Intelligent paid waste recycling system

Abstract

The invention discloses an intelligent paid waste recovery system, which relates to the technical field of waste recovery and comprises an intelligent recovery box main body, a central control module, a user interaction module, a settlement module and a user side applet, the intelligent recycling bin main body comprises a single recycling bin door, an independent identification channel, a sorting execution module and a double-category sorting bin body; the double-category sorting cabin body comprises a paper-plastic cabin body and a fabric cabin body, electric sorting gates are arranged at the tops of the paper-plastic cabin body and the fabric cabin body, and weighing settlement units are arranged at the bottoms of the paper-plastic cabin body and the fabric cabin body; a dangerous goods detection module, a delivery quantity limiting module, an overlapping recognition module, an overlapping goods dispersing module and a self-adaptive illumination category recognition module are sequentially integrated in the independent recognition channel in the goods conveying direction; according to the scheme, identification precision fluctuation can be improved, and a stable and reliable classification effect is formed.

Description

Technical Field

[0001] This invention relates to the field of waste recycling technology, specifically to an intelligent paid waste recycling system. Background Technology

[0002] With the increasing demand for resource recycling, intelligent waste recycling systems have become an important component of urban waste sorting and resource recycling systems. Their core value lies in reducing manual intervention and improving recycling efficiency through automation technology. However, current mainstream intelligent recycling technologies still face a series of common technical bottlenecks in practical applications: In the pre-sorting mode, the recycling system places rigid demands on users' category recognition abilities, leading to a simultaneous increase in sorting operation thresholds and mis-disposal rates; the configuration and operation mechanism of the identification module lack systematic optimization, and factors such as differences in item shape and environmental interference can easily cause fluctuations in identification accuracy, making it difficult to achieve stable and reliable sorting results; the design of the safety detection link has shortcomings, with insufficient pre-interception capabilities for hazardous materials, posing safety hazards to system operation and subsequent processing; at the same time, system architecture design often focuses on single-function implementation, failing to fully consider the synergistic improvement of identification accuracy, operational economy, and user experience. As a result, while some technical solutions can meet basic recycling needs, they are limited in large-scale promotion due to high costs and maintenance difficulties, failing to fully unleash the application value of intelligent recycling technology. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent paid waste recycling system, which solves the problems existing in the background technology.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent paid waste recycling system, comprising: an intelligent recycling bin body, a central control module, a user interaction module, a settlement module, and a user terminal applet;

[0005] The main body of the intelligent recycling bin includes a single recycling compartment door, an independent identification channel, a sorting execution module, and a dual-category sorting compartment; the dual-category sorting compartment includes a paper-plastic compartment and a fabric compartment, both of which are equipped with electric sorting gates at the top and weighing and settlement units at the bottom;

[0006] The independent identification channel integrates, in sequence along the direction of item transport, a dangerous goods detection module, a quantity restriction module, an overlap identification module, an overlap item dispersion module, and an adaptive lighting category identification module.

[0007] The user interaction module is used to guide users to select the category of the items to be delivered and to obtain the target category selected by the user.

[0008] The hazardous materials detection module is used to perform pre-deployment safety screening of the materials;

[0009] The quantity limit module is used to control the quantity of items delivered in a single delivery.

[0010] The overlap recognition module is used to determine the overlap status of the placed items and generate an overlap determination result;

[0011] The overlapping item dispersion module is used to disperse the deployed items using a dispersion structure based on the overlap determination result.

[0012] The adaptive lighting category recognition module is used to identify the actual category of the placed items and generate category recognition results;

[0013] The sorting execution module is used to perform sorting actions based on the category identification results;

[0014] The central control module is used to control the operation of the entire system, receive data from each module, and execute control commands.

[0015] The settlement module is used to settle revenue based on the weighing results from the weighing settlement unit.

[0016] Furthermore, the user interaction module displays category icons for paper and plastic products and textile products, along with corresponding descriptions for each category, on a touchscreen display, and works in conjunction with a voice acquisition unit to recognize specific delivery commands.

[0017] Furthermore, the hazardous materials detection module includes a metal detection unit, a dielectric constant detection unit, and an X-ray fluorescence scanning unit, which are used to detect the metal composition, dielectric constant, and elemental composition characteristics of the placed items, respectively. The hazardous level of the placed items is comprehensively determined by calculating a comprehensive hazardous materials judgment coefficient. If the comprehensive hazardous materials judgment coefficient is greater than or equal to a preset hazardous materials threshold, the item is determined to be a hazardous material, and an alarm is triggered and the warehouse door is opened.

[0018] Furthermore, the delivery quantity restriction module constructs a three-dimensional model of the item using an infrared ranging sensor array to calculate its volume, and uses a weight pre-sensor to collect the item's weight. Based on the target category selected by the user, it calls the differentiated delivery threshold formula for the corresponding category to determine whether the volume or weight exceeds the limit. If the limit is exceeded, the user is prompted to deliver the item in multiple batches. Once the limit is met, the item is allowed to proceed to the overlap recognition module.

