Automatic AI-based trash sorting bin that integrates image, sound, and weight information

Abstract

We provide a trash can device that automatically separates trash into designated sorting bins by integrating and assessing image, sound, and weight information using AI. The system includes a garbage input detection sensor (2), a photography camera (1), lighting means (3), a weight measurement sensor (16), a sound acquisition means (14), a turntable (5), a gate servo motor (7), and a control computer (10). The control computer uses AI to evaluate images and sounds, score weight measurements, calculate reliability, and dynamically adjust the weighting of each score to improve sorting accuracy. This configuration enables highly accurate and stable automatic sorting of similar garbage, which was previously difficult, contributing to the efficiency of waste disposal at public facilities and other facilities. Previous technology struggled to accurately classify similar garbage using image recognition alone, resulting in a high risk of misidentification and sorting errors. This was particularly difficult when distinguishing between glass bottles, aluminum cans, steel cans, white paper trays, and plastic trays, which have similar shapes and materials.

Description

[Technical Field]

[0001] This invention relates to a trash can equipped with a mechanism that uses sound, weight, and images to automatically and comprehensively identify trash using AI and separate it with high accuracy. [Background technology]

[0002] At home, in the office, at events, and elsewhere, there is an increasing need to properly sort waste in accordance with recycling and local government waste collection rules.

[0003] Traditionally, waste sorting has often been done manually, which requires a great deal of effort, especially at event venues and large facilities, and the tedious process has led to problems such as littering. Furthermore, manual sorting is prone to errors and contamination.

[0004] AI-based trash sorting bins using images have already been proposed, but in recent years, the amount of trash that differs not only in appearance but also in shape, size, and material has increased, and there are limits to how much can be distinguished based on images alone.

[0005] For example, there are glass bottles and plastic bottles that are similar in shape, aluminum cans and steel cans that have similar appearances and sizes, and white plastic trays and white paper trays, and it is difficult to accurately distinguish between them based on images alone.

[0006] Prior art (Patent Documents 1 and 2, etc.) proposes automatic sorting using multiple sensors, but does not disclose a specific configuration for integrating images, sound, and weight to make a comprehensive judgment using AI.

[0007] Conventional automatic trash sorting bins rely solely on images to make judgments, making it difficult to distinguish between trash that looks similar, and resulting in insufficient sorting accuracy.

[0008] To solve this problem, this invention integrates the analysis of images, sound, and weight as AI judgment criteria, making the image judgment accuracy less susceptible to the influence of lighting conditions and the shooting environment, achieving high-precision and stable sorting. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] US Patent Publication No. US20180016096A1 (Automatic sorting of waste) URL: https: / / patents.google.com / patent / US20180016096A1 / en [Patent Document 2] US Patent Publication No. US20210188541A1 (Smart waste bin sensor device and methods) URL: https: / / patents.google.com / patent / US20210188541A1 / en [Non-patent literature]

[0010] [Non-Patent Document 1] Bin-e's "Smart Waste Bin" product introduction page URL: https: / / bine.world / [Non-patent document 2] (Smart Recycling Bin TrashBot Sorts Waste While Delivering Feedback and Education to Users), PR Newswire, February 16, 2023 URL: https: / / www.prnewswire.com / news-releases / smart-recycling-bin-trashbot-sorts-waste-while-delivering-feedback-and-education-to-users-301748231.html Summary of the Invention [Problem to be solved by the invention]

[0011] The purpose of this invention is to automatically separate similar garbage stably and with high accuracy by using AI to comprehensively analyze image information, sound information, and weight information. [Means for solving the problem]

[0012] This device uses an ultrasonic sensor to detect when trash is dropped in, a camera to capture images using lighting such as LEDs, a weight sensor to measure the weight, and a microphone to record the sound generated by the trash. All of this image, sound, and weight information is input into a control computer such as a Raspberry Pi, which then scores and evaluates the trash using AI. Based on the evaluation results, the turntable servo motor and gate servo motor are driven to automatically sort the trash into the designated sorting bins. [Effects of the Invention]

[0013] This invention enables highly accurate and stable classification of similar types of garbage, which was difficult to distinguish using conventional technology. Furthermore, the combined analysis of images, sound, and weight makes it less susceptible to the effects of lighting conditions and the environment, reducing the chance of misclassification. In addition, by streamlining sorting work, it can be effectively used in a variety of places, including homes, offices, public facilities, and event venues. This reduces the burden of manual sorting work, improves recycling rates, reduces waste disposal costs, and contributes to environmental conservation. Furthermore, by adding AI learning data, sorting accuracy can be continuously improved. [Brief explanation of the drawings]

[0014] [Figure 1] Isometric view of the entire device [Figure 2] Diagram of the ceiling inside the device [Figure 3] View from ceiling to floor and range of motion of sorting ramp [Figure 4] Front view of the device [Figure 5] Side view of sorting ramp [Figure 6] Flowchart of input and integration judgment using image, sound, and weight information DETAILED DESCRIPTION OF THE INVENTION

[0015] The garbage sorting device of this invention uses AI to identify the garbage that is put in using image, sound, and weight information, and automatically sorts it into designated sorting bins.

