurately, economically, and efficiently classifying, identifying, and recycling beverage bottles remains a challenge in this field. Using machine vision for waste sorting has become a popular research direction in this area.

[0005] Prior art 1: CN109201514B;

[0006] Prior art 2: CN109165568A;

[0007] Prior art 3: CN109684979B;

[0008] The prior art 1 discloses a scheme for classifying garbage based on deep learning neural networks, which overcomes the problems of high cost and low efficiency of manual sorting.

[0009] In view of the shortcomings of existing technologies that require barcode recognition or intact bottle body for identification of beverage bottles, prior art 2 discloses a technical solution that can identify deformed beverage bottles.

[0010] Existing technology 3 also classifies garbage based on neural networks, and it notes that the system should not idle most of the time but should remain in standby mode. Garbage classification and identification are only performed after garbage to be classified is detected, which reduces system power consumption to some extent. The scheme for detecting the presence of garbage includes sampling and then identifying images, but this scheme is inherently a high-power scheme.

[0011] The common drawbacks of existing machine vision-based intelligent waste sorting solutions are high standby power consumption and low accuracy.

[0012] Regarding power consumption, the system requires continuous comparison of the current image with the previous background image to detect the presence of recyclable waste, a process that consumes significant resources. A smart trash can is a typical end-user device; if it requires mains power to operate, the system's power consumption is extremely high, thus diminishing the economic benefits of intelligent waste sorting.

[0013] Regarding accuracy, since the disposal of beverage bottles and other trash by users is a dynamic process with various random elements, traditional AI classification methods based on static images struggle to handle this dynamic spatiotemporal problem, especially when the images captured during image acquisition are not of optimal quality. In other words, existing recognition technologies face an accuracy bottleneck due to randomness.

[0014] Based on these technological backgrounds, this invention discloses a waste sorting scheme that can achieve low power consumption and high precision, especially using AI image recognition and other technologies to achieve the technical goal of sorting and identifying waste beverage bottles. This enables residents to self-deposit within the community and earn income, which is conducive to cultivating user habits, reducing the pollution of waste beverage bottles to the environment, and increasing the social benefits of waste recycling. Summary of the Invention

[0015] The purpose of this invention is to provide an automatic sorting, identification, and recycling system and device for waste beverage bottles, thereby effectively improving the recycling efficiency of waste beverage bottles and, to some extent, increasing people's enthusiasm for recycling waste beverage bottles. It can also alleviate the environmental pollution caused by waste beverage bottles to a certain extent.

[0016] An automatic sorting, identification, and recycling device for waste beverage bottles is provided. This device includes: an image acquisition unit comprising a low-power event capture unit and a camera; wherein the low-power event capture unit includes an event camera and an SNN processor, the SNN processor performing inference based on the event camera's perception of the environment; when the event camera detects a beverage bottle being delivered within its field of view, the SNN processor generates a trigger signal based on the pulse event output by the event camera; a control unit configured to, upon receiving the trigger signal, control the camera to delay capturing the current image after a first delay; an image classification unit configured to classify the image captured by the camera and obtain a classification result; and a recognition result processing unit configured to analyze and process the classification result of the beverage bottle, analyzing whether the beverage bottle type is one of the types required for recycling set by the administrator: if so, it is recycled; otherwise, the user is notified that the beverage bottle is not a recyclable type.

[0017] In one embodiment, when the event camera detects a beverage bottle being delivered within its field of view, the SNN processor generates a trigger signal based on the input pulse event output by the event camera, after a first delay; the control unit is configured to control the camera to delay capturing the current image upon receiving the trigger signal.

[0018] In one embodiment, the automatic classification, identification, and recycling device further includes: a recognition model training unit, which trains images and types of beverage bottles into corresponding recognition models, and deploys the trained recognition models on the image classification unit.

