Smart city iot device data collection management platform and implementation method thereof

Abstract

This invention relates to the field of Internet of Things (IoT) data management technology, and more particularly to a smart city IoT device data acquisition and management platform and its implementation method. The method includes the following steps: deploying IoT sensing devices inside trash cans to construct an IoT front-end sensing network; collecting data on incremental littering events in the trash cans through the IoT front-end sensing network and calculating the urgency of the trash can's state to obtain a dynamic urgency index; performing regulatory-weighted collection urgency processing based on the dynamic urgency index to dynamically plan the theoretically optimal collection sequence for target collection vehicles; and processing the theoretically optimal collection sequence into segmented navigation instructions to manage the smart trash can collection operation. This invention, by deploying a multi-sensor network inside trash cans to monitor the trash status in real time, achieves dynamic monitoring of trash thermal decomposition and intelligent prioritization of regionally differentiated collection, thus accurately guiding the entire sanitation operation process.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) data management technology, and in particular to a smart city IoT device data acquisition and management platform and its implementation method. Background Technology

[0002] The main problems in urban waste collection and management include: frequent overflowing of garbage bins leading to environmental sanitation hazards; unreasonable collection routes causing resource waste; and inefficiency due to a mismatch between collection frequency and actual demand. Especially in densely populated areas and bustling commercial districts, waste generation fluctuates greatly, and garbage bins easily overflow during peak hours, affecting not only the city's appearance but also causing public health problems and citizen complaints. Statistics show that large and medium-sized cities receive tens of thousands of environmental sanitation complaints annually due to overflowing garbage bins, with related emergency response costs accounting for 15%-20% of the total waste collection and transportation costs. Existing intelligent waste monitoring systems typically use a single parameter to measure the filling height of garbage bins and infer their filling status. However, this simple technical solution has significant drawbacks: measuring only the surface height cannot accurately reflect the true capacity utilization rate of the garbage bins. Because the accumulation, density, and distribution of garbage are extremely uneven, it is easy to see situations where the filling level is "falsely high" or "falsely low." For example, large-volume, lightweight waste (such as packaging foam, cardboard boxes, etc.) will quickly occupy the upper space of the trash can, triggering a high fill rate alarm; conversely, small-volume, high-density waste (such as kitchen waste, construction waste, etc.) will concentrate at the bottom, and although the filling height is not high, it is actually close to the capacity limit. Summary of the Invention

[0003] Based on this, the present invention provides a smart city IoT device data acquisition and management platform and its implementation method to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for implementing a smart city IoT device data acquisition and management platform includes the following steps:

[0005] Step S1: Deploy IoT sensing devices inside the trash can to build an IoT front-end sensing network; collect data on incremental throwing events in the trash can through the IoT front-end sensing network, analyze the changes in structural voxels in the trash can caused by incremental throwing events, and generate three-dimensional structural voxel data of the trash.

[0006] Step S2: Calculate the total waste filling rate based on the three-dimensional structure voxel data of the waste; evaluate the dynamic thermal decomposition degree data based on the three-dimensional structure voxel data of the waste, and calculate the urgency of the state inside the bin based on the total waste filling rate to obtain the dynamic urgency index; perform regulatory weighted collection urgency processing based on the dynamic urgency index, and then identify the cluster of sites to be processed.

[0007] Step S3: Extract the coordinate data of the sites to be visited in the cluster of sites to be processed, and dynamically plan the theoretically optimal collection sequence of the target collection vehicles;

[0008] Step S4: Process the theoretically optimal collection and transportation sequence into segments for navigation instructions, and add detailed single-point operation annotations using the three-dimensional structure voxel data of the waste to manage the smart waste bin collection and transportation operation.

[0009] Preferably, the present invention also provides a smart city IoT device data acquisition and management platform, comprising: a data layer, a service layer, and an application layer. The service layer is configured at the output end of the data layer, and the application layer is configured at the output end of the service layer. The data layer is used to collect and analyze smart city information. The service layer is used to provide different services to meet the access requirements of the application layer. The application layer enables access to smart terminal devices through an access interface. The data layer is used to execute the implementation method of the smart city IoT device data acquisition and management platform as described above. The data layer in this smart city IoT device data acquisition and management platform includes:

[0010] The IoT data acquisition module is used to deploy IoT sensing devices inside the trash can and build an IoT front-end sensing network. It collects data on incremental throwing events in the trash can through the IoT front-end sensing network, analyzes the changes in structural voxels in the trash can caused by incremental throwing events, and generates three-dimensional structural voxel data of the trash.

[0011] The dynamic risk assessment module is used to calculate the total waste filling rate based on the three-dimensional structure voxel data of the waste; to assess the dynamic thermal decomposition degree data based on the three-dimensional structure voxel data of the waste; and to calculate the urgency of the state inside the bin based on the total waste filling rate, thereby obtaining a dynamic urgency index; to perform regulatory weighted collection urgency processing based on the dynamic urgency index, and then identify the cluster of sites to be processed.

[0012] The collection and transportation route planning module is used to extract the coordinate data of the sites to be visited in the cluster of sites to be processed, and to dynamically plan the theoretically optimal collection and transportation sequence for the target collection and transportation vehicles.

[0013] The intelligent operation management module is used to process the theoretically optimal collection and transportation sequence into segments and navigation instructions, and to perform single-point fine operation annotations through the three-dimensional structure voxel data of waste, so as to realize the intelligent management of waste collection and transportation in smart cities.

[0014] This invention achieves accurate assessment of the true filling state of garbage bins by precisely capturing three-dimensional structural data of waste, rather than simply measuring surface height. This three-dimensional monitoring mode based on an IoT front-end sensing network can effectively distinguish between lightweight, large-volume waste and high-density, small-volume waste, accurately reflecting capacity utilization and making the early warning mechanism more reliable. Through comprehensive analysis of dynamic thermal decomposition data and filling rate, the actual urgency level of the garbage bin's internal state can be intelligently judged, avoiding misjudgments caused by single-parameter judgments and significantly reducing sanitation emergency handling costs by approximately 15%-20%. The platform's regulatory weighted collection and transportation processing based on a dynamic urgency index realizes the transformation from "fixed-cycle collection" to "on-demand collection," greatly improving collection efficiency. Through intelligent identification and cluster analysis of disposal sites, combined with the theoretically optimal collection and transportation sequence of dynamic programming, the garbage collection and transportation route is significantly optimized, reducing empty mileage and fuel consumption, and lowering carbon emissions and operating costs. Refined segmented navigation instructions and operation annotation functions provide clear operational guidance for frontline sanitation workers, improving operation quality and efficiency. Furthermore, through precise data collection and analysis, the platform provides city managers with a comprehensive view of waste generation, distribution patterns, and processing efficiency, supporting scientific decision-making and optimized resource allocation. The platform's intelligent prediction function can anticipate peak waste generation based on historical data patterns, allowing for advance adjustments to collection plans and effectively preventing overflow incidents. In densely populated areas and bustling commercial districts, precise responses to significant fluctuations in waste generation ensure urban environmental sanitation and appearance, reduce citizen complaints, and improve public satisfaction. Therefore, the implementation method of this invention's smart city IoT device data collection and management platform constructs three-dimensional structural voxel data of waste through a multi-sensor front-end sensing network, analyzes waste decomposition activity using infrared thermal imaging, assesses risk urgency using meteorological data and environmental parameters, achieves differentiated regional collection priority ranking, and predicts dynamic unloading impact values ​​based on internal structural cavity identification and centroid offset analysis, generating refined operational guidance. This enables intelligent management of the entire process of waste monitoring, assessment, scheduling, and collection. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the implementation method of the smart city IoT device data acquisition and management platform of the present invention.

[0016] Figure 2 A line graph illustrating the voxel activity score in a smart city IoT device data acquisition and management platform.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0020] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for implementing a smart city IoT device data acquisition and management platform, comprising the following steps:

[0022] Step S1: Deploy IoT sensing devices inside the trash can to build an IoT front-end sensing network; collect data on incremental throwing events in the trash can through the IoT front-end sensing network, analyze the changes in structural voxels in the trash can caused by incremental throwing events, and generate three-dimensional structural voxel data of the trash.

[0023] In this embodiment of the invention, various IoT sensing devices are deployed within a municipal waste bin via embedded installation. These include an array of eight HC-SR04 ultrasonic sensors arranged in a ring on the upper inner wall, four strain gauge pressure sensors fixed at the four corners of the bin bottom, a 32×32 pixel infrared thermal imaging sensor mounted in the upper center, a gas sensor group mounted in the middle of the bin wall, and a temperature and humidity sensor in the upper part of the bin. All sensors are connected to an edge computing unit equipped with an ARM Cortex-A53 processor via IP67 waterproof shielded cables, constructing a complete front-end sensing network. Pressure sensor data is acquired every 100 milliseconds via a 12-bit ADC. When a weight change exceeding 100 grams is detected, it is marked as an incremental throwing event, triggering the ultrasonic sensor array to operate. Each sensor continuously acquires the sound echo time five times, and the average value after removing the maximum and minimum values ​​is taken as the distance data. The measured distance value is combined with the sound velocity value to convert it into spatial rectangular coordinates, forming the height coordinates of the waste surface. The edge computing unit divides the trash can space into a 10×10×15 three-dimensional grid. Each grid cell represents a voxel. The density distribution of each voxel is calculated using surface height coordinates and weight data to generate three-dimensional structure voxel data of the trash containing location and density information.

[0024] Step S2: Calculate the total waste filling rate based on the three-dimensional structure voxel data of the waste; evaluate the dynamic thermal decomposition degree data based on the three-dimensional structure voxel data of the waste, and calculate the urgency of the state inside the bin based on the total waste filling rate to obtain the dynamic urgency index; perform regulatory weighted collection urgency processing based on the dynamic urgency index, and then identify the cluster of sites to be processed.