[0019] Furthermore, the overlap recognition module includes a high-definition industrial camera and a TOF depth sensor, used to acquire two-dimensional images and three-dimensional coordinate information of the object, respectively, and extract the contour features, area features, and depth features of the placed object. An overlap coefficient is used to determine the degree of overlap of the placed objects. The formula for calculating the overlap coefficient is as follows: ,in, The area of ​​the overlapping region in the object image is obtained through contour intersection calculation; The sum of the total areas of all items. The number of detected items is determined through contour connectivity analysis; Let be the average depth of the i-th and j-th items. The average depth of all items. , These are the weighting coefficients. This is the category correction factor.

[0020] If the overlap coefficient of the placed items is greater than the preset overlap judgment threshold, it is judged as significant overlap, and the central control module starts the overlapping item dispersion module; otherwise, it directly enters the adaptive lighting category recognition module.

[0021] Furthermore, the overlapping item dispersion module adapts the corresponding dispersion structure according to the target category, and constructs a dispersion parameter adjustment model based on the target category and the overlap coefficient to adjust the dispersion parameters of the dispersion structure and disperse the items that are determined to be significantly overlapping.

[0022] Furthermore, the adaptive lighting category recognition module collects the light intensity within the channel through an ambient light sensor. If the light intensity exceeds the optimal range, it adjusts the camera exposure time and gain, adjusts the brightness of the LED fill light array, and extracts effective features based on the characteristic differences between paper-plastic and fabric categories using a differentiated feature fusion model. These features are then input into an improved SVM classifier to complete category recognition, and the category recognition results are fed back to the central control module.

[0023] Furthermore, the sorting execution module is used to receive the category matching signal from the central control module. If the actual category matches the user's selection, it controls the main gate and the corresponding category compartment's sub-gate to open sequentially, guiding the items to slide into the corresponding category compartment via the guide slide. If the categories do not match, the gate remains closed, waiting for the user to adjust their selection before executing the sorting action.

[0024] Furthermore, the settlement module is used to obtain the actual weight data of the weighing and settlement unit after the items are delivered to the target compartment. Based on the recycling unit price of the corresponding category, the revenue is calculated using the revenue calculation formula and automatically credited to the user's mini-program wallet, allowing users to withdraw at any time, while simultaneously retaining the delivery and settlement records.

[0025] This invention provides an intelligent paid waste recycling system, which has the following beneficial effects:

[0026] This solution utilizes a combination of single-door and dual-category selection, simplifying interaction through graphic guidance and voice assistance. The dual-category division aligns with users' daily understanding, making it easier for new users to get started and reducing mis-disposal rates, significantly improving classification accuracy compared to traditional systems. A unified, independent identification channel replaces the dispersed identification modules across multiple doors, and the dual-compartment structure reduces manufacturing costs compared to traditional multi-compartment designs. Simultaneously, controlled disposal volume and overlapping handling reduce subsequent maintenance losses, lowering overall operating costs. The overlapping identification module accurately determines the overlapping status, and a category-differentiated dispersion mechanism eliminates stacking or entanglement. An adaptive lighting compensation algorithm stabilizes the identification environment, improving the accuracy of both categories. The error rate is further reduced after overlapping handling, enhancing identification precision and stability. A composite detection system employing metal detection, dielectric constant detection, and X-ray scanning, using quantified judgment coefficients, enables precise interception of hazardous materials, improving detection accuracy and preventing hazardous materials from entering the recycling process. Differentiated volume or weight thresholds are set based on the physical characteristics of paper-plastic and fabric materials, reducing identification errors caused by excessive disposal at the source and improving the system's continuous operational efficiency. Attached Figure Description

[0027] Figure 1 This is a block diagram of an intelligent paid waste recycling system according to the present invention; Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0029] Example 1

[0030] Please see Figure 1 This invention provides an intelligent paid waste recycling system, comprising:

[0031] The smart recycling bin consists of a main body, a central control module, a user interaction module, a settlement module, and a user-end mini-program.

[0032] The main body of the intelligent recycling bin includes a single recycling compartment door, an independent identification channel, a sorting execution module, and a dual-category sorting compartment; the dual-category sorting compartment includes a paper-plastic compartment and a fabric compartment, both of which are equipped with electric sorting gates at the top and weighing and settlement units at the bottom;

[0033] The independent identification channel integrates, in sequence along the direction of item transport, a dangerous goods detection module, a quantity restriction module, an overlap identification module, an overlap item dispersion module, and an adaptive lighting category identification module.

[0034] The user interaction module is used to guide users to select the category of the items to be delivered and to obtain the target category selected by the user.