[0019] In one embodiment, the automatic classification, identification, and recycling device further includes: a mode selection unit, which allows the user to select manual or automatic identification mode after scanning the code and activating the identification command; in manual mode, the user needs to manually press the identification button to identify the beverage bottle; in automatic identification mode, the system automatically identifies the type of beverage bottle; and an image classification unit sets the mode according to the mode selected by the mode selection unit, and classifies the types of beverage bottles according to the identification model deployed by the image acquisition unit and the identification model training unit.

[0020] In one embodiment, the automatic sorting, identification, and recycling device is further configured to: query the price information set by the administrator in the web management unit, and estimate the value of the recycled beverage bottles based on the sorting results and the price information set by the administrator.

[0021] In one embodiment, the automatic sorting, identification, and recycling device further includes: an automatic recycling unit that uses hydraulic technology; after determining that the beverage bottle is one that needs to be recycled, the automatic recycling unit is activated to compress the volume of the beverage bottle and store it in the recycling bin; a recycling management unit that automatically senses the current capacity of the recycling bin through an infrared sensor module installed on the recycling bin; and an administrator that can view the status of the recycling bin through a terminal device and determine whether the recycling bin has reached its maximum capacity based on the status of the recycling bin; and a cloud unit that uploads the user's identification information and the identified amount to a cloud server after the automatic recycling is completed, providing a data access interface through backend technology.

[0022] In one embodiment, the automatic classification, identification, and recycling device is configured such that: after a user logs in on a WeChat mini-program terminal, the device accesses the user's identification data and identification amount through backend technology and server, displays the user's information and identification amount using frontend technology, and the user withdraws the identified amount through WeChat.

[0023] In one embodiment, the trigger signal is generated by the SNN processor at the end of the first network decision window.

[0024] In one embodiment, the first delay is between 70 and 90 milliseconds.

[0025] In one embodiment, at the end of the first network decision window, the number of active pixels of the corresponding event camera is a first number; at a first delay after the end of the first network decision window, the number of active pixels of the corresponding event camera is a second number; and the first number is less than the second number.

[0026] In one embodiment, within the interference window after the first delay after the first network decision window ends, the maximum number of active pixels of the event camera is also less than the second number.

[0027] The present invention has the following beneficial technical effects:

[0028] This invention features a small-sized device that can be deployed in high-traffic areas such as shops, schools, and hotels. By offering rewards for recycling, it increases people's motivation to recycle waste beverage bottles. Through visual and image recognition technologies, it overcomes the problems of cumbersome processes, long cycles, and low efficiency associated with traditional waste beverage bottle recycling. Crucially, this invention allows for low-power standby, especially beneficial for non-reward-based intelligent recycling systems, reducing operating costs, enhancing commercial value, and promoting the widespread practical application of intelligent recycling solutions. Furthermore, this invention not only overcomes the issue of decreased accuracy that often arises after introducing a low-power event capture unit, but also overcomes the accuracy bottleneck inherent in traditional solutions in this field. Attached Figure Description

[0029] Figure 1 This is a framework diagram of an intelligent recycling system;

[0030] Figure 2 This is a schematic diagram of an automatic sorting, identification, and recycling system and device for discarded beverage bottles;

[0031] Figure 3 This is a diagram showing the relationships between the core components of an automated sorting, identification, and recycling system and device for discarded beverage bottles.

[0032] Figure 4 This is a schematic diagram of the improved image acquisition unit of the present invention;

[0033] Figure 5 This is a graph showing the change in the number of activated pixels of the event camera over time during a delivery process;

[0034] Figure 6 This is a sample image of the training data;

[0035] Figure 7 This is a diagram illustrating the principle of the optimal window for capturing RGB images;

[0036] Figure 8 This is a schematic diagram of the physical structure of an automatic sorting, identification, and recycling system and device for discarded beverage bottles;

[0037] Figure 9 This is a business process diagram of a self-service recycling terminal;

[0038] Figure 10 This is a schematic diagram of a system capacity detection scheme based on ultrasound.