[0025] In this embodiment of the invention, three-dimensional structure voxel data of the garbage is acquired, the number of filled voxels is counted and divided by the total number of voxels to calculate the total filling rate of the garbage. Simultaneously, an infrared thermal imaging sensor is used to collect the internal temperature distribution of the garbage bin, and spatial mapping of the temperature data is performed to generate three-dimensional temperature-scale voxel data. Continuous areas with temperatures more than 5°C above the ambient temperature are extracted as effective heat-generating areas, and their average filling density is calculated. When the density is greater than 0.8 g / cm³, a weight of 1.5 is set for high-risk organic matter; when the density is less than 0.3 g / cm³, a weight of 0.5 is set for low-risk organic matter. The voxels of the effective heat-generating areas are scored for activity, and the highest activity score is extracted as dynamic thermal decomposition degree data. Subsequently, piecewise function compensation is performed based on the ambient humidity, and an odor diffusion influence coefficient is set by combining real-time gas concentration values ​​obtained from a gas sensor and ambient wind speed values ​​obtained from an urban meteorological data interface. Based on these parameters, a dynamic urgency index is calculated, and a regulatory weight coefficient is matched according to the functional area category label of the garbage bin (commercial area, residential area, or public place), performing a weighted collection urgency calculation and normalization to generate collection score data. Garbage bins with a collection score greater than 75 are selected to form a high-priority site list. The site with the highest collection score is selected as the cluster generation center point. Sites within a range defined by this center are then aggregated to form a cluster of sites to be processed.

[0026] Step S3: Extract the coordinate data of the sites to be visited in the cluster of sites to be processed, and dynamically plan the theoretically optimal collection sequence of the target collection vehicles;

[0027] In this embodiment of the invention, the geographical coordinates of all sites are extracted from the cluster of sites to be processed, and a dataset of site coordinates to be visited is constructed. The garbage bin site with the highest collection score is set as the first target point for the target collection vehicle. The real-time driving time from this point to all remaining points is queried through the map service interface to obtain a table of candidate point time consumption. The minimum value search algorithm is executed to select the point with the shortest time as the next optimal operation point. When the driving time of multiple points does not differ by more than 60 seconds, the optimal point is determined by considering both driving time and collection score through a comprehensive scoring mechanism. The selected operation point is added to the vehicle task list and removed from the site coordinate data to be visited. The above route planning operation is repeated until the vehicle capacity limit or shift duration limit is met, generating a feasible operation sequence. Subsequently, a secondary weight adjustment is performed on the key points in the sequence whose collection scores are significantly higher than the average value. The order of points is adjusted through a local optimization strategy, while considering the peak-hour avoidance rules for garbage bins in commercial areas and public places. Finally, the theoretically optimal collection sequence is generated, realizing intelligent planning of garbage collection routes.

[0028] Step S4: Process the theoretically optimal collection and transportation sequence into segments for navigation instructions, and add detailed single-point operation annotations using the three-dimensional structure voxel data of the waste to manage the smart waste bin collection and transportation operation.

[0029] In this embodiment of the invention, the theoretically optimal collection sequence is extracted into an ordered latitude and longitude array, and a complete navigation path is generated through the driving route planning interface of the map open platform. The continuous path is divided into multiple independent segments according to the stations. Each segment constitutes a segmented navigation instruction containing information such as the coordinates of the origin and destination, the set of waypoints, the set of turning points, and the estimated time and distance. For each target station, its three-dimensional structure voxel data of the waste is analyzed, and a set of voxels with a density value of less than 0.3 kg / m³ and spatial continuity is identified as the internal structural cavity. The ratio of the total cavity volume to the volume of the waste bin and its vertical distribution are calculated to assess the structural instability risk index. At the same time, the position of the centroid of the waste pile is calculated, and whether the horizontal offset from the geometric center of the bin exceeds 30% of the bin radius is used to determine the off-center loading state. Combining the structural instability risk and the off-center loading state, the dumping impact force is quantified, and a five-level estimated dynamic unloading impact value is generated. Based on this, corresponding operational suggestions are formulated, such as conventional dumping, slow dumping, segmented dumping, multiple dumping at small angles, or special treatment. The segmented navigation instructions and single-point detailed operation annotations are integrated into a complete operation instruction package, which is pushed to the vehicle terminal of the target collection vehicle through the 4G / 5G network to realize the whole process of refined management of smart garbage bin collection operations.

[0030] Preferably, step S1 includes the following steps:

[0031] Step S11: Deploy IoT sensing devices inside trash cans in the smart city and connect them to an edge computing unit with at least 4 general input / output interfaces and 1 analog-to-digital conversion interface to build an IoT front-end sensing network; wherein, the IoT sensing devices include ultrasonic sensor arrays, pressure sensors, infrared thermal imaging sensors, gas sensors and temperature and humidity sensors;

[0032] Step S12: Read the voltage signal of the pressure sensor in the IoT front-end sensing network in real time, and obtain the real-time total weight data of the garbage filling after analog-to-digital conversion;

[0033] Step S13: Based on the real-time total weight data of the garbage, mark each instance of a weight change greater than 100 grams as an incremental throwing event;

[0034] Step S14: Based on the time of the incremental throwing event, use the ultrasonic sensor array in the IoT front-end sensing network to obtain the echo time from each sensor to the garbage surface to obtain the original echo duration.

[0035] Step S15: Convert the original echo duration into the corresponding spatial rectangular coordinates of the sensor according to the preset sound velocity value, thereby determining the surface height coordinates inside the trash can;

[0036] Step S16: Analyze the structural voxel changes of the throwing event based on the surface height coordinates and the real-time total weight of the waste, and generate three-dimensional structural voxel data of the waste.

[0037] In this embodiment of the invention, when deploying IoT sensing devices inside the trash can, each sensor is fixed to a preset position on the inner wall of the trash can using an embedded installation method. The ultrasonic sensor array consists of eight HC-SR04 type sensors, evenly distributed in a ring on the upper inner wall of the trash can; the pressure sensor uses four distributed strain gauge sensors, fixed at the four corners of the bottom of the trash can; the infrared thermal imaging sensor uses a 32×32 pixel array, installed in the center of the upper part of the trash can; the gas sensor group includes methane, ammonia, and hydrogen sulfide detection units, installed in the middle of the can wall; and the temperature and humidity sensor is installed in the upper part of the can. All sensors are connected to an edge computing unit equipped with an ARM Cortex-A53 processor, eight GPIO interfaces, and two 12-bit ADC conversion interfaces via IP67 waterproof shielded cables, forming a complete front-end sensing network.

[0038] During the pressure sensor acquisition process, the edge computing unit triggers a sampling interrupt every 100 milliseconds via a timer to perform analog-to-digital conversion. The analog voltage signals output from the four pressure sensors are amplified by a preamplifier circuit and then digitized by the edge computing unit's built-in 12-bit ADC, achieving a conversion accuracy of 0.1 g / bit. The converted digital signal is then filtered by a low-pass filter to remove high-frequency noise, with a cutoff frequency set to 10 Hz. The digitized values ​​from the four pressure sensors are weighted and summed, with weighting coefficients determined based on the force distribution characteristics at the bottom of the trash can: 0.26, 0.24, 0.25, and 0.25 respectively. The final weighted sum is multiplied by a pre-calibrated linear scaling factor of 5.78 using standard weights to obtain accurate real-time total weight data of the filled trash, in grams.

[0039] Continuous monitoring of the total weight of waste filling in real time is performed, and a sliding window method is used for data smoothing, with the window size set to 5 sampling points. When the absolute value of the difference between two adjacent smoothed data points exceeds a 100-gram threshold, the edge computing unit records the current timestamp in its internal time register and generates an incremental throwing event data structure containing three fields: timestamp, weight change value, and duration of change. For cases where the weight suddenly increases and then quickly decreases, a 50-millisecond dead zone is set to avoid false judgments. After the throwing event is determined, the event information is stored in a circular buffer with a size of 256 event records, and the event index table is updated simultaneously.

[0040] Upon triggering of the incremental throwing event, the edge computing unit immediately sends a data acquisition command to the ultrasonic sensor array. Each ultrasonic sensor receives a 10-microsecond-wide trigger pulse via its I / O trigger pin and then emits an ultrasonic signal beam at a frequency of 40 kHz. The ultrasonic signal is received by the sensor after being reflected from the surface of the trash can. The edge computing unit precisely measures the time interval from transmission to reception using a hardware counter with a clock frequency of 1 MHz, providing a time resolution of 1 microsecond. To improve measurement accuracy, each sensor is measured five times consecutively. The maximum and minimum values ​​are removed, and the average of the remaining three measurements is taken as the raw echo duration of that sensor, in microseconds. Considering the multipath reflection problem inside the trash can, the maximum detection distance is set to 1.5 meters, corresponding to a maximum echo duration of 8824 microseconds.

[0041] When converting the original echo duration to spatial coordinates, the sound velocity is first corrected based on the current ambient temperature inside the trash can. At a standard temperature of 25 degrees Celsius, the sound velocity is set to 346.3 m / s, and the temperature correction formula is: actual sound velocity = 346.3 + 0.6 × (current temperature - 25). The corrected sound velocity is multiplied by the original echo duration and then divided by 2 to obtain the distance from each ultrasonic sensor to the trash surface. Based on the preset installation coordinates of each sensor inside the trash can and the measured distance, the spatial rectangular coordinates of each reflection point on the trash surface are calculated using trigonometric functions. A right-handed coordinate system is used, with the center of the trash can bottom as the origin and the vertically upward direction as the positive z-axis. For areas where sensor echoes are lost, linear interpolation between adjacent valid points is used to fill in the gaps.

[0042] The space of the trash can was divided into a 10×10×15 three-dimensional grid, with each grid cell representing a voxel, each voxel measuring 5 cm × 5 cm × 3 cm. Based on the surface height coordinates, the occupancy status of each voxel was determined using back projection; voxels below the surface were marked as occupied, and those above the surface were marked as unoccupied. Simultaneously, the average density of occupied voxels was calculated using real-time total weight data of the trash. For newly occurring throwing events, the set of newly occupied voxels was identified by comparing the changes in voxel occupancy status before and after the event, and the density distribution of these newly occupied voxels was calculated based on the weight change. By comparing different density thresholds (low density 0.2 g / cm³, medium density 0.6 g / cm³, high density 1.2 g / cm³), the three-dimensional structure of the trash was divided into different category regions, ultimately generating three-dimensional structure voxel data of the trash containing location and density information.

[0043] Preferably, step S16, which involves analyzing the structural voxel changes of the throwing event based on the surface height coordinates and the real-time total weight of the waste, includes:

[0044] Based on incremental throwing events, the ratio of volume increment to weight increment for each throwing event is calculated using surface height coordinates and real-time total weight data of the waste, and the density of different batches of waste is evaluated to obtain the stratified waste density value.