[0035] The hazardous materials detection module is used to perform pre-deployment safety screening of the materials;

[0036] The quantity limit module is used to control the quantity of items delivered in a single delivery.

[0037] The overlap recognition module is used to determine the overlap status of the placed items and generate an overlap determination result;

[0038] The overlapping item dispersion module is used to disperse the deployed items using a dispersion structure based on the overlap determination result.

[0039] The adaptive lighting category recognition module is used to identify the actual category of the placed items and generate category recognition results;

[0040] The sorting execution module is used to perform sorting actions based on the category identification results;

[0041] The central control module is used to control the operation of the entire system, receive data from each module, and execute control commands.

[0042] The settlement module is used to settle revenue based on the weighing results from the weighing settlement unit.

[0043] In this embodiment, the main body of the smart recycling bin adopts a three-level series structure of "single recycling bin door - independent identification channel - dual-category sorting compartment". The dual-category sorting compartment includes a paper and plastic compartment and a fabric compartment. Both types of compartments are equipped with electric sorting gates at the top and weighing and settlement units at the bottom.

[0044] The paper-plastic bin is used to recycle paper and plastic waste, specifically including cardboard boxes, waste paper, cartons, books, newspapers, plastic bottles, plastic cups, plastic buckets, and plastic toys. The bin has a volume of 80L and its inner wall is coated with a wear-resistant polyethylene coating. The fabric bin is used to recycle fabric and footwear waste, specifically including handbags, down jackets, sweaters, T-shirts, plush toys, and footwear. The bin has a volume of 120L and its inner wall is coated with an anti-snagging nylon coating. The single recycling bin door is the only item loading entrance, equipped with an electromagnetic drive device and an anti-pinch sensor, with an opening angle of 90° and a response time of ≤0.2s.

[0045] The system, with a central control module at its core, executes the following five-level processing flow:

[0046] The interactive selection process includes the user selecting the target category K of the item to be delivered through the user interaction module. K=1 represents paper and plastic, K=2 represents fabric, and the central control module records the target category K and sends an opening command to the door control unit.

[0047] The safety inspection process includes the following steps: after the recycling bin door is opened, the user puts the item into the independent identification channel, the dangerous goods detection module starts composite detection, calculates the comprehensive judgment coefficient D of dangerous goods, if D≥0.8, it is judged as a dangerous goods, the central control module triggers an alarm and keeps the bin door open for the user to retrieve the item; if D<0.8, it proceeds to the next step.

[0048] The quantity control screening process includes a delivery quantity limit module that matches the corresponding volume threshold V(K) and weight threshold m(K) according to the target category K, and detects the item volume V and weight m in real time. If V>V(K) or m>m(K), the user is prompted to deliver the item in multiple batches; if both conditions are met, the process proceeds to the next step.

[0049] The overlap processing and recognition process includes the overlap recognition module detecting the overlap coefficient O of the items. If O > 0.25 for paper-plastic items or O > 0.35 for fabric items, the overlap item dispersion module is activated to perform adaptive dispersion processing. After dispersion, the adaptive lighting category recognition module completes the recognition of the actual category k of the item, where k=1 represents paper-plastic items and k=2 represents fabric items. If dispersion is not required, the recognition operation is performed directly.

[0050] The matching, sorting, and settlement process includes a central control module comparing k and K. If k=K, the electric sorting gate of the corresponding category compartment is opened, and the item slides into the compartment. The actual weight m is measured by the weighing and settlement unit, and the revenue is calculated based on the recycling unit price and synchronized to the user's mini-program. If k≠K, the prompt module is activated to guide the user to change the target category K'. After the change, the overlap processing and identification process is re-executed. If the user refuses to change, a timeout alarm is triggered.

[0051] The user interaction module consists of a touch screen, a voice acquisition unit, and a voice broadcasting unit. It is installed 15cm outside the recycling bin door at a height of 1.2m to accommodate adult operation. The touch screen has a resolution of 1920×1080. In standby mode, it displays the prompt "Please select the type of item to deposit" and two category icons: the "Paper & Plastic" icon shows images of cardboard boxes and plastic bottles along with the text "Includes paper and plastic waste"; the "Textiles" icon shows images of sweaters and sneakers along with the text "Includes clothing and footwear waste". The voice acquisition unit uses a microphone array and supports specific commands such as depositing paper / plastic waste and old clothes waste, with an accuracy rate of ≥95%. The voice broadcasting unit uses an 8Ω / 5W speaker, and the broadcast volume can be automatically adjusted according to ambient noise (30dB-70dB).

[0052] This module has a built-in user delivery record database that automatically stores the product categories delivered by the user in the last 10 deliveries. It places high-frequency product categories in the priority display position on the left side of the touch screen and sets up a "repeated delivery" touch button. Clicking the button will directly trigger the warehouse door opening command for the corresponding product category, shortening the interaction time.