[0039] Figure 11 This is a diagram of a web-based remote management system;

[0040] Figure 12 This is a diagram illustrating the functions of the WeChat client. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Figure 1This is a framework diagram of the entire intelligent recycling system of this invention. The intelligent recycling system includes an AI classification module, a self-service recycling terminal, a web management system, a WeChat mini-program, and a server. The AI ​​classification module consists of an Edgeboard and is responsible for deploying the AI ​​classification model and classifying image data. The self-service recycling terminal consists of a Raspberry Pi, a screen, an STM32 microcontroller, an ultrasonic module, a servo motor, a speaker, and a camera. It uses MQTT technology and the WeChat mini-program for communication and is responsible for providing a user interface, capacity monitoring, and automatic recycling. The web management system adopts a front-end and back-end separation technology, with the front-end using a Vue2 architecture and the back-end using a Spring Boot architecture. It uses MQTT technology to manage the devices and is responsible for providing administrators with the ability to view recycling information, modify recycling prices, and manage devices. The WeChat mini-program communicates with the self-service recycling terminal via MQTT technology and is responsible for functions such as user login via QR code, balance withdrawal, points redemption, and viewing historical data. The server uses an Nginx reverse proxy, stores data through MySQL, and uses Tomcat to deploy the web management system and the WeChat mini-program backend, responsible for storing and accessing user data and deploying the web management system and the WeChat mini-program backend.

[0043] Figure 2 This illustration shows an automatic sorting, identification, and recycling system and apparatus for discarded beverage bottles (hereinafter referred to as the recycling system and apparatus, or self-service recycling terminal) disclosed in one embodiment of the present invention. It includes an image acquisition unit, a mode selection unit, an image classification unit, a recognition result processing unit, an automatic recycling unit, a human-computer interaction unit, a recycling management unit, and a user scanning unit, as well as a recognition model training unit, a cloud unit, a user terminal, and a web management unit for information interaction with the automatic sorting, identification, and recycling system and apparatus. The units and modules mentioned in this invention are synonymous.

[0044] The image acquisition unit acquires image data of the recognition area, and the image classification unit classifies the beverage bottles in the recognition area according to the image data. The image classification unit classifies the bottles according to the mode selected by the mode selection unit. Optionally, the recycling system and device also include a mode selection unit. After the user scanning unit processes the user's scanning operation via WeChat mini-program on the user's terminal and initiates the recognition command, the mode selection unit allows the user to select between manual and automatic recognition modes. In manual mode, the user needs to manually press the recognition button to identify the beverage bottle; in automatic recognition mode, the user only needs to place the beverage bottle into the recycling system and device, and the system will automatically identify the type of beverage bottle, at least under the capture of the image acquisition unit's camera.

[0045] The identification result processing unit analyzes and processes the bottle classification information after the valuation of the recycled bottles is completed. It analyzes whether the bottle type is the type that the administrator has set for recycling. If so, the bottle is recycled; otherwise, the user is reminded that the bottle type is not the type to be recycled.

[0046] The automatic sorting, identification, and recycling device is also configured to: query the price information set by the administrator in the web management unit, and estimate the value of the recycled beverage bottles based on the sorting results and the price information set by the administrator.

[0047] The automatic recycling unit uses hydraulic technology. After analyzing whether the bottle placed by the user in the identification area is a bottle that needs to be recycled, it will activate the automatic recycling device to compress the bottle's volume and store it in the recycling bin.

[0048] The recycling management unit automatically senses the current capacity of the recycling bin through an infrared sensor module installed on the bin. Administrators can view the status of the recycling bin on the terminal device and determine whether the recycling bin has reached its maximum capacity based on the status.

[0049] After the automatic recycling process ends, the cloud-based unit will upload the user's identification information and the amount collected to the cloud server, providing a data access interface through backend technology.

[0050] After logging in to the WeChat mini-program terminal, the user terminal can access the user's identification data and identification amount through backend technology and server, and display the user's information and identification amount using frontend technology. The user can then withdraw the identified amount through WeChat.