[0045] Acquire historical surface height time-series data;

[0046] Using historical surface height time series data and the three-dimensional geometric parameters of the trash can pre-set in the edge computing unit, coordinate mapping and basic voxel mesh calibration are performed in real time, and density weights are assigned to voxels at the corresponding depths using the layered trash density values ​​to generate a density layered voxel model;

[0047] The density-layered voxel model is divided into lower-layer voxel meshes, and then the lower-layer voxel meshes are simulated and corrected for gravity settlement effect, that is, the layered waste density values ​​corresponding to the lower-layer voxel meshes are compressed and compensated by 1-5%, so as to obtain the three-dimensional structure voxel data of waste.

[0048] In this embodiment of the invention, for each incremental throwing event, the total weight data of the garbage before and after the event is first obtained, and the weight increment Δm is calculated. Simultaneously, based on the surface height coordinate data before and after throwing, the volume increment Δv is calculated using a three-dimensional discrete integration method. Specifically, the cross-section of the garbage bin is divided into a 10×10 grid, with each grid area being 25 square centimeters. For each grid cell, the surface height difference before and after throwing is calculated, multiplied by the grid area to obtain the volume increment for that grid, and then summed over all grids to obtain the total volume increment Δv. When calculating density, the formula ρ = Δm / Δv is used, where ρ represents the stratified garbage density value (g / cm³), Δm represents the weight increment (g), and Δv represents the volume increment (cm³). The calculation results are stored in a three-dimensional array according to the current height of the garbage surface, forming a vertical stratified density record with a resolution of 3 cm / layer.

[0049] The edge computing unit maintains a three-dimensional circular buffer to store historical surface height data. This buffer is 10×10×24 in size, corresponding to a 10×10 grid of the trash can's cross-section and a time window of the most recent 24 hours. When a throwing event occurs, the current surface height coordinates are mapped onto the 10×10 grid, generating a height matrix, which is then timestamped and stored in the circular buffer. To reduce storage space, a relative height storage method is used, recording only the height value relative to the bottom of the trash can with an accuracy of 0.5 cm. Simultaneously, for areas that have not changed for a long time (more than 30 minutes), a time-stamped compression algorithm is used, recording only the start and end times to reduce data redundancy. The circular buffer is updated using a first-in, first-out (FIFO) principle; when new data is written, the oldest historical data is overwritten. Surface height data at different time points is quickly retrieved using a buffer index table.

[0050] The edge computing unit pre-determines the three-dimensional geometric parameters of the trash can, including a bottom diameter of 50 cm, a mouth diameter of 58 cm, a height of 90 cm, and a wall inclination angle of 3.5 degrees. Based on these parameters, a frustum model is established, and a right-handed coordinate system is set within the edge computing unit, with the origin at the center of the bottom and the z-axis pointing vertically upwards. The trash can space is divided into a 10×10×30 basic voxel grid, with a horizontal resolution of 5 cm and a vertical resolution of 3 cm. For each voxel, its center point coordinates are calculated, and it is determined whether it falls within the geometric range of the frustum. Subsequently, historical surface height time-series data is interpolated along the time axis to determine the surface morphology at each moment. For voxels below the current surface, the corresponding layer density value is queried based on their depth, and density weighting is applied. If a density value for a certain depth layer is not directly measurable, a distance-weighted average algorithm is used, taking into account the density values ​​of the four adjacent layers to calculate the interpolation result. Finally, a density-layered voxel model incorporating both location and density attributes is formed.

[0051] For the density-stratified voxel model, the voxel model is divided into three layers: upper, middle, and lower, based on the depth of the waste layer. The upper layer contains voxels within 0-15 cm below the surface, the middle layer contains voxels within 15-45 cm below the surface, and the lower layer contains voxels above 45 cm below the surface. For the lower voxel grid, a gravity settling effect simulation is performed. Specifically, the compression compensation coefficient k is calculated based on the waste retention time t (in hours), using the formula k = 0.01 + 0.0002 × t, where t does not exceed 200 hours and k does not exceed 0.05 (i.e., a maximum of 5%). For the density value ρ of each voxel in the lower grid, a correction calculation is performed: ρ correction = ρ × (1 + k), where ρ correction represents the density value corrected for the gravity settling effect. The density correction also considers the voxel's depth d; the greater the depth, the higher the compression ratio. Specifically, a depth factor of 0.0005 × (d - 45) is added to the basic compression compensation coefficient, where d represents the depth from the surface (in centimeters). After completing the density correction of all voxels, the location information and the corrected density values ​​are integrated to generate three-dimensional structure voxel data of the waste that includes the effect of gravity settling.

[0052] Of particular importance is the process of meshing the lower-level voxel model into a density-layered voxel grid, followed by simulation corrections for gravity settlement effects on the lower-level voxel grid, including:

[0053] Using the lowest layer of filled voxels in the density-layered voxel model as a benchmark, all filled voxels above this layer are identified as upper load-bearing voxels.

[0054] By summing up the weights of all the upper load-bearing elements, the total cumulative pressure acting on the top surface of the lower waste layer is calculated.

[0055] Based on the total cumulative pressure value, a dynamic compression compensation coefficient between 1% and 5% is determined by querying and using a preset pressure-compression coefficient mapping table;

[0056] Voxels located in all unfilled areas of the density-layered voxel model and below the upper load-bearing voxels are identified as lower voxels to be corrected.

[0057] The density value of the layered waste corresponding to the lower voxel to be corrected is multiplied by the dynamic compression compensation coefficient to obtain the density value after pressure correction. This value is then used to update the original model, thereby obtaining the three-dimensional structure voxel data of the waste.

[0058] In this embodiment of the invention, a density-layered voxel model is first used as a basis, and the lowest layer position of filled voxels is determined through a voxel state traversal algorithm. In implementation, starting from the bottom of the trash can (z=0), all voxels in a 10×10 grid are scanned upwards layer by layer, and the z-coordinate of the first non-zero density voxel appearing in each grid column is recorded. When at least one non-zero density voxel is detected in all grid columns, the minimum z-coordinate is taken as the reference plane for the lowest layer of filled voxels, and this z-value is denoted as zmin. Subsequently, a region growing algorithm is executed to mark all voxels located at z-coordinates greater than or equal to zmin and with a density value greater than 0.05 g / cm³ as upper-layer load-bearing voxels. To improve computational efficiency, a three-dimensional array is used to record the state flag of each voxel, where a flag value of 1 indicates that the voxel is an upper-layer load-bearing voxel. The scanning process employs a depth-first search strategy to ensure that all connected filled voxels are correctly identified.

[0059] After determining the upper load-bearing voxels, the edge computing unit performs vertical pressure accumulation calculations. It processes each of the 10×10=100 grid columns in the vertical direction, traversing all voxels marked as upper load-bearing voxels from top to bottom within each column. For each load-bearing voxel, its pre-assigned density value ρi (g / cm³) is read and multiplied by its voxel volume vi (cm³) to calculate the voxel mass mi, where vi = 5×5×3 = 75 cm³. For grid column j, the total mass of all upper load-bearing voxels in that column is calculated as Mj = ∑mi, where i represents the index of all upper load-bearing voxels in that grid column. The mass is then converted to force: Fj = Mj × 9.8 × 10⁻⁶. ―3 Where Fj is the pressure generated by grid column j (in Newtons), 9.8 is the gravitational acceleration (in meters per second squared), and 10^(-3) is the unit conversion factor. Finally, the pressures of all grid columns are summed to obtain the total cumulative pressure value acting on the top surface of the lower layer of waste, Ftotal = ∑Fj, where j represents the grid column number (1 to 100).

[0060] The edge computing unit contains a pre-built pressure-compression coefficient mapping table, obtained through materials mechanics testing. The testing process uses standard waste samples (containing typical municipal waste components such as plastics, paper, and organic matter, mixed in a 4:3:3 ratio) to measure compressibility under different pressure conditions. The mapping table contains 11 discrete data points: [0 N, 1.0%], [50 N, 1.4%], [100 N, 1.8%], [150 N, 2.2%], [200 N, 2.6%], [250 N, 3.0%], [300 N, 3.4%], [350 N, 3.8%], [400 N, 4.2%], [450 N, 4.6%], [500 N and above, 5.0%]. Based on the total cumulative pressure value Ftotal, a binary search is performed in the mapping table to locate the two closest discrete points. If Ftotal falls between two discrete points, the corresponding dynamic compression compensation coefficient k is calculated using linear interpolation. The interpolation formula is k = k1 + (k2 - k1) × (Ftotal - F1) / (F2 - F1), where F1 and F2 are the pressure values ​​at adjacent discrete points, and k1 and k2 are the corresponding compression coefficients. The calculation result is rounded to three decimal places.

[0061] The lower-layer voxels to be corrected are identified by executing a 3D scanning algorithm. First, a Boolean 3D array of the same size as the voxel model is generated, with all initial values ​​set to false. For each upper-layer load-bearing voxel, its xy-coordinate position on the horizontal plane is recorded, and the Boolean values ​​of all voxel positions below that position whose z-coordinate is less than that of the load-bearing voxel are marked as true. Simultaneously, the density value in the original density-layered voxel model at that position is checked; only voxels with a density value of zero (indicating an unfilled area) are marked as true, otherwise they are reset to false. Using a horizontal plane projection method, all voxels located directly below the upper-layer load-bearing voxels in each grid column and currently unfilled are identified. To avoid erroneous corrections, a connectivity check is also performed to ensure that there is a connected path between the voxel to be corrected and the lowest-layer reference plane, excluding void areas blocked by other objects.

[0062] The density stratification voxel model is updated based on the location set of the lower voxels to be corrected and the dynamic compression compensation coefficient k. First, a temporary 3D array is created to store the corrected density values. For each lower voxel to be corrected, the nearest non-zero density load-bearing voxel directly above it is retrieved, and the stratified waste density value ρ0 of that load-bearing voxel is obtained. Then, the pressure-corrected density value ρ1 = ρ0 × (1 + k / 100) is calculated, where k is the dynamic compression compensation coefficient in percentage form. For example, when k = 3.5%, ρ1 = ρ0 × 1.035. To simulate the changes in physical properties after compression, a secondary correction is performed based on the depth difference; the farther away from the load-bearing voxel, the smaller the correction coefficient. The correction formula is ρ2 = ρ1 × (1 - 0.02 × Δz / 30), where Δz is the difference in z-coordinate (in centimeters) between the voxel to be corrected and its corresponding load-bearing voxel, and 30 is the normalization coefficient. The calculated ρ2 is the final corrected density value, which is written to the corresponding position in the temporary array. After processing all voxels to be corrected, the temporary array is merged with the original density-layered voxel model to obtain complete three-dimensional structure voxel data of the waste containing the gravity settling effect.