[0053] The hazardous materials detection module integrates a metal detection unit, a dielectric constant detection unit, and an X-ray fluorescence scanning unit, arranged sequentially along the entrance of the independent identification channel. The detection coverage area is the cross-section of the channel (30cm×30cm), and the response time is ≤0.5s.

[0054] The metal detection unit uses a high-frequency electromagnetic induction coil with a working frequency of 100kHz. It can identify metal blocks with a volume ≥0.5cm³ and outputs a metal detection coefficient D: D=1 when the metal volume is ≥5cm³; D=0.3 when the metal volume is ∈[0.5cm³,5cm³); and D=0 when there is no metal.

[0055] The dielectric constant detection unit uses a 2.4GHz microwave sensor to calculate the dielectric constant ε of the item by the phase difference between the transmitted and received microwave signals, and outputs the dielectric constant coefficient D: when ε∈[1.5,3.5] (the dielectric constant range of conventional waste products), D=0; when ε<1.5 or ε>3.5, D=|ε-2.5| / 2.5, with a maximum value of 1.

[0056] The X-ray fluorescence scanning unit uses a low-dose X-ray source (tube voltage 50kV, tube current 0.5mA) to rapidly analyze the elemental composition of items. It can identify hazardous elements such as lead, mercury, lithium, and potassium, as well as their corresponding compounds. The output X-ray detection coefficient D is: D=1 when a hazardous element is detected, otherwise D=0.

[0057] The formula for calculating the comprehensive judgment coefficient D of hazardous materials is: ,in, , , The weighting coefficients are 0.4, 0.3, and 0.3, respectively, set according to the priority of hazard levels. Sharp metal objects pose the highest risk, followed by flammable and explosive materials, and then toxic substances, with the sum of the weighting coefficients being 1. When D ≥ 0.8, the central control module immediately controls the recycling bin door to open in the reverse direction and remain open. The touchscreen displays a red warning image and text: "Hazardous material detected. Do not dispose of. Please retrieve immediately." The voice broadcast unit repeatedly broadcasts "Hazardous material, please retrieve" at a volume of 70dB. Simultaneously, the user-end mini-program pushes a warning message including the detection time and an image of the item.

[0058] The quantity control module consists of an infrared ranging sensor array and a weight pre-sensor, installed 5cm inside the entrance of the independent identification channel. The infrared ranging sensor array contains 6 sensors arranged in a 3×2 matrix on the inner wall of the channel. It uses the triangulation principle, with a measurement range of 0.1m-1m and an accuracy of ±1mm. It calculates the volume V by constructing a three-dimensional spatial model of the item, with a measurement error ≤5%. The weight pre-sensor is a pressure strain gauge sensor with a range of 0-5kg and an accuracy of ±0.01kg. It is installed on the load-bearing plate at the bottom of the channel to collect the weight m of the item in real time.

[0059] Based on the differences in physical properties between paper-plastic materials (high density, stable shape) and fabric materials (low density, easily fluffy), a differentiated dosage threshold mechanism is adopted. The threshold calculation formula is as follows: Where K∈{1,2}, K=1 represents paper-plastic products, and K=2 represents fabric products; The maximum single-dispensing volume for product category K is expressed in liters (L). The maximum single delivery weight for category K is in kilograms (kg). Based on 100 sets of experiments, the following values ​​were set: paper-plastic type m(1) = 2kg (high density, strong load-bearing capacity), and fabric type m(2) = 3kg (fluffy but with a high upper limit on weight). 1.3 represents the minimum density of category K, expressed in kilograms per liter (kg / L). Experiments have shown that the minimum density for paper-plastic products (a mixture of cardboard boxes and plastic bottles) is 0.4 kg / L, and the minimum density for fabric products (a mixture of sweaters and plush toys) is 0.15 kg / L. 1.3 is the volume redundancy coefficient, used to compensate for volume calculation deviations caused by the irregular shape of paper-plastic products (such as cardboard boxes) and the fluffy characteristics of fabric products.

[0060] Substituting the parameters, the calculation yields: V(1) = 2 / 0.4 × 1.3 = 6.5L for paper and plastic products, and V(2) = 3 / 0.15 × 1.3 = 26L for fabric products. When V > V(K) or m > m(K) is detected, the system pauses the process, and the touch screen displays a yellow prompt message: "Single delivery amount is too large, please deliver in multiple deliveries (Current category: XX category, maximum single delivery: weight XXkg / volume XXL)". The voice broadcast unit simultaneously broadcasts the prompt message, and the recycling bin door remains open for the user to take away some items. After the user adjusts, the system re-executes the detection.

[0061] The overlap recognition module consists of a high-definition industrial camera and a TOF (Time of Flight) depth sensor, installed in the middle of the independent recognition channel, 10cm away from the quantity restriction module. The high-definition industrial camera has a resolution of 1920×1080, a frame rate of 30fps, and a lens focal length of 8mm, acquiring two-dimensional images of the object; the TOF depth sensor has a measurement range of 0.1m-5m and an accuracy of ±1mm, acquiring three-dimensional coordinate information of the object and outputting a depth image.