[0051] For image acquisition units, in existing technologies, OpenCV image technology is usually used to preprocess the frames, convert them into grayscale images, apply Gaussian blur, set the initial image frame as the background, capture the image of each frame, calculate the difference between the current frame and the background frame, and obtain a difference map to determine whether there is object movement in the recognition area.

[0052] The image classification unit uses MSE (mean squared error) technology to calculate the similarity between two adjacent frames to determine whether the two frames are stationary. The calculation formula is as follows:

[0053]

[0054] Where m and n represent the width and height of the image, respectively, and I(·) and K(·) represent the pixel values ​​of the two test images. The pixel values ​​of corresponding positions in the two test images are subtracted, and the results are accumulated and normalized. The similarity between the two frames is obtained based on the processing result to determine whether the scene is still. When the scene is still, the acquired image data is sent to the image classification unit for classification.

[0055] However, this type of solution suffers from high power consumption because it requires constant calculations to determine whether there is movement in the recognition area and whether the image is static. Both the image classification unit and the sensor require a large amount of computational resources to continuously output image frames to acquire environmental images, which is particularly evident in non-reward-based intelligent recycling systems. Furthermore, not all scenarios provide static images for image classification, and moving objects to be identified exhibit image blurring on traditional image sensors. This significantly reduces the accuracy of AI recognition because the randomness introduced by the user creates a precision bottleneck. A technical solution to this problem will be disclosed later in this invention.

[0056] Figure 3 This diagram illustrates the core components of an automatic sorting, identification, and recycling system and apparatus for waste beverage bottles according to an embodiment of the present invention. The following description is merely an example illustrating the main components of the recycling system and apparatus, and does not constitute a limitation on the inventive concept; it may also include other components.

[0057] The automatic sorting, identification and recycling system and device for waste beverage bottles includes a main control unit 21, an image acquisition unit 23, an AI image analysis and processing unit 22, a human-computer interaction unit 24, a recycling management unit 26, and a recycling processing unit 25.

[0058] The image acquisition unit 23 includes a Spedal 902 camera, which has a dynamic imaging range of 69.5° and a field of view of 100°, making it more suitable for acquiring image data. The camera is mounted directly above the recycling bin and, in this embodiment, is used to acquire image data of the identification area.

[0059] The AI ​​image analysis and processing unit 22 (i.e., the image classification unit) uses the Baidu EdgeBoard computing card. This unit adopts an FPGA chip architecture, which is small in size and high in performance, with a maximum computing power of 1.2 GOPS. In this embodiment, it provides the function of image data analysis.

[0060] The human-computer interaction unit 24 uses a 10-inch high-definition touch screen with a resolution of 1024×600px. In this embodiment, it provides human-computer interaction functions, and users can view classification results and mode selection functions on the screen.

[0061] The recycling management unit 26 uses an STM32F103C8T6 minimum system with two SG90 servos. In this embodiment, after the image classification unit finishes classification, it sends instructions to the STM32F103C8T6 main control chip via serial port. The STM32F103C8T6 drives the SG90 to open and close the recycling bin according to the instructions.

[0062] The recycling unit 25 uses the HC-SR501 unit group, which is an automatic control unit based on infrared technology. It employs the LHI799 probe design, offering high sensitivity. When the beverage bottle capacity reaches its sensing range, it outputs a high level. The STM32F103C8T6 processes the sensor level data and sends it to the main control unit 1 via serial port. The main control unit 1 then uploads the data to the server. Administrators can view the recycling bin capacity through a web management unit or mobile terminal. The relationship between the recycling bin capacity and the infrared sensor's location is shown in Table 1.