[0063] Preferably, step S2, which involves evaluating the degree of dynamic thermal decomposition based on the three-dimensional structure voxel data of the waste and calculating the urgency of the condition inside the bin based on the total waste filling rate, includes:

[0064] Infrared thermal imaging sensors in the front-end sensing network of the Internet of Things are used to collect temperature distribution information inside the trash can, and infrared temperature data inside the can is obtained.

[0065] Using the three-dimensional structure voxel data of the waste as a spatial reference, pixel-level registration and overlay are performed based on the infrared temperature data inside the bin to generate three-dimensional temperature scale voxel data.

[0066] Ambient temperature and humidity are obtained by using temperature and humidity sensors in the front-end sensing network of the Internet of Things.

[0067] The dynamic thermal decomposition degree was assessed based on three-dimensional temperature scale voxel data and ambient temperature.

[0068] Environmental risk compensation is performed based on dynamic thermal decomposition data using ambient humidity, and the urgency of the state inside the bin is calculated based on the total waste filling rate to obtain a dynamic urgency index.

[0069] In this embodiment of the invention, the infrared thermal imaging sensor in the IoT front-end sensing network adopts a 32×32 pixel array structure, with a sampling frequency set to once every 10 minutes. During the acquisition process, the edge computing unit sends an acquisition command to the infrared thermal imaging sensor through a serial peripheral interface, triggering the sensor to perform a complete scan. The sensor uses an uncooled microbolometer array with a temperature measurement range of -20℃ to 120℃ and a temperature resolution of 0.1℃. The sensor has a built-in 16-bit ADC conversion circuit to convert the thermal radiation signal into a digital temperature value. Since the sensor is installed at the top center of the trash can, a 120° wide-angle lens is used to cover the entire interior space of the can, and the acquired raw data is a planar thermal image. After receiving the 32×32 pixel temperature matrix, the edge computing unit performs Gaussian filtering to eliminate thermal noise, with a filter kernel size of 3×3 and a standard deviation of 0.8.

[0070] The obtained 3D structure voxel data of the waste was used as a spatial reference datum for temperature data spatial mapping. First, a geometric transformation matrix was established between the infrared thermal imaging sensor and the 3D coordinate system of the waste bin. This transformation matrix was determined by the correspondence of five pre-calibrated feature points, located at the top center and four edges of the waste bin. For each pixel in the 32×32 temperature matrix, its ray equation in 3D space was determined using back projection, and then intersected with the surface voxels in the waste's 3D structure voxel data to obtain the 3D spatial coordinates of the temperature point. For voxels that could not be directly observed (such as those obscured by other waste), a temperature attenuation model was used for estimation, with an attenuation coefficient set to 0.15 / cm. During registration, the Iterative Closest Point (ICP) algorithm was used to eliminate deviations caused by sensor installation errors, ensuring registration accuracy within 2 cm. After spatial mapping, each voxel, in addition to density information, was assigned a corresponding temperature attribute value, forming 3D temperature-scale voxel data containing spatial location, density, and temperature attributes.

[0071] The temperature and humidity sensor in the IoT front-end sensing network is a digital composite sensor with a temperature measurement range of -40℃ to 80℃ and an accuracy of ±0.5℃; the humidity measurement range is 0% to 100% relative humidity with an accuracy of ±3%. The sensor is fixedly installed on the upper inner wall of the trash can, 5 cm from the opening, avoiding direct contact with the trash to ensure accurate measurement of ambient air parameters. The edge computing unit sends a data acquisition command to the temperature and humidity sensor every 5 minutes via a single-bus interface. The sensor returns a 16-bit digital signal, with the high 8 bits representing humidity and the low 8 bits representing temperature. Verification and checks are performed on the received data to confirm its validity. To eliminate the influence of instantaneous fluctuations, a mean-mode filter is applied to three consecutive data acquisitions to obtain stable ambient temperature and humidity readings. For abnormal temperature and humidity values ​​(temperature <-10℃ or >50℃, humidity <10% or >95%), a self-test program is initiated and data is re-acquired. Three consecutive abnormal values ​​indicate a sensor malfunction.

[0072] The degree of waste thermal decomposition activity was assessed based on three-dimensional temperature-scale voxel data. First, the difference between the waste voxel temperature and the ambient temperature, ΔT = Tv - Te, was calculated, where Tv represents the voxel temperature (°C) and Te represents the ambient temperature (°C). Based on the ΔT value, the waste voxels were divided into four categories: inert zone (ΔT < 2°C), low-activity zone (2°C ≤ ΔT < 5°C), medium-activity zone (5°C ≤ ΔT < 10°C), and high-activity zone (ΔT ≥ 10°C). The proportion of voxels in each active zone to the total number of filling voxels was calculated, resulting in three proportionality coefficients: low-activity proportion Pl, medium-activity proportion Pm, and high-activity proportion Ph. Hotspot clustering analysis was performed, using the DBSCAN algorithm to identify temperature anomaly clusters. The clustering parameters were set as follows: neighborhood radius of 12 cm, minimum number of points of 5, and the identified cluster centers were marked as hotspots. For each hotspot, the difference between its highest temperature Tmax and the ambient temperature, ΔTmax, was calculated, and the hotspot volume Vh was recorded. Finally, based on the proportionality coefficient and hotspot parameters, the dynamic thermal decomposition degree H is calculated as H = 0.2 × Pl + 0.5 × Pm + 1.0 × Ph + 0.05 × n × (ΔTmax / 10) × (Vh / 100), where n is the number of hotspots.

[0073] Risk compensation is applied to dynamic thermal decomposition data based on ambient humidity. First, the humidity risk factor R is calculated as: R = 1 + [(RH - 50) / 100]. 2 RH represents the percentage of relative humidity. When RH < 50%, a dry environment inhibits decomposition but increases the risk of fire; when RH > 50%, a humid environment accelerates the decomposition of organic matter and the generation of gas. Subsequently, the total filling rate F = Vf / Vt is calculated, where Vf is the total volume of filled voxels (cubic centimeters), and Vt is the total volume of the garbage bin (cubic centimeters). The F value is divided into four intervals corresponding to different basic urgency coefficients Cf: light filling (F < 30%) corresponds to Cf = 0.6, moderate filling (30% ≤ F < 60%) corresponds to Cf = 1.0, heavy filling (60% ≤ F < 85%) corresponds to Cf = 1.5, and critical filling (F ≥ 85%) corresponds to Cf = 2.5. A comprehensive calculation yields the dynamic urgency index U = H × R × Cf, where H is the degree of dynamic thermal decomposition, R is the humidity risk factor, and Cf is the filling urgency coefficient. When the calculated result U exceeds the warning threshold of 12, an abnormal status alarm for the trash can is triggered, and the decision to use it is executed by the dispatch center through the smart city IoT device data acquisition and management platform.

[0074] Preferably, the dynamic thermal decomposition degree assessment data based on three-dimensional temperature-scale voxel data and ambient temperature includes:

[0075] Based on the three-dimensional temperature scale voxel data, a continuous region with a temperature 5°C or higher than the ambient temperature is extracted as the effective heat-generating region.

[0076] The average filling density of waste in the effective heat-generating area was calculated using three-dimensional structure voxel data of the waste.

[0077] When the average filling density in the effective heating area is greater than the preset first content threshold, the risk weight factor of the contents in the effective heating area of ​​the trash can is set to 1.5, which is recorded as the high-risk organic matter weight.

[0078] When the average filling density in the effective heating area is less than the preset second content threshold, the content risk weight factor of the effective heating area in the trash can is set to 0.5, which is recorded as the low-risk organic matter weight; wherein, the preset first content threshold is greater than the preset second content threshold;

[0079] The voxel activity scores of the effective heat-generating areas are processed based on the weights of high-risk and low-risk organic matter, and then the highest activity score among the voxels is extracted as dynamic thermal decomposition degree data of the activity level of waste decomposition.

[0080] In this embodiment of the invention, temperature anomalous voxels are first extracted from the three-dimensional temperature scale voxel data. The measured ambient temperature Te is read, and a temperature screening threshold of Te+5℃ is set. By comparing the temperature value Tv of each voxel with the screening threshold, voxels that meet the condition Tv>Te+5℃ are marked as thermal anomalous voxels. Subsequently, three-dimensional connected component analysis is performed, and all thermal anomalous voxels are clustered using the 26-connectivity criterion (considering the adjacency of voxel faces, edges, and vertices). The connected component analysis uses a depth-first search algorithm, starting from an unlabeled thermal anomalous voxel, recursively searching for all adjacent thermal anomalous voxels and marking them with the same region number. After the search is completed, connected components with more than 5 selected voxels are considered as effective heat-generating regions, and a unique region identifier is assigned to each effective heat-generating region. The three-dimensional coordinate range of the effective heat-generating region and the list of voxels it contains are stored in the region attribute table, and the average temperature, maximum temperature, and volume information of each region are recorded.

[0081] The effective heat-generating regions are spatially registered with the 3D structure voxel data of the waste using a region matching algorithm. First, a spatial index table is established, organizing each voxel in the waste 3D structure voxel data into a 10×10×30 3D grid structure for easy retrieval. For each effective heat-generating region R, all voxel locations within it are traversed, and the density value ρi (unit: g / cm³) for the corresponding location is extracted from the waste 3D structure voxel data. When a density record is not found in the original data for a certain location, the nearest neighbor interpolation method is used to obtain an estimated value from neighboring voxels. The arithmetic mean of all extracted density values ​​is calculated to obtain the average filling density ρavg of the region, R = ∑ρi / nR, where nR represents the number of voxels contained in region R. Extreme outliers (density values ​​less than 0.05 g / cm³ or greater than 2.0 g / cm³) are removed during the calculation to reduce the impact of measurement errors. For waste bins containing multiple effective heat-generating regions, the average filling density value is calculated and recorded separately for each region and stored in the density field of the region attribute table.