[0062] After image preprocessing (including Gaussian filtering for noise reduction, Otsu binarization, and edge detection), three types of core features are extracted: for paper and plastic items, the focus is on extracting contour hardness features (edge ​​gradient value ≥ 50) and area features; for fabric items, the focus is on extracting depth distribution features (depth value variance ≥ 10 mm). The degree of overlap between items is quantified using an overlap coefficient O, calculated using the following formula: ,in, The unit is square centimeters (cm²), obtained through contour intersection operations; The sum of the total areas of all items is expressed in cm²; n is the number of detected items, determined by contour connectivity analysis; H and H' are the mean depths of the i-th and j-th items, respectively, expressed in millimeters (mm). Let be the average depth of the i-th and j-th items. This is the average depth of all items, in mm. , These are the weighting coefficients. =0.6、 =0.4, balancing the two features of area overlap and depth difference, focusing on area overlap for paper and plastic, and focusing on depth difference for fabric; ζ(K) is the category correction coefficient, when K=1 (paper and plastic) ζ=0.05 (increasing the sensitivity of overlap judgment, because overlap of paper and plastic is prone to recognition error), when K=2 (fabric) ζ=-0.05 (reducing the judgment sensitivity, because slight overlap of fabric does not affect recognition).

[0063] The overlap determination threshold is set according to the characteristics of the product category: for paper and plastic products, O > 0.25 is considered to be significantly overlapping, and for fabric products, O > 0.35 is considered to be significantly overlapping. At this time, the central control module starts the overlapping product dispersion module; otherwise, it directly enters the adaptive lighting category recognition module.

[0064] The dispersion system adopts a dual-factor adaptive adjustment mechanism based on the characteristics of the product category and the overlap state. Based on the physical properties of the target recycled product category and the real-time detected overlap coefficient, it dynamically matches dispersion parameters (paddle speed, air pressure, vibration frequency) to ensure that the dispersion force is adapted to the tolerance of the items. The higher the overlap, the greater the dispersion force. At the same time, it avoids damage to the items by using category-differentiated safety thresholds, thus achieving efficient and gentle overlap separation.

[0065] In this embodiment, the overlapping item dispersion module adopts a composite dispersion structure of rotating paddles, airflow assistance, and vibration assistance, and is installed in the middle of the independent identification channel, adjacent to the overlapping identification module. Three sets of rotating paddles are circumferentially arranged on the inner wall of the channel. The paddles on the paper / plastic side are made of hard rubber with a Shore hardness of 70°, while those on the fabric side are made of flexible silicone with a Shore hardness of 30° to prevent snagging. Four air nozzles are symmetrically installed at the top and bottom of the channel. The air nozzles for paper / plastic are direct-fire type (airflow diffusion angle ≤ 15°), and those for fabric are diffusion type (airflow diffusion angle ≥ 45°). A small vibration unit is integrated into the bottom load-bearing plate of the channel, which is activated only when dispersing fabric items.

[0066] The dispersion parameters are adaptively adjusted based on the target product category K and the overlap coefficient O. During dispersion, the overlap recognition module detects the O value in real time, and dispersion is immediately stopped when the safety threshold is reached (O≤0.25 for paper and plastic products, O≤0.35 for fabric products). The dispersion parameter adjustment model includes the following dispersion parameters: paddle rotation speed. The linear velocity of the pick as it combs an item is expressed in r / min; the air pressure of the nozzle is also measured. The pressure intensity characterizing airflow-assisted separation, in kPa; vibration frequency. , representing the frequency of vibration of loose items on the pallet, in Hz, where vibration frequency is only enabled for fabric items. Specifically, the expression for the dispersion parameter adjustment model is as follows: (O) = (1 + O), (O) = (1 + O), (O) = (1 + O);

[0067] When K=1 (for paper-plastic products), the expression for the dispersion parameter adjustment model is: (O) = (1 + O), (O) = (1 + O);

[0068] When K=2 (for fabrics), the expression for the dispersion parameter adjustment model is: (O) = (1 + O), (O) = (1 + O), (O) = (1 + O);

[0069] In the formula, O is the real-time overlap coefficient (dimensionless, 0≤O≤1), which is detected in real time by the overlap recognition module; The standard shift speed (r / min) is used for this product category. > Hard materials can withstand higher combing speeds; The reference pressure (kPa) for this product category is the valve pressure. > Hard materials can withstand higher airflow pressure; The reference vibration frequency (Hz) for fabrics; The paddle speed adjustment coefficient (dimensionless, 0.3≤kᵥ≤0.6) controls the rate at which the speed increases with the degree of overlap. The air pressure adjustment coefficient (dimensionless, 0.5≤k_P≤0.8) controls the rate of increase of air pressure with the degree of overlap. The vibration frequency adjustment coefficient (dimensionless, 0.2 ≤ k_f ≤ 0.4) controls the rate of increase of the vibration frequency with the degree of overlap. In this embodiment, all adjustment coefficients are experimentally calibrated to ensure that 1 + k O≤2, meaning the maximum dispersion force should not exceed twice the baseline value, to avoid exceeding the tolerance range of the item.