[0063] Table 1

[0064]

[0065] The main control unit 1 is a Raspberry Pi 4B. This unit features a 64-bit quad-core processor running at 1.5GHz and supports dual displays with a 4K resolution that refreshes at up to 60fps; it has up to 4GB of RAM, making it small in size and high in performance. The main control unit 1 connects to the camera 3 via a USB interface to acquire image information from the recognition area; it connects to the AI ​​image analysis and processing unit 2 via an Ethernet interface to acquire image information processed by the AI ​​image analysis and processing unit 2; it connects to the human-computer interaction unit 4 via a USB interface and an HDMI to micro-HDMI interface for screen display information and human-computer interaction; and it connects to the recycling management unit 6 and the recycling processing unit 5 via a USB interface for processing and managing recycling information.

[0066] Figure 4 This invention demonstrates an improved image acquisition unit. The image acquisition unit includes a low-power event capture unit and a camera, wherein the low-power event capture unit comprises an SNN processor and an event camera. The event camera is a novel image sensor, unlike traditional image frame sensors. Each pixel of this sensor operates independently, independently sensing corresponding changes in light intensity and emitting a corresponding pulse event when the light intensity changes. In most cases where no user collects trash, there are no image changes within the field of view, and therefore no image data (events) are generated.

[0067] The pulse events generated by the camera are fed into a Spike Neural Network (SNN) processor for processing, yielding the corresponding results. When there is no target object, the SNN processor outputs sparse, irregular pulse events. Only when a target object is present does it output a large number of continuous pulse events.

[0068] When an object is detected, such as a common beverage bottle, the spike neurons corresponding to that category fire a large number of pulse events, thereby generating a trigger signal for the low-power event processing unit.

[0069] Upon receiving a trigger signal, the control module typically wakes up the camera to capture the current image data (image frame) and sends the captured image to the AI ​​image analysis and processing unit. The AI ​​image analysis and processing unit then sends the accurate recognition results to the control unit.

[0070] However, during the actual development process, the inventors discovered that if the control unit immediately triggers the camera to capture an image after receiving the trigger signal, the recognition accuracy would always be far lower than expected (the simulation performance of the artificial neural network model in the server).

[0071] refer to Figure 5 After repeated research and testing, the inventors discovered that the control unit should not immediately prompt the camera to capture an image after receiving the trigger signal. If the image is captured immediately, it is not captured in the optimal window, and the captured image is mostly of the beverage bottle still at a tilted angle. This is significantly different from the data used by the AI ​​image analysis and processing unit when training the network.

[0072] For example, as samples in the training data, the visual information of beverage bottles is basically as follows: Figure 6 As shown, the bottle is directly facing the camera. Therefore, the AI ​​image analysis and processing unit trained on this dataset has poor recognition accuracy when the image captured by the camera is still at an angle. The aforementioned existing intelligent recycling systems, because they process dynamic spatial information based on various random variables introduced by the user, generally suffer from the problem of low recognition accuracy for non-frontal input image data (i.e., the aforementioned accuracy bottleneck), and this problem has generally not been well resolved.

[0073] The aforementioned findings, resulting from the application of event cameras and SNN processors to capture events, led the inventors to discover, in order to address the low recognition accuracy issue of artificial neural networks, that the optimal window for RGB image capture should appear only after the first network decision window ends, i.e., after a certain delay following the trigger signal. While this result contradicts conventional experience, numerous tests have consistently shown this pattern. RGB images captured within this optimal RGB image capture window, after being processed by the AI ​​image analysis unit, achieve higher recognition accuracy. Furthermore, this delay cannot be too long, otherwise it is easy to enter an interference window (i.e., a small peak in the number of subsequent pixel activations), such as interference from a human hand or the bouncing of a beverage bottle after impact.

[0074] If the number of active pixels of the corresponding event camera at the end of the first network decision window is denoted as the first number, and the number of active pixels of the corresponding event camera at the first delay after the end of the first network decision window is the second number, then the first number is less than the second number. Furthermore, within the interference window after the first delay after the end of the first network decision window, the maximum number of active pixels of the event camera is also less than the second number.