[0082] The initial content threshold was set at 0.8 g / cm³, obtained through laboratory testing. The testing process used representative organic waste samples (including food scraps, fruit peels, vegetable leaves, and other wet waste) at different compaction levels to determine the bulk density. When the average filling density ρavg of the effective heating zone, R, was greater than 0.8 g / cm³, a high-risk organic matter determination process was implemented. First, this area was marked as a high-density organic zone, and the zone type flag was updated in the zone attribute table. Then, the content risk weight factor WR = 1.5 was set for this zone, denoted as the high-risk organic matter weight. Simultaneously, the monitoring frequency for this zone was increased, shortening the temperature sampling interval from the standard 10 minutes to 5 minutes to more closely monitor the temperature change trend in the high-risk zone.

[0083] The second content threshold is preset to 0.3 g / cm³, determined by density measurement of dry waste (such as paper, plastic, and textiles). The first content threshold of 0.8 g / cm³ is ensured to be greater than the second content threshold of 0.3 g / cm³, forming a three-tiered density classification interval. When the average filling density ρavg and R of the effective heat-generating area are less than 0.3 g / cm³, the low-risk organic matter determination process is executed. This area is marked as a low-density area, and the area type flag is updated to 1 in the area attribute table. The content risk weight factor WR = 0.5 is set for this area, denoted as the low-risk organic matter weight. For areas with an average filling density between the second and first content thresholds (i.e., 0.3 g / cm³ ≤ ρavg, R ≤ 0.8 g / cm³), their content risk weight factor is set to 1.0, denoted as a medium-risk area. The area type flag is updated to 2 in the area attribute table, and the standard 10-minute temperature sampling interval is maintained.

[0084] Voxel activity scoring is performed on all effective heat-generating areas. For voxel i in each effective heat-generating area R, the temperature difference between it and the ambient temperature is calculated as ΔTi = Ti - Te, where Ti is the voxel temperature and Te is the ambient temperature. Then, a base activity score Si,base = 0.1 × ΔTi is calculated based on the temperature difference. When ΔTi ≥ 20℃, the base activity score is capped at 2.0. Further considering the temperature gradient factor, the maximum temperature difference ΔTi,max between the voxel and its six adjacent voxels is calculated, and a gradient activity component Si,grad = 0.05 × ΔTi,max is added. Based on the risk weighting factor WR, the comprehensive activity score Si = WR × (Si,base + Si,grad) is calculated. All voxels in all effective heat-generating areas are traversed, and the maximum comprehensive activity score Smax = max{Si} is found as the dynamic thermal decomposition degree data for waste decomposition activity. When multiple effective heat-generating areas exist within the waste bin, the global maximum activity score is taken as the final result. This value typically ranges from 0 to 5. A higher value indicates more intense thermal decomposition of waste and a more active decomposition process.

[0085] In one implementation of this invention, please refer to Figure 2 The image shows a line graph of voxel activity scores. For example, in a practical application at a smart waste treatment station, three-dimensional temperature scale voxel data was collected, and the ambient temperature Te was measured to be 25℃. Figure 2 As shown, a temperature screening threshold of Te+5℃ = 30℃ was set, and voxel temperatures were screened one by one. In a typical trash can monitoring case, a total of 1847 anomalous voxels with temperatures ranging from 30℃ to 55℃ were detected. A 26-connectivity criterion was used to perform three-dimensional connected component analysis, and 23 connected components were identified using a depth-first search algorithm. After voxel count screening (>5 voxels), eight effective heat-generating regions were finally identified, labeled R1-R8. The largest heat-generating region, R1, contained 312 voxels, with temperatures distributed between 32℃ and 48℃, and an average temperature of 39.5℃. A spatial registration algorithm was used to match the effective heat-generating regions with the three-dimensional structure voxel data of the trash. Taking heat-generating region R1 as an example, after extracting the density data at the corresponding location, two extreme outliers (density <0.05g / cm³) were removed. 3 The average filling density ρavg was calculated, and R1 = 0.85 g / cm³. 3 .like Figure 2 The risk grading criteria shown have R1 = 0.85 g / cm³ due to ρavg. 3 >0.8g / cm 3(Based on the first content threshold), the system classifies region R1 as a high-risk organic matter area and sets the risk weighting factor WR1 = 1.5. Meanwhile, the average filling density of region R3 is 0.25 g / cm³. 3 <0.3g / cm 3 (Second content threshold), classified as a low-risk area, with a weighting factor WR3 = 0.5. The density of area R2 is 0.6 g / cm³. 3 The voxel, falling between two thresholds, is classified as a medium-risk area with a weighting factor WR2 = 1.0. Voxel activity scoring is performed on each effective heat-generating area. Taking the voxel with the highest temperature in region R1 as an example, its temperature Ti = 48℃, and the temperature difference ΔTi = 48 - 25 = 23℃. Since ΔTi > 20℃, the base activity score reaches the capped value Si, base = 2.0. Further calculation of the temperature gradient factor reveals that the maximum temperature difference between this voxel and its six adjacent facet voxels, ΔTi,max = 12℃, results in a gradient activity component of 0.05 × 23 = 1.15. Combining this with the high-risk weighting factor WR1 = 1.5, the comprehensive activity score for this voxel is: Si = (2.0 + 1.15) × 1.5 = 4.73. Figure 2 The high-risk weighted curve (red line) shows an activity score of approximately 4.7 at a temperature of 48℃, which is highly consistent with the calculated results. Traversing all voxels of all eight effective heat-generating regions, the highest activity scores for each region were calculated as follows: R1 = 4.73, R2 = 3.12, R3 = 1.85, R4 = 2.96, R5 = 3.45, R6 = 2.21, R7 = 4.15, and R8 = 3.78. Through comparison of monitoring data over seven consecutive days, the accuracy rate of assessing the thermal decomposition degree of garbage cans using the method of this invention reached 92.3%, a 35% improvement compared to traditional single-point temperature monitoring methods. Figure 2 The multi-weighted factor scoring mechanism shown effectively distinguishes the decomposition activity levels of different types of waste.

[0086] Preferably, environmental risk compensation is performed based on dynamic thermal decomposition data using ambient humidity, and the urgency of the bin's condition is calculated based on the total waste filling rate, including:

[0087] Gain compensation is applied to the dynamic thermal decomposition degree data based on ambient humidity to generate compensated thermal decomposition degree data.

[0088] Real-time gas concentration values ​​are obtained using gas sensors in the front-end sensing network of the Internet of Things (IoT).

[0089] Acquire local meteorological data and extract the current ambient wind speed as the local ambient wind speed value;

[0090] Determine whether the real-time gas concentration value is greater than 50 ppm and the local ambient wind speed value is less than 1 m / s. If yes, set the influence coefficient to 1.4; otherwise, set it to 1.0 to obtain the odor diffusion influence coefficient.

[0091] The public health sensitivity of the data on the degree of thermal decomposition was processed by using the odor diffusion influence coefficient, and the urgency of the state inside the bin was calculated based on the total waste filling rate, thus obtaining the dynamic urgency index.

[0092] In this embodiment of the invention, the ambient humidity value RH (relative humidity percentage) is obtained. A piecewise function compensation model is applied to process the dynamic thermal decomposition degree data H according to different humidity ranges. When the ambient relative humidity is below 30%, a low humidity compensation coefficient KL = 0.85 is used to suppress thermal decomposition activity; when the ambient relative humidity is between 30% and 70%, a linear transition compensation coefficient KM = 0.85 + 0.015 × (RH - 30) is used; when the ambient relative humidity is above 70%, a high humidity compensation coefficient KH = 1.45 + 0.01 × (RH - 70) is used to accelerate thermal decomposition activity. The compensation calculation formula is HC = H × K, where HC is the compensated thermal decomposition degree data, H is the original dynamic thermal decomposition degree data, and K is the compensation coefficient for the corresponding humidity range. When the ambient relative humidity exceeds 85% and the original thermal decomposition degree data is greater than 2.5, additional anaerobic fermentation risk correction is performed, adding a fixed compensation value of 0.8 to the compensation result.

[0093] The gas sensor array in the IoT front-end sensing network comprises three independent sensing units: a methane sensor (model MQ-4), an ammonia sensor (model MQ-137), and a hydrogen sulfide sensor (model MQ-136). The sensor array is connected to the edge computing unit via a dedicated acquisition circuit, with a sampling frequency set to once every 2 minutes. During acquisition, the edge computing unit sequentially selects the three sensor channels using a multiplexer, performing a 16-bit ADC conversion on the analog voltage signal output by each sensor. Each sensor acquires data five times, removing the highest and lowest values ​​and taking the average to eliminate the influence of instantaneous fluctuations. The converted digital signal is then converted into the actual gas concentration using a pre-calibrated sensor response curve. The conversion formula is C = a × (V / V0). b Where C is the gas concentration (ppm), V is the current ADC reading, V0 is the ADC reading under clean air, and a and b are sensor calibration coefficients. For the concentrations of methane, ammonia, and hydrogen sulfide, a weighted average method is used to calculate the comprehensive gas index, with weights of 0.3, 0.4, and 0.3 respectively, yielding the real-time gas concentration value G in ppm. When the concentration of any single gas exceeds the preset alarm threshold (methane 10000 ppm, ammonia 100 ppm, hydrogen sulfide 50 ppm), a gas exceedance alarm is immediately triggered.

[0094] Local weather information is obtained through the meteorological data interface of the city's IoT data center. The MQTT protocol is used to subscribe to meteorological data packets released by the city's meteorological service every hour on the hour. These packets contain parameters such as temperature, humidity, air pressure, wind direction, and wind speed, and are in JSON format. The edge computing unit parses the received packets and extracts the wind speed field value as the local environmental wind speed benchmark. Due to the distance difference between the city's meteorological station and the actual location of the trash can, a geographic information mapping table is used for location correction. The correction formula is WS = WSbase × Kara, where WS is the corrected local environmental wind speed value (m / s), WSbase is the wind speed benchmark value measured by the meteorological station (m / s), and Kara is the regional terrain correction coefficient. The regional terrain correction coefficient is obtained through pre-analysis of wind speed attenuation characteristics in different areas of the city and is stored in the regional parameter table of the edge computing unit. If the meteorological data interface has not been updated for more than 15 minutes, the most recent valid data is used with a 10% uncertainty correction. When the local environmental wind speed value is less than 0.2 m / s, it is considered a windless state, and the wind speed value is fixed at 0.2 m / s.