[0070] In this embodiment, for paper-plastic composites (K=1): (O) = 70 (1 + 0.4 O), (O) = 60 (1 + 0.6) O);

[0071] For fabrics (K=2): (O) = 30 (1 + 0.5 O), (O) = 25 (1 + 0.7 O), (O) = 5 (1 + 0.3 O); where the reference speed of the paddle is 70 r / min for paper and plastic and 30 r / min for fabric; the reference air pressure of the air nozzle is 60 kPa for paper and plastic and 25 kPa for fabric; the reference vibration frequency of fabric is 5 Hz. If the O standard is still not met after repeating the dispersion operation 3 times, the system determines that it cannot be effectively dispersed, the touch screen displays "Items are overlapping / entangled and cannot be separated, please manually sort them and put them back in", the voice broadcast is synchronized, and the recycling bin door opens for the user to take away and sort them.

[0072] The adaptive lighting category recognition module consists of an ambient light sensor, an LED supplementary light array, and a multispectral camera. It is installed at the end of the independent recognition channel, 15cm away from the overlapping item dispersion module. It is responsible for accurately identifying the actual category k of the item and provides the core basis for sorting and matching.

[0073] Adaptive Illumination Adjustment: The ambient light sensor uses a photoresistor with a measurement range of 0-20000 lux and an accuracy of ±10 lux, acquiring the illumination intensity L within the channel in real time. When L < 500 lux or L > 5000 lux, the central control module activates illumination compensation, adjusting the brightness of the LED supplementary lighting array and camera parameters to stabilize the effective illumination within the optimal recognition range of 1000-3000 lux. The LED supplementary lighting array is arranged on both sides: cool-light LEDs (color temperature 6000K) on the paper and plastic side, and warm-light LEDs (color temperature 3000K) on the fabric side, enhancing the characteristic contrast of different product categories.

[0074] The illumination compensation algorithm is as follows:

[0075]

[0076]

[0077] in, The standard illumination intensity is 2000 lux (matching the standard conditions of T0), L is the real-time illumination intensity, i.e., L_env (unit: lux) collected by the ambient light sensor, T is the camera exposure time (unit: ms), T0 is the reference exposure time under standard illumination L0, with a value of 10ms, G is the camera gain, G0 is the reference gain under standard illumination, with a value of 1.2, η(K) is the category exposure correction coefficient, when K=1 (paper and plastic), η=1.0 (hard materials reflect light stably), when K=2 (fabric), η=1.1 (soft materials absorb light and require longer exposure); the log term is used to suppress gain abrupt changes under extreme illumination to ensure image stability.

[0078] A multispectral camera covering the 400-900nm wavelength band was used to acquire spectral images of the items. For paper and plastic items, color features (HSV space) and shape features (rectangularity, roundness) were extracted. For fabric items, texture features (gray-level co-occurrence matrix) and material features (multispectral reflectance) were extracted. A differential feature fusion model was used to extract effective features, which were then input into an improved SVM classifier to identify the actual item category k. The feature fusion formula for the differential feature fusion model is as follows: ,in, This is the fusion feature vector of the k-th item, with a dimension of 24 (the basic feature vectors are concatenated after normalization). , , , These are feature vectors for color, texture, material, and shape, all of which have been normalized. , , , The feature weights are set according to the differences in product category characteristics;

[0079] Feature fusion formula for paper-plastic composites (k=1):

[0080] = 0.4 + 0.3 + 0.2 + 0.1

[0081] Fabric feature fusion formula (k=2):

[0082] = 0.1 + 0.2 + 0.4 + 0.3

[0083] Wherein, F is the HSV color space feature vector (3-dimensional, including the mean and variance of hue, saturation, and brightness); F is the shape feature vector (2-dimensional, including rectangularity and roundness); F is the material feature vector (8-dimensional, including reflectance of 8 feature bands); and F is the texture feature vector (4-dimensional, including entropy, contrast, energy, and correlation of the gray-level co-occurrence matrix). All feature vectors are normalized to the [0,1] interval using min-max normalization. The weight coefficients are set based on the training results of 1000 sets of samples, with color and shape features accounting for 70% for paper and plastic types, and material and texture features accounting for 70% for fabric types, to ensure recognition accuracy.

[0084] The feature vector is input into the improved SVM (Support Vector Machine) classifier, which uses RBF (Radial Basis Function) with a penalty coefficient C=10 and a kernel parameter σ=0.1. The test results show that the recognition accuracy for paper and plastic and fabric categories is ≥99%, and the recognition time is ≤0.3s. After the recognition is completed, the central control module records the actual category k and weight pre-detection information of the item.