[0016] A control computer such as a Raspberry Pi is used as the control unit. For example, in this embodiment, a Raspberry Pi is used, which automatically runs a Python program at startup and displays "Please put in garbage" on the LCD display.

[0017] When garbage is placed into the sorting box, an ultrasonic sensor detects it, an LED or other light comes on, an image is taken with a camera, and at the same time, a weight sensor measures the weight.

[0018] By capturing the sound of the dust using impacts from a servo motor equipped with a striking rod or other methods, it is possible to make an auxiliary judgment on dust that is difficult to determine using image scores alone.

[0019] The acquired image, sound, and weight data are scored on the Raspberry Pi (control unit), and an overall judgment is made based on weighting according to the reliability of each piece of information. The final garbage type classification result is determined by combining the judgment score and reliability of each piece of information.

[0020] Depending on the result of the judgment, the Raspberry Pi (control unit) operates the turntable servo motor and the gate servo motor, and automatically transports the garbage via the sorting slope to the designated sorting box.

[0021] With this invention, operators can automatically separate similar types of garbage with high accuracy based on an integrated assessment of images, sound, and weight, without having to worry about any special operations when putting garbage in.

[0022] Figure 6 is a flowchart showing the input and integrated judgment flow of the AI ​​model in the garbage sorting device of this invention. The process is carried out in the following order: garbage input detection, image, sound, and weight acquisition, data preprocessing such as normalization and filtering, scoring of each data by AI judgment, score weighting adjustment based on reliability, integrated judgment of garbage type, and sorting operation by servo control. This enables stable automatic sorting that is less affected by environmental conditions or individual judgment accuracy. [Example]

[0023] In this embodiment, a garbage sorting device is created based on the configuration shown in FIGS.

[0024] First, a Raspberry Pi is installed as the control unit, and a Python program is automatically executed at startup, displaying the message "Please put in garbage" on the LCD screen.

[0025] When garbage is dropped in, an ultrasonic sensor detects it, lights up an LED or other light, and a camera captures an image. At the same time, a weight sensor measures the weight of the garbage.

[0026] Next, as an example, a servo motor equipped with a striking rod that generates sound is driven to strike the garbage, obtaining the sound and providing auxiliary judgment when it is difficult to judge from images alone.

[0027] Data preprocessing is performed using normalization, filtering, etc. to make each piece of data suitable for AI judgment.

[0028] The acquired image, sound, and weight information is scored by AI and programs on the Raspberry Pi.

[0029] An integrated judgment is made based on the weighting of scores according to the reliability of each type of garbage.

[0030] Depending on the judgment result, the turntable servo motor and gate servo motor are driven to transport the garbage via the sorting slope to the designated sorting box.

[0031] This allows the operator to automatically separate garbage with a stable and high degree of accuracy, without having to worry about any special operations, by integrating image, sound, and weight judgments to reduce false positives even when the garbage has similar shapes and materials. [Industrial Applicability]

[0032] This device is effective when high-precision garbage sorting is required in homes, offices, event venues, public facilities, etc., and contributes to improving sorting work efficiency and recycling rates. [Explanation of symbols]

[0033] 1 Camera: Takes pictures of the trash 2 Ultrasonic sensor: Detects the distance to the garbage and detects when garbage is thrown in 3. Lighting means: Light source that illuminates the subject when shooting (e.g. LED, etc.) 4 Sorting box: A box for storing sorted garbage 5 Rotating platform servo motor: Adjusts the position of the dust 6 Servo motor for striking rod: Drives the striking rod 7 Gate servo motor: Opens and closes the sorting gate 8 Sorting ramp: A ramp for flushing garbage 9 Microphone: Records the sound of hitting the trash 10 Control unit (e.g., Raspberry Pi): Equipped with an AI model to make a comprehensive judgment 11 LCD display: Displays the judgment result 12 Cover: Enclosure that covers the entire unit 13 Garbage: Target items (aluminum cans, bottles, etc.) 14 Striking stick: Hits the trash to make a sound 15 Garbage gate: Holds garbage and opens and closes 16 Weight sensor: Measures the weight of the garbage

Claims

1. A garbage bin device that automatically separates similar garbage with high accuracy, comprising a garbage insertion detection sensor, a photographing camera, lighting means, a weight measurement sensor, sound acquisition means, a turntable and gate servo motor, and a control computer, and based on the detection of insertion by the garbage insertion detection sensor, acquires image, sound, and weight information, and combines the acquired image, sound, and weight information to separate the garbage into designated sorting bins.

2. 2. The device according to claim 1, wherein the control computer scores each piece of acquired image, sound, and weight information, calculates a reliability score based on the judgment accuracy of each piece of information, and makes an integrated judgment based on this.

3. 3. The apparatus according to claim 2, wherein the weighting based on the reliability score can be dynamically changed depending on conditions, thereby improving the accuracy of the determination.

4. 4. The device according to claim 1, wherein the sound acquisition means is capable of acquiring the sound of the garbage by impact with a striking stick or other method.

Images