[0095] The real-time gas concentration value G is compared with a threshold of 50 ppm, and the local ambient wind speed value WS is compared with a threshold of 1 m / s. The threshold of 50 ppm is determined based on the human olfactory perception threshold and public area odor control standards, while the threshold of 1 m / s is determined based on the critical wind speed value in the outdoor gas diffusion dynamics model. A four-branch judgment logic is designed: when G > 50 ppm and WS < 1 m / s, the conditions of high gas concentration and low wind speed are met, and the odor diffusion influence coefficient ID is set to 1.4; when G > 50 ppm but WS ≥ 1 m / s, the high-concentration gas can diffuse effectively under sufficient wind speed, and ID is set to 1.2; when G ≤ 50 ppm but WS < 1 m / s, although the gas concentration is low, diffusion is limited, and ID is set to 1.1; when G ≤ 50 ppm and WS ≥ 1 m / s, the low-concentration gas diffuses sufficiently under sufficient wind speed, and ID is set to 1.0.

[0096] The odor diffusion influence coefficient ID was used to process the compensated thermal decomposition data HC for public health sensitivity, calculated as HS = HC × ID, where HS is the sensitivity-processed thermal decomposition data. Subsequently, the total fill rate F was calculated. The fill rate F = Nfilled / Ntotal × 100 was obtained by dividing the number of filled voxels Nfilled in the three-dimensional structure voxel data of the waste by the total number of voxels in the waste bin space. The filling urgency coefficient CF was determined based on the fill rate F. When F < 40%, CF = 0.8; when 40% ≤ F < 65%, CF = 1.0 + 0.01 × (F - 40); when 65% ≤ F < 85%, CF = 1.25 + 0.0375 × (F - 65); when F ≥ 85%, CF = 2.0. A comprehensive calculation is performed to obtain a dynamic urgency index. When U > 5, the trash can is marked as requiring attention; when U > 8, it is marked as needing to be emptied as soon as possible; and when U > 12, it is marked as needing urgent emptying. The dynamic urgency index U and the corresponding status markings are uploaded in real time to the smart city IoT device data acquisition and management platform via wireless network.

[0097] Preferably, step S3, which involves weighted processing of the urgency of waste collection based on a dynamic urgency index, includes:

[0098] Obtain the functional area category labels of trash cans; match regulatory weight coefficients based on the functional area category labels; where functional area category labels include commercial areas, residential areas, and public places;

[0099] The urgency of waste disposal is calculated by weighting the regulatory weight coefficient and the dynamic urgency index, and the priority score is normalized to generate waste disposal score data.

[0100] Set the collection and transportation start threshold to 75, filter out trash cans with a collection and transportation score data greater than 75 to obtain a high-priority site list; otherwise, obtain a low-priority site list.

[0101] The site with the highest waste disposal score is selected from the list of high-priority sites as the initial center of the regional cluster and the cluster generation center point.

[0102] Using the cluster generation center point as the center, a radius of 1 kilometer is defined on the map. All high-priority sites and some secondary-priority sites falling within this range are aggregated to obtain the potential cluster member set.

[0103] Based on the three-dimensional structure voxel data of the waste, the estimated total volume and weight of the waste to be transported are calculated to obtain the estimated load of the cluster.

[0104] Match the target collection vehicle that is closest to the cluster generation center point;

[0105] The estimated load of the cluster is compared with the carrying capacity of the target collection vehicle. If the load exceeds the limit, the site with the lowest urgency is removed from the potential cluster member set until the condition is met, thus generating a cluster of sites to be processed.

[0106] In this embodiment of the invention, an electronic map database of urban functional areas is established within the smart city IoT management platform. This database contains detailed regional functional attribute information and is indexed by latitude and longitude coordinates. Based on the location coordinates (accuracy ±3 meters) uploaded by the GPS positioning module built into the trash can, the corresponding functional area category tags are extracted using a spatial query matching algorithm. The functional area category tags are divided into three categories: commercial areas (including shopping malls, office buildings, restaurant streets, etc.), residential areas (including residential communities, apartments, community centers, etc.), and public places (including parks, squares, stations, schools, etc.). Based on the pedestrian density, waste generation characteristics, and regulatory requirements of different functional areas, corresponding regulatory weight coefficients are set: a weight coefficient of 1.4 for commercial areas, 1.0 for residential areas, and 1.2 for public places. These weight coefficients are derived from statistical analysis of the city sanitation department's collection records and complaint hotspot distribution over the past three years.

[0107] The weighted collection urgency is calculated based on the regulatory weight coefficient W and the dynamic urgency index U. The calculation formula is P = U × W × 10, where P is the original collection urgency score, U is the dynamic urgency index, W is the regulatory weight coefficient, and 10 is the base score adjustment factor. Priority score normalization is then performed using the Min-Max standardization method, with the formula S = (P - Pmin) / (Pmax - Pmin) × 100, where S is the normalized collection score, P is the original collection urgency score, Pmin is the minimum original collection urgency score among all current trash cans, and Pmax is the maximum. When there are fewer than 10 trash cans, Pmin is fixed at 0, and Pmax is fixed at 150 to avoid fluctuations in normalization due to small samples. To ensure the real-time validity of the collection score data, all trash cans in the network are recalculated and sorted every 30 minutes to generate updated collection score data. The calculation results retain the integer part, ranging from 0 to 100.

[0108] A collection and transportation initiation threshold of 75 points is set. This threshold is determined based on the service standards set by the city's sanitation department, indicating that collection should be prioritized when the collection score reaches or exceeds this value. A data filtering operation is performed, comparing the collection score data of all trash cans with the threshold of 75. For trash cans with a collection score greater than 75, their equipment identification number, geographic coordinates, collection score data, and functional area category label are organized into structured data entries and added to the high-priority site list. For trash cans with a collection score less than or equal to 75, data entries of the same format are added to the second-priority site list. Both lists are sorted in descending order of collection score data for easier subsequent processing. A 20-minute refresh cycle is set for the high-priority site list. When the high-priority site list is empty, the collection and transportation initiation threshold is temporarily lowered to 65, and the filtering operation is re-executed to ensure continuous collection operations.

[0109] From the list of high-priority sites, extract the site with the highest waste disposal score using a heap sort algorithm. Read the site's geographical coordinates (longitude λ, latitude λ). This is set as the initial center for regional clustering, i.e., the cluster generation center point. When multiple sites have the same collection score (score difference less than 0.5), a secondary sorting is performed, prioritizing sites with the functional area category label of commercial area, followed by public places, and lastly residential areas. If multiple sites are still tied, the site closest to the nearest collection vehicle is selected as the center point. A cluster identifier is created for the selected cluster generation center point, in the format of area code-date-serial number, such as HD-0718-01. The relevant information of the cluster generation center point (including site identifier, geographical coordinates, collection score, and functional area category label) is stored in the center point record table of the cluster management database, and the status is marked as active.

[0110] A spatial range query is performed with the cluster generation center point as the center. Geographical distance is calculated using latitude and longitude coordinates, employing the Havesing formula. Calculate the distance between two points, where d is the distance (in kilometers) and r is the Earth's radius (6371 kilometers). λ1 represents the latitude and longitude (in radians) of the center point. λ2 represents the latitude and longitude (in radians) of the point to be detected. All trash can sites within a distance of 1 km or less are selected, prioritizing all eligible sites from the high-priority site list for inclusion in the potential cluster member set. For sites in the secondary-priority site list, a multi-level selection criterion is applied: secondary-priority sites within 0.5 km of the center point must have a collection score greater than 65; secondary-priority sites between 0.5 and 0.8 km must have a collection score greater than 70; secondary-priority sites between 0.8 and 1 km are included only if their collection score is greater than 72 and their functional area category label matches that of the center point. All eligible sites are organized into a potential cluster member set, recording each member's site identifier, geographic coordinates, collection score, functional area category label, and distance from the center point.

[0111] Retrieve 3D structure voxel data of each trash can from the potential cluster member set from the database. Process the voxel data of each trash can sequentially, calculating its estimated collection volume and weight. For volume calculation, count the number n of non-zero density voxels in each trash can, multiply by the volume of a single voxel v (75 cubic centimeters), to obtain the trash volume V = n × v. For weight calculation, sum the density value ρi of each non-zero voxel, multiply by the voxel volume v, to obtain the trash weight m = ∑(ρi × v). Considering the trash compression factor, apply a compression correction factor of 0.85 to the volume calculation, obtaining the compressed volume V′ = V × 0.85. Summarize the volume and weight data of all trash cans in the potential cluster member set, and calculate the total volume Vtotal = ∑V′ and the total weight mtotal = ∑m. Combine these two indicators to generate a cluster estimated load representation in the format "Total Weight: Total Volume", for example, "350 kg: 0.8 cubic meters".

[0112] Maintain a real-time updated database of collection vehicle status, recording each vehicle's vehicle number, model specifications, maximum load capacity, maximum loading volume, current location coordinates, and working status. Using the cluster generation center point as a reference point, calculate the distance between all idle collection vehicles and that point. The distance calculation also uses the Havesing formula, with results accurate to meters. Sort all idle vehicles in ascending order of distance, prioritizing the closest vehicle. If multiple idle vehicles exist within a 2-kilometer radius, perform a multi-factor scoring: Vehicle distance score = 100 - distance (km) × 20, Load matching score = 100 - |Maximum load capacity - Estimated total weight| / Estimated total weight × 100 (maximum cap is 100), Total score = 0.6 × Distance score + 0.4 × Load matching score. Select the vehicle with the highest total score as the target collection vehicle. When all vehicles are busy, the system will query the vehicle that is expected to complete its current task earliest, calculate its expected idle time and the urgency of the current cluster. If the waiting time exceeds 30 minutes and the cluster contains stations with a score of over 90, the emergency dispatch mechanism will be triggered.