[0075] Specifically, still refer to Figure 5 Due to manufacturing defects or lighting conditions, the event camera will consistently have background noise, continuously outputting noisy events to the SNN. During the process of a user throwing a beverage bottle into the smart recycling system, the number of activated pixels on the event camera increases rapidly within a short timeframe. The first network decision window appears randomly, and the SNN processor generates a trigger signal at the end of this window. However, it cannot immediately trigger the camera to capture the beverage bottle; otherwise, it will suffer from a consistently lower-than-expected recognition accuracy.

[0076] However, after a preset first delay (test results show that a first delay within the range of 70-90ms is optimal), the image quality captured by the camera is at its best, and this is also when the number of activated pixels (not the number of pulses, because some hot-pixels emit a large number of pulses, but all from the same pixel) is at its highest. This overcomes the aforementioned shortcomings, and the solution also overcomes the shortcomings of existing technologies where low recognition accuracy is caused by randomness in certain situations. In other words, this invention not only provides a low-power intelligent recycling solution, overcoming the aforementioned persistent low-precision problem, but also unexpectedly eliminates the accuracy bottleneck defect commonly found in existing technologies due to randomness.

[0077] refer to Figure 7 It explained Figure 5The SNN processor generates pulses from the spike neurons in each readout layer during a recognition process. If the beverage bottle submitted by the user falls into category 4, the spike neurons corresponding to category 4 will generate a large number of pulses in a short period. However, the first network decision window often arrives before the optimal window for capturing the RGB image, and a trigger signal is generated at the end of this first network decision window. After a first delay, the trigger signal is sent to the control unit, which can then capture the highest quality image within the aforementioned optimal window for subsequent recognition by the AI ​​image analysis and processing unit. In another alternative embodiment, after receiving the trigger signal from the SNN processor, the control unit triggers the camera to capture the image after a first delay.

[0078] Figure 8 The diagram below shows the physical structure of an automatic sorting, identification, and recycling system and device for waste beverage bottles according to one embodiment of the present invention. It includes a beverage bottle recycling bin 1, an STM32 development board 2, a stepper motor 3, a camera 4, a fill light 5, an infrared sensor 6, a small switch 7, a main control switch 8, a Raspberry Pi terminal 9, a bin shell 10, a QR code 11, a speaker 12, a beverage bottle identification area 13, a screen display area 14, and EdgeBoard AI vision hardware 15. The upper edge of the beverage bottle recycling bin 1 is connected to the infrared sensor 6. The speaker 12, EdgeBoard AI vision hardware 15, screen display area 14, STM32 development board 2, and Raspberry Pi terminal 9 are connected. The fill light 5 and stepper motor 3 are connected to STM32 development board 2. The main control switch 8 controls the power supply of the entire device. The small switch 7 provides a Raspberry Pi interface.

[0079] After the user logs into the mini-program, the recognition function is automatically activated. Based on the image information captured by the camera 4, and the recognition results returned by the EdgeBoardAI vision hardware 15, the function intelligently distinguishes the types of beverage bottles. The information is then uploaded to the cloud via the Raspberry Pi terminal 9. Users can view the relevant information on their terminal devices and receive a cash reward based on the beverage bottle identified.

[0080] Figure 9 This diagram illustrates the workflow of a self-service recycling terminal. The terminal uses a Raspberry Pi as the main controller, PyQt5 for the user interface, and Python for the user interaction logic. The user interface includes a QR code scanning interface, a main interface, a user information interface, a category query interface, and a recognition interface; the interface is simple and easy to use, facilitating user adoption. It communicates with a WeChat mini-program via MQTT to receive user commands. Data transmission to the sorting results is performed via an Ethernet interface and Edgeboard.

[0081] On the user interface, users can access the server to view their personal information and the types of items that can be recycled (this can also be viewed within the WeChat app). After the user activates the recognition system, the system automatically acquires camera data and uses computer vision technology to determine if a beverage bottle has been placed in the image. If so, the system automatically starts the recognition process and transmits the image data to the EdgeBoard via Ethernet. Simultaneously, it accesses the server to retrieve the recycling information set by the administrator. After receiving the classification results from the EdgeBoard, the Raspberry Pi checks if the result falls within the recycling range set by the administrator. If it does, the servo motor is activated to perform the recycling; otherwise, the user is notified. If the recycling is successful, the recycling record is saved to the server database, and the user receives a certain amount of cash and points as a reward (set by the administrator).