[0085] The sorting execution module consists of an electric main gate, dual-channel branch gates, and guide rails, integrated between the independent identification channel and the dual-category sorting compartments. The electric main gate is located at the exit of the channel, while the dual-channel branch gates are installed at the top entrances of the paper / plastic and fabric compartments, respectively. Both are electromagnetically driven, operating at 24V, with a response time of ≤0.1s. After the gates are fully closed, a limit switch feedback signal is triggered.

[0086] The guide slides feature an inclined design with an optimized structure tailored to different product categories: paper-plastic slides have a 35° inclination angle and an inner wall coated with PTFE (coefficient of friction ≤0.04) to ensure smooth sliding of hard materials; fabric slides have a 30° inclination angle and an inner wall with anti-slip bumps spaced 5cm apart (2mm high) to prevent soft materials from getting stuck. A cushioning pad is installed at the bottom of the slides to reduce impact noise by ≤50dB.

[0087] The sorting logic is uniformly controlled by the central control module: when k=K, the main gate opens immediately, and the corresponding category gate opens at the same time. The item slides into the target compartment through the guide slide, and the main gate and the category gate close sequentially (with an interval of 0.5s); when k≠K, the sorting module remains closed, waiting for subsequent instructions from the user interaction module. After the user changes the selection to K', the system re-executes the overlap processing and recognition process and performs sorting according to the new K'.

[0088] In this embodiment, an alert module is also included. This alert module is linked with the user interaction module and the central control module, and adopts a triple feedback method of sound and light alerts and mini-program push notifications to output differentiated alerts for different abnormal scenarios.

[0089] Dangerous goods alerts include red warning images and text, a 70dB voice loop, and push notifications via a mini-program.

[0090] The alert for exceeding the limit includes a yellow graphic notification (containing category threshold information), a 50dB voice broadcast once, and a mini-program reminder;

[0091] The category mismatch prompt includes a blue interactive graphic that reads, "Item identified as [Category XX], which does not match your selected [Category XX]. Do you want to change your selection?", along with "Confirm" and "Cancel" touch buttons, 55dB voice broadcast, and push notifications of matching results from the mini-program. If the user selects "Confirm," K is updated to k, and sorting begins. If the user selects "Cancel," "Please retrieve your item" is displayed, and the hatch opens.

[0092] The timeout alarm includes items that need to be retrieved by the user (dangerous items, mismatched items, and items whose changes have been cancelled). If the item is not retrieved within 10 minutes, a 1000Hz high-frequency alarm will be triggered (continuously looping), the touchscreen will flash red, and an "item inactivity" abnormal information (including device number, location, and time) will be sent to the operations backend. The mini-program will push a reminder every 2 minutes until the item is retrieved.

[0093] The central control module uses an STM32H743 microcontroller with a core frequency of 400MHz, runs on the RT-Thread real-time operating system, has 1MB of RAM and 16MB of storage, and is responsible for receiving data from various modules and executing control commands. The response latency of each component is ≤100ms. The module has four communication interfaces: RS485, CAN, WiFi, and Bluetooth, which communicate with the sensor module, actuator, cloud platform, and user terminal app, respectively.

[0094] The settlement module is linked to the weighing and settlement unit at the bottom of the dual-category compartment. The weighing unit uses a resistance strain gauge sensor with a range of 0-10kg and an accuracy of ±0.001kg. After the item slides into the compartment and stabilizes for 2 seconds, the actual weight m is measured. The settlement formula is as follows: Where p(k) is the real-time recycling price per k unit (in yuan / kg), which is remotely updated and synchronized to the system by the operation backend. The default price is p(1) = 1.5 yuan / kg for paper and plastic and p(2) = 2.0 yuan / kg for textiles. The revenue R is automatically credited to the user's mini-program wallet, with two decimal places. Users can apply for withdrawal at any time. The withdrawal process is linked with the WeChat Pay API. After initiating a withdrawal, the arrival time is ≤1 minute. The withdrawal record is synchronized to the central control module database in real time.

[0095] The central control module establishes a related database of "user ID - selected category K - detection data - identification result k - sorting action - actual weight - revenue - timestamp". The data is retained for 90 days and supports the operation backend to query, count and export via HTTP protocol, providing data support for recycling scheduling.