[0113] The maximum load capacity (mmax, kg) and maximum loading volume (Vmax, m³) of the target collection vehicle are retrieved from the vehicle database. The total weight (mtotal) and total volume (Vtotal) in the estimated cluster load are compared with the vehicle parameters. When mtotal > mmax or Vtotal > Vmax, a cluster member adjustment operation is performed. First, all stations are sorted in ascending order of collection score from the potential cluster member set. Then, stations are removed one by one from the bottom, while decreasing the total weight and total volume values. After each station is removed, the mtotal and Vtotal of the remaining stations are recalculated and compared with mmax and Vmax until the conditions of mtotal ≤ mmax and Vtotal ≤ Vmax are met. To ensure a reasonable number of stations within the cluster, a minimum of 3 stations and a maximum of 15 stations are set. If the number of stations meeting the load condition exceeds 15, only the 15 stations with the highest scores are retained. If the load condition is still not met after removing low-scoring stations and the number of remaining stations has reached 3, the target collection vehicle will be replaced with a vehicle model with a larger load capacity. The final set of sites constitutes a cluster of sites to be processed, generating a collection task order containing detailed information on all sites within the cluster, and assigning the task to the target collection vehicle.

[0114] Preferably, step S3 includes the following steps:

[0115] Step S31: Take the trash can with the highest priority in the coordinate data of the site to be visited as the first target point of the target collection vehicle. Then, starting from the first target point, query the real-time driving time of all remaining points in the coordinate data of the site to be visited through the preset map service interface to obtain the time taken for the candidate points.

[0116] Step S32: Select the point with the shortest time from the candidate points and determine it as the next optimal work point for the vehicle;

[0117] Step S33: Add the next optimal work point to the task list of the current vehicle and remove the point from the coordinate data of the site to be visited to obtain the updated work path;

[0118] Step S34: Repeat steps S31 to S33 until the capacity or shift duration of the target collection and transportation vehicles is met, and use the updated operation route as a feasible operation sequence.

[0119] Step S35: Adjust the priority of the feasible operation sequence by secondary weighting based on the collection and transportation score data to obtain the theoretically optimal collection and transportation sequence.

[0120] In this embodiment of the invention, the garbage bin station with the highest collection score is extracted from the cluster of sites to be processed, and its geographical coordinates (longitude λ1, latitude λ2) are determined. The first target point for the target collection vehicle is specified. A batch route planning request is constructed by calling the preset Gaode Map service interface via a RESTful API. Request parameters include: starting point coordinates (latitude and longitude coordinates of the first target point), destination coordinate set (latitude and longitude coordinates of all remaining points in the target site coordinate data), driving strategy (set to shortest time), and vehicle type (set to heavy truck based on the target collection vehicle model). An HTTP POST request is sent, carrying a JSON data body in the specified format, and the request timeout is set to 3 seconds. The map service interface returns JSON format data containing all route options. The estimated driving time (in seconds), distance (in meters), and real-time traffic information for each route are extracted using a JSON parser. If the interface call fails, a backup offline route calculation module is started, estimating the driving time based on Euclidean distance and historical average driving speed (18 km / h).

[0121] A minimum value search algorithm is applied to the candidate location time table. First, a hash table structure is built with the trash can station identifier as the key and driving time as the value. Then, a linear scan is performed to record the index of the location with the shortest driving time. When multiple locations are found to have driving times differing by no more than 60 seconds, a multi-factor evaluation mechanism is triggered: the comprehensive score S = 0.7 × (1 - T / Tmax) + 0.3 × (P / Pmax) is calculated, where T is the driving time (seconds), Tmax is the maximum driving time among all candidate locations, P is the location's waste collection score, and Pmax is the highest waste collection score among all candidate locations. The location with the highest comprehensive score S is selected as the next optimal work point. If the comprehensive scores are the same, the location with the higher waste collection score is prioritized. Road congestion is also considered; when the congestion index of a route exceeds 0.8, the driving time of that route is penalized by 30% in the calculation. Finally, the coordinates (longitude λ2, latitude λ2) of the station with the shortest driving time or the highest comprehensive score under the comprehensive conditions are determined. () is taken as the next optimal work point.

[0122] The detailed information of the next optimal work point is added to the current vehicle's task list. The task list is stored in a linked list structure, with each node containing five fields: site identifier, geographic coordinates, estimated arrival time, estimated operation duration, and collection score. The estimated operation duration is calculated based on the 3D structure voxel data of the garbage bins using the formula t = 2 + V / 40, where t is the estimated operation duration (minutes), V is the garbage volume (cubic meters), 2 is the basic operation time (minutes), and 40 is the standard collection rate (cubic meters / minute). The vehicle's current location is updated to the coordinates of the newly added work point, and the cumulative travel time and distance are calculated. Simultaneously, the location information is removed from the coordinate data structure of the sites to be visited using a marked deletion method, and the status field of the corresponding entry is set to "assigned". The total volume and weight of the remaining sites to be visited are recalculated, and the vehicle's remaining capacity information is updated. After the operation is completed, an operation log is generated, containing a timestamp, vehicle number, work point information, and remaining capacity data, and stored in the log database to form the updated operation path.

[0123] The iterative algorithm is executed, encapsulating steps S31 to S33 into a path planning function, which is repeatedly called to allocate points. Before each iteration, two termination conditions are checked: the capacity limit of the target collection vehicle and the shift duration limit. The capacity limit check formula is Vc + Vnext ≤ Vmax and Mc + Mnext ≤ Mmax, where Vc is the current total volume of allocated waste, Vnext is the volume of waste at the next candidate point, Vmax is the maximum vehicle capacity, Mc is the current total weight of allocated waste, Mnext is the weight of waste at the next candidate point, and Mmax is the maximum vehicle load. The shift duration limit check formula is Tc + Tnext + Toperation ≤ Tmax, where Tc is the current cumulative time, Tnext is the driving time to the next point, Toperation is the estimated working time at the next point, and Tmax is the maximum working time of the shift (480 minutes). The iteration process terminates when either condition is not met. After each iteration, the vehicle load status and remaining capacity are updated, and the total travel distance and total time are recalculated. The iteration terminates when the coordinate data of the site to be visited is empty or all points have been determined to be unavailable. After all iterations are completed, the generated task list is marked as a feasible operation sequence, and the total travel distance, total travel time, and total operation time of the sequence are calculated.

[0124] The average collection score for all locations in the sequence is calculated, and then key locations with collection scores significantly higher than the average are identified. For each key location, the difference Δi = i - i' between its current location index i and its ideal location index i′ is calculated, where the ideal location index is determined based on the collection score ranking. When Δi > 3, a sequence reordering mechanism is triggered, moving the key location forward by up to 3 positions and simultaneously moving the affected locations backward. The reordering process employs a local optimization strategy to ensure that the total travel distance after adjustment does not increase by more than 8% of the original plan. Time window constraints are also considered; for trash cans in commercial areas and public places, peak-hour (7:00-9:00 and 17:00-19:00) avoidance rules are added to adjust sensitive locations during these periods to more suitable times. After all adjustments are completed, all indicators for the entire sequence are recalculated to generate the theoretically optimal collection sequence, and the complete route navigation information is pushed to the onboard terminal equipment of the target collection vehicles via a wireless network.

[0125] Preferably, step S4 includes the following steps:

[0126] Step S41: Plan the navigation path for the theoretically optimal collection sequence and process the segmented navigation instructions to obtain segmented navigation instructions;

[0127] Step S42: Based on each target station in the segmented navigation instructions, fill in the structural features through intelligent analysis of the three-dimensional structure voxel data of the garbage, and generate a single-point detailed operation annotation;

[0128] Step S43: Combine the segmented navigation instructions with the single-point detailed operation annotations for each station and send them to the terminal of the target collection vehicle to manage the smart garbage bin collection operation.

[0129] In this embodiment of the invention, the list of station coordinates in the theoretically optimal collection sequence is extracted into an ordered latitude and longitude array, containing the current location coordinates of the collection vehicle as the starting point and the coordinates of all stations to be visited. A complete path is generated by calling the driving route planning interface of the map open platform. The interface call uses a batch request method, processing a maximum of 10 waypoints at a time; if more than 10 points are needed, they are processed in batches. The driving strategy parameter is set to the shortest time, the vehicle type parameter is set to heavy truck, and the avoidance area parameter includes restricted areas and congested road sections. The JSON data returned by the interface contains the coordinate point set of the complete driving path, turning prompts, road names, and travel direction information. Navigation instructions are processed in segments, dividing the continuous path into multiple independent road segments according to the stations. Each road segment constitutes a segmented navigation instruction. The segmented navigation instruction includes: starting point coordinates, ending point coordinates, a set of waypoint coordinates (one point every 10 meters), a set of turning points, estimated travel time, estimated travel distance, road type, and traffic status, among other key information.

[0130] For each target site in the segmented navigation instructions, a three-dimensional structure voxel data analysis of the waste is performed. The latest three-dimensional structure voxel data of the waste at each target site is retrieved from the database, and a three-dimensional morphological feature extraction algorithm is executed. First, a histogram of the waste filling height distribution is calculated to identify areas with concentrated filling height. Then, density clustering analysis is performed, dividing voxels into high-density areas (ρ>1.0 g / cm³), medium-density areas (0.5≤ρ≤1.0 g / cm³), and low-density areas (ρ<0.5 g / cm³). Based on the clustering results, filling characteristic labels are generated, including concentrated heavy materials (bottom high-density proportion >60%), predominantly light materials (low-density area proportion >70%), mixed filling (similar proportions in each density area), and presence of foreign materials (detection of single-point density anomalies). Furthermore, waste activity is analyzed based on dynamic thermal decomposition data; when the thermal decomposition degree data is >5, it is marked as high-temperature activity. Based on the above analysis, structured single-point detailed operation notes are generated, including estimated waste volume, estimated weight, physical property markings, treatment difficulty level (1-5), and special treatment prompts.

[0131] The segmented navigation instructions are matched and merged with the detailed operation annotations for each single point. A unified operation instruction data structure is established, comprising three parts: basic route information block, station operation information block, and control information block. For each target station, the road segment data in the navigation instruction is associated with the detailed operation annotations for the corresponding station to form a complete station operation instruction package. The operation instruction package is pushed to the vehicle-mounted terminal equipment of the target collection vehicle via a 4G / 5G wireless communication network using the MQTT protocol. The push process is set to a service quality level of QoS2 (ensuring one-time and only-once delivery) and 256-bit AES encryption is enabled to protect data security. After receiving the operation instruction package, the vehicle-mounted terminal automatically parses it and loads the navigation information into the vehicle navigation system. At the same time, the detailed operation annotations for the station are displayed in a card layout on the driver's operating interface. The execution status of the instructions is monitored in real time. When the vehicle arrives at the station, the terminal automatically displays the corresponding station's operation annotations at the top and reminds the driver of special handling requirements through voice prompts. After the vehicle completes the station operation, the driver confirms the completion status on the terminal, and the navigation instructions and operation annotations for the next station are automatically loaded, achieving intelligent management throughout the entire process.