[0082] Figure 10 This diagram illustrates a system capacity detection scheme based on ultrasound. The STM32 ultrasonic capacity monitoring system uses an STM32F103C8T6 as the main control chip. It monitors the capacity of the recycling bin via an HC-SR04 ultrasonic ranging module and sends the monitoring data to a Raspberry Pi every 5 seconds via serial port for data processing. Ultrasonic modules 1 and 2 are deployed on one half and the top of the bin, respectively. When the distance measured by ultrasonic module 1 is less than 15cm, the recycling bin capacity is half; otherwise, it is considered low capacity. When the distance measured by ultrasonic ranging module 2 is less than 15cm, the device capacity has reached its limit, at which point the device is disabled and the administrator is notified. When the distance measured by ultrasonic ranging module 2 is greater than 15cm, the device is restarted. The Raspberry Pi receives capacity data, processes the data, disables or enables the device, and saves the data to a server for administrator viewing.

[0083] Figure 11 The diagram illustrates a web-based remote management system. Administrators log in to the web management system using their account and password. After logging in, they can view a recycling overview on the homepage. The overview summarizes recycling data for the past week and generates visual reports, comparing the data with the previous week's data to help administrators monitor recycling progress and provide better data support for pricing. On the recycling management page, administrators can modify the recycling price for beverage bottles, appropriately lowering prices for bottles with high recycling volumes and raising prices for others to maximize revenue. On the device management page, administrators can view their owned device information, check device capacity, prepare to add devices (each device is uniquely identified by a device ID, which facilitates identification and management), and enable / disable devices (by sending enable / disable commands to specified devices via the MQTT protocol).

[0084] Figure 12This is a diagram illustrating the functions of the WeChat client. On the WeChat mini-program side, after logging in, the homepage displays three modules: a recycling module, a user information module, a history module, and a category overview module.

[0085] The recycling module is divided into three parts. The scanning part is responsible for opening the camera to scan the QR code on the device. After successful scanning, it will prompt the user whether to open the device. If the user opens the device, the user's OpenId will be sent to the Raspberry Pi via the MQTT protocol; otherwise, it will return. When the recycling volume is large, the administrator can be contacted directly for recycling.

[0086] The user information module accesses the database on the server to query user data (balance, points, etc.) and displays this data. Users can choose to withdraw their balance or use their points to redeem gifts in the points mall. If a user wants to purchase equipment, they can also click to join and contact the administrator. The history module displays the user's identification records (type of beverage bottle, rewards received, and time information), and the call history is a record of user calls to the administrator. The category overview module allows users to quickly view the types of items that can be recycled and their pricing information.

[0087] In summary, this invention is an automatic sorting, identification, and recycling system and device for waste beverage bottles. By combining AI vision technology with front-end and back-end technologies and using cash rewards, it improves the recycling efficiency of waste beverage bottles.

[0088] The order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments; moreover, the above description focuses on specific embodiments, while other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0089] The foregoing description has fully disclosed the specific embodiments of the present invention. It should be noted that any modifications made to the specific embodiments of the present invention by those skilled in the art do not depart from the scope of the claims. Accordingly, the scope of the claims is not limited to the foregoing specific embodiments.