[0096] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent paid waste recycling system, characterized in that, include: The smart recycling bin consists of a main body, a central control module, a user interaction module, a settlement module, and a user-end mini-program. The main body of the intelligent recycling bin includes a single recycling compartment door, an independent identification channel, a sorting execution module, and a dual-category sorting compartment; the dual-category sorting compartment includes a paper-plastic compartment and a fabric compartment, both of which are equipped with electric sorting gates at the top and weighing and settlement units at the bottom; The independent identification channel integrates, in sequence along the direction of item transport, a dangerous goods detection module, a quantity restriction module, an overlap identification module, an overlap item dispersion module, and an adaptive lighting category identification module. The user interaction module is used to guide users to select the category of the items to be delivered and to obtain the target category selected by the user. The hazardous materials detection module is used to perform pre-deployment safety screening of the materials; The quantity limit module is used to control the quantity of items delivered in a single delivery. The overlap recognition module is used to determine the overlap status of the placed items and generate an overlap determination result; The overlapping item dispersion module is used to disperse the deployed items using a dispersion structure based on the overlap determination result. The adaptive lighting category recognition module is used to identify the actual category of the placed items and generate category recognition results; The sorting execution module is used to perform sorting actions based on the category identification results; The central control module is used to control the operation of the entire system, receive data from each module, and execute control commands. The settlement module is used to settle revenue based on the weighing results from the weighing settlement unit.

2. The intelligent paid waste recycling system according to claim 1, characterized in that: The user interaction module displays category icons for paper and plastic products and textiles, along with corresponding descriptions for each category, on a touchscreen display, and works in conjunction with a voice acquisition unit to recognize specific delivery commands.

3. The intelligent paid waste recycling system according to claim 1, characterized in that: The hazardous materials detection module includes a metal detection unit, a dielectric constant detection unit, and an X-ray fluorescence scanning unit, which are used to detect the metal composition, dielectric constant, and elemental composition characteristics of the placed items, respectively. The module calculates a comprehensive hazardous materials judgment coefficient to comprehensively determine the degree of danger of the placed items. If the comprehensive hazardous materials judgment coefficient is greater than or equal to the preset hazardous materials threshold, the item is determined to be a hazardous material, and an alarm is triggered and the warehouse door is opened.

4. The intelligent paid waste recycling system according to claim 1, characterized in that: The delivery quantity restriction module constructs a three-dimensional model of the item using an infrared ranging sensor array to calculate its volume, and uses a weight pre-sensor to collect the item's weight. Based on the target category selected by the user, it calls the differentiated delivery threshold formula for the corresponding category to determine whether the volume or weight exceeds the limit. If the limit is exceeded, the user is prompted to deliver the item in multiple batches. Once the limit is met, the item is allowed to proceed to the overlap recognition module.

5. The intelligent paid waste recycling system according to claim 1, characterized in that: The overlap recognition module includes a high-definition industrial camera and a TOF depth sensor, used to acquire two-dimensional images and three-dimensional coordinate information of the object, respectively, and extract the contour features, area features, and depth features of the object. An overlap coefficient is used to determine the degree of overlap of the object. The formula for calculating the overlap coefficient is as follows: ,in, The area of ​​the overlapping region in the object image is obtained through contour intersection calculation; The sum of the total areas of all items. The number of detected items is determined through contour connectivity analysis; Let be the average depth of the i-th and j-th items. The average depth of all items. , These are the weighting coefficients. This is the category correction factor. If the overlap coefficient of the placed items is greater than the preset overlap judgment threshold, it is judged as significant overlap, and the central control module starts the overlapping item dispersion module; otherwise, it directly enters the adaptive lighting category recognition module.

6. The intelligent paid waste recycling system according to claim 1, characterized in that: The overlapping item dispersion module adapts the corresponding dispersion structure according to the target category, and constructs a dispersion parameter adjustment model based on the target category and the overlap coefficient to adjust the dispersion parameters of the dispersion structure and disperse the items that are determined to be significantly overlapping.

7. The intelligent paid waste recycling system according to claim 1, characterized in that: The adaptive lighting category recognition module collects the light intensity in the channel through an ambient light sensor. If the light intensity exceeds the optimal range, it adjusts the camera exposure time and gain, adjusts the brightness of the LED fill light array, and extracts effective features based on the characteristic differences between paper-plastic and fabric categories using a differentiated feature fusion model. The extracted features are then input into an improved SVM classifier to complete category recognition, and the category recognition results are fed back to the central control module.

8. The intelligent paid waste recycling system according to claim 1, characterized in that: The sorting execution module is used to receive the category matching signal from the central control module. If the actual category matches the user's selection, it controls the main gate and the corresponding category compartment's sub-gate to open sequentially, and guides the items to slide into the corresponding category compartment through the guide slide. If the categories do not match, the gate remains closed, and the sorting action is performed after the user adjusts their selection.

9. The intelligent paid waste recycling system according to claim 1, characterized in that: The settlement module is used to obtain the actual weight data of the weighing and settlement unit after the items are delivered to the target compartment. Based on the recycling unit price of the corresponding category, the revenue is calculated using the revenue calculation formula and automatically credited to the user's mini-program wallet. Users can withdraw the money at any time, and the delivery and settlement records are stored synchronously.

Images