Claims

1. An automatic sorting, identification, and recycling device for waste beverage bottles, used for automatically identifying and recycling beverage bottle types, characterized in that, The automatic sorting, identification, and recycling device includes: an image acquisition unit, which comprises a low-power event capture unit and a camera; wherein, The low-power event capture unit includes an event camera and an SNN processor, which performs inference based on the event camera's perception of the environment. When the event camera detects a beverage bottle being delivered within its field of view, the SNN processor generates a trigger signal based on the pulse event output by the event camera. The trigger signal is generated by the SNN processor at the end of the first network decision window, and at the end of the first network decision window, the number of activated pixels of the corresponding event camera is a first number. The control unit is configured to, upon receiving a trigger signal, control the camera to delay capturing the current image after a first delay; wherein, at the first delay after the end of the first network decision window, the number of active pixels of the corresponding event camera is a second number, and the first number is less than the second number; The image classification unit is an AI image analysis and processing unit configured to classify images captured by the camera and obtain classification results. The identification result processing unit is configured to analyze and process the classification results of beverage bottles, and analyze whether the type of beverage bottle is one of the types that need to be recycled as set by the administrator: if so, it is recycled; if not, the user is reminded that the beverage bottle is not a type to be recycled. The human-computer interaction unit is used by users to view classification results and select modes on the screen; The recycling management unit includes a microcontroller and two servo motors. The microcontroller drives the servo motors to open and close the recycling bin according to instructions. The recycling unit is an automatic control unit based on infrared technology, which includes a probe that outputs a high level when the capacity of the beverage bottle reaches its sensing range. The main control unit connects to a camera via a USB interface to acquire image information from the recognition area, and connects to an AI image analysis and processing unit via an Ethernet interface to acquire image information processed by the AI ​​image analysis and processing unit. It also connects to a human-computer interaction unit for screen display and human-computer interaction, and connects to a recycling management unit and recycling processing unit via a USB interface for processing and managing recycling information; and... When the event camera detects a beverage bottle being delivered within its field of view, the SNN processor generates a trigger signal based on the input pulse event output by the event camera, after a first delay; wherein the first delay is between 70 and 90 milliseconds.

2. The automatic sorting, identification, and recycling device for waste beverage bottles according to claim 1, characterized in that: The recognition model training unit trains corresponding recognition models based on images and types of beverage bottles, and then deploys the trained recognition models on the image classification unit.

3. The automatic sorting, identification, and recycling device for waste beverage bottles according to claim 2, characterized in that: The automatic sorting, identification, and recycling device also includes: The mode selection unit allows the user to choose between manual or automatic recognition mode after scanning the code and initiating the recognition command. In manual mode, users need to manually press the recognition button to recognize beverage bottles; In automatic identification mode, the system will automatically identify the type of beverage bottle; The image classification unit sets the mode according to the mode selected by the mode selection unit, and classifies the types of beverage bottles according to the recognition model deployed by the image acquisition unit and the recognition model training unit.

4. The automatic sorting, identification, and recycling device for waste beverage bottles according to claim 3, characterized in that: The automatic sorting, identification, and recycling device also includes: The automatic sorting, identification, and recycling device is also configured to: query the price information set by the administrator in the web management unit, and estimate the value of the recycled beverage bottles based on the sorting results and the price information set by the administrator.

5. The automatic sorting, identification, and recycling device for waste beverage bottles according to claim 4, characterized in that: The automatic recycling unit uses hydraulic technology. After determining that the beverage bottle is one that needs to be recycled, it will activate the automatic recycling device to compress the volume of the beverage bottle and store it in the recycling bin. The recycling management unit automatically senses the current capacity of the recycling bin through an infrared sensor module installed on the recycling bin. Administrators can view the status of the recycling bin through a terminal device and determine whether the recycling bin has reached its maximum capacity based on the status of the recycling bin. as well as, The cloud-based unit uploads the user's identification information and the amount collected to the cloud server after the automatic recycling process ends, and provides a data access interface through backend technology.

6. The automatic sorting, identification, and recycling device for waste beverage bottles according to claim 5, characterized in that: The automatic sorting, identification, and recycling device is configured such that after a user logs in on the WeChat mini-program terminal, the device accesses the user's identification data and identification amount through backend technology and server, displays the user's information and identification amount using frontend technology, and the user can withdraw the identified amount through WeChat.

Images