[0132] Of particular importance is that, based on the segmented navigation instructions, each target station's structural features are filled in through intelligent analysis of garbage 3D structure voxel data, including...

[0133] Based on the three-dimensional structure voxel data of waste, voxel sets with a density value less than the preset threshold of 0.3 kg / m³ and spatial continuity are identified and extracted, and these sets are marked as internal structural cavities.

[0134] Calculate the ratio of the total volume of the internal structural cavity to the effective volume of the trash can, and the vertical position distribution in the three-dimensional structural voxel data of the trash can, to assess the structural instability risk index due to gravity during vehicle bumps or tipping.

[0135] The centroid position of the entire garbage pile is calculated using the voxel density and three-dimensional coordinates in the three-dimensional structure voxel data of the garbage. It is then determined whether the horizontal offset of the centroid from the geometric center of the garbage bin is greater than 30% of the bin radius. If so, it is determined that there is an off-center load and an off-center load imbalance state marker is obtained.

[0136] By combining the structural instability risk index and the off-center load imbalance state marker, the impact force generated by the target collection vehicle when dumping the garbage bin is quantified, and the estimated dynamic unloading impact value is generated.

[0137] Intelligently generate detailed operation notes for each point based on the estimated dynamic unloading impact value.

[0138] In this embodiment of the invention, the latest three-dimensional structural voxel data of the target site is extracted from the database, including the three-dimensional coordinates (x, y, z) and corresponding density value ρ of each voxel. A preset threshold ρ0 = 0.3 kg / m³ is set as the cavity judgment standard. This threshold is obtained based on laboratory tests and represents the critical density value of loosely piled garbage. A three-dimensional region growing algorithm is executed to scan layer by layer starting from the bottom of the garbage bin. When a voxel with a density value ρ < ρ0 is found, it is marked as a candidate cavity point. Subsequently, a 26-connectivity check (considering face contact, edge contact, and vertex contact of voxels in three-dimensional space) is performed on each candidate cavity point, and voxels that are spatially adjacent and whose density values ​​are all less than the threshold are grouped into the same cavity set. Small cavities with fewer than 10 voxels are eliminated, and only large cavities with structural significance are retained. For each identified internal structural cavity, a unique identifier is assigned, and its voxel coordinate set, total volume, shape feature parameters, and spatial location information in the garbage bin are recorded. The total volume Vc of the internal structural cavities is calculated by multiplying the number of voxels marked as cavities by the volume v (cubic centimeters) of a single voxel. The effective volume Vt (cubic meters) of this model of trash can is retrieved from the equipment parameter library, and the cavity proportion r = Vc / Vt is calculated. Simultaneously, the vertical distribution characteristics of the cavities are analyzed, and the height of the trash can is divided into three equal layers (lower, middle, and upper), and the volume proportions rl, rm, and ru of each layer are statistically analyzed. A structural instability risk assessment is performed. First, the basic risk factor R0 = r × 10 is calculated, with a value ranging from 0 to 10. Then, adjustments are made based on the vertical distribution: when the lower layer cavity proportion rl > 0.5, the risk coefficient increases by 30%; when the middle layer cavity proportion rm > 0.4, the risk coefficient increases by 50%; and when there are continuous cavities penetrating two or more layers, the risk coefficient increases by 70%. The final structural instability risk index Rs is calculated as Rs=R0×(1+0.3×I1+0.5×I2+0.7×I3), where I1, I2, and I3 are Boolean indicator variables, which take the value 1 when the corresponding conditions are met, and 0 otherwise.

[0139] The centroid position of the entire waste pile is calculated based on the three-dimensional structure voxel data of the waste. For each non-cavity voxel i, its three-dimensional coordinates (xi, yi, zi) and density value ρi are obtained. The centroid coordinate calculation formulas are: xc = ∑(xi × ρi × v) / M, yc = ∑(yi × ρ × v) / M, zc = ∑(zi × ρi × v) / M, where v is the volume of a single voxel, and M = ∑(ρi × v) is the total mass of the waste. The calculated centroid coordinates (xc, yc, zc) are taken with the center of the bottom of the waste bin as the origin, the positive x-axis direction to the right, the positive y-axis direction forward, and the positive z-axis direction upward. The horizontal offset of the centroid is calculated. Where (x0, y0) are the coordinates of the geometric center of the trash can. The radius R of the trash can is obtained from the equipment parameter library, and the relative offset ratio p = d / R is calculated. When p > 0.3 (i.e., the horizontal offset is greater than 30% of the can radius), it is determined that the trash can is unbalanced, and the unbalanced state flag B = 1 is set; otherwise, B = 0 is set. The unbalanced direction angle θ = arctan((yc―y0) / (xc―x0)) is also recorded.

[0140] The static foundation impact force F0 is calculated as follows: F0 = M × g × h / t, where M is the total mass of the garbage (kg), g is the acceleration due to gravity (9.8 m / s²), h is the tipping height of the garbage bin (m), and t is the typical tipping time (seconds). When there is an internal structural cavity, the additional impact caused by the collapse effect is considered, and the structural instability gain coefficient Ks = 1 + 0.15 × Rs is calculated. When there is an off-center load, the lateral moment caused by unbalanced rotation is considered, and the off-center load gain coefficient Kb = 1 + 0.2 × B × p is calculated, where p is the relative offset ratio. Taking both factors into account, the dynamic unloading impact value F = F0 × Ks × Kb is estimated. The F-value is mapped to a 5-level quantization scale: F < 500 Newtons is Level 1 (mild impact), 500 ≤ F < 1000 Newtons is Level 2 (moderate impact), 1000 ≤ F < 2000 Newtons is Level 3 (strong impact), 2000 ≤ F < 3500 Newtons is Level 4 (strong impact), and F ≥ 3500 Newtons is Level 5 (extremely strong impact).

[0141] For sites with an impact value of Level 1, a standard operating procedure prompt for regular dumping is generated. For Level 2 sites, a slow dumping prompt is generated and marked with a yellow warning. For Level 3 sites, a segmented dumping prompt is generated, suggesting tilting at a 45-degree angle and pausing for 3 seconds until the contents stabilize before continuing to dump. For Level 4 sites, a prompt for multiple small-angle dumping is generated, suggesting repeated small-angle shaking before complete dumping. For Level 5 sites, a special handling prompt is generated, marked with a red warning, suggesting using an auxiliary push rod to assist unloading. When the unbalanced state is marked B=1, an unbalanced load direction prompt is added to the operating notes; for example, if the right side's center of gravity shifts, adjust the position to the left before dumping. When the internal cavity is located in the lower middle layer and the cavity ratio r>0.2, a special reminder is added regarding internal looseness and the possibility of sudden material slippage during dumping. These detailed operating notes are integrated into structured text and diagrams for collection and transportation personnel to refer to and implement.

[0142] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0143] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An implementation method of a smart city Internet of Things device data collection management platform, characterized in that, Includes the following steps: Step S1: Deploy IoT sensing devices inside the trash can to build an IoT front-end sensing network; Data on incremental throwing events in trash cans is collected through the Internet of Things front-end sensing network, and the changes in structural voxels in trash cans caused by incremental throwing events are analyzed to generate three-dimensional structural voxel data of trash. Step S2: Calculate the total waste filling rate based on the three-dimensional structure voxel data of the waste; The dynamic thermal decomposition level is assessed based on the three-dimensional structure voxel data of the waste, and the urgency of the in-bin condition is calculated based on the total waste filling rate to obtain the dynamic urgency index; which includes: Infrared thermal imaging sensors in the front-end sensing network of the Internet of Things are used to collect temperature distribution information inside the trash can, and infrared temperature data inside the can is obtained. Using the three-dimensional structure voxel data of the waste as a spatial reference, pixel-level registration and overlay are performed based on the infrared temperature data inside the bin to generate three-dimensional temperature scale voxel data. Ambient temperature and humidity are obtained by using temperature and humidity sensors in the front-end sensing network of the Internet of Things. The dynamic thermal decomposition degree was assessed based on three-dimensional temperature-scale voxel data and ambient temperature data; this included: Based on the three-dimensional temperature scale voxel data, a continuous region with a temperature 5°C higher than the ambient temperature was extracted as the effective heat-generating region. The average filling density of waste in the effective heat-generating area was calculated using three-dimensional structure voxel data of the waste. When the average filling density in the effective heating area is greater than the preset first content threshold, the risk weight factor of the contents in the effective heating area of ​​the trash can is set to 1.5, which is recorded as the high-risk organic matter weight. When the average filling density in the effective heating area is less than the preset second content threshold, the content risk weight factor of the effective heating area in the trash can is set to 0.5, which is recorded as the low-risk organic matter weight; wherein, the preset first content threshold is greater than the preset second content threshold; The voxel activity score of the effective heat-generating area is processed according to the weight of high-risk organic matter and low-risk organic matter, and then the highest activity score in the voxel is extracted as dynamic thermal decomposition degree data of the activity of waste decomposition. Environmental risk compensation is performed based on dynamic thermal decomposition data of ambient humidity, and the urgency of the state inside the bin is calculated based on the total waste filling rate to obtain the dynamic urgency index. The urgency of waste removal is assessed based on a dynamic urgency index, and then clusters of sites awaiting processing are identified. Step S3: Extract the coordinate data of the sites to be visited in the cluster of sites to be processed, and dynamically plan the theoretically optimal collection sequence of the target collection vehicles; Step S4: Process the theoretically optimal collection and transportation sequence into segments for navigation instructions, and add detailed single-point operation annotations using the three-dimensional structure voxel data of the waste to manage the smart waste bin collection and transportation operation.

2. The implementation method of the smart city IoT device data collection management platform according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Deploy IoT sensing devices inside trash cans in the smart city and connect them to an edge computing unit with at least 4 general input / output interfaces and 1 analog-to-digital conversion interface to build an IoT front-end sensing network; wherein, the IoT sensing devices include ultrasonic sensor arrays, pressure sensors, infrared thermal imaging sensors, gas sensors and temperature and humidity sensors; Step S12: Read the voltage signal of the pressure sensor in the IoT front-end sensing network in real time, and obtain the real-time total weight data of the garbage filling after analog-to-digital conversion; Step S13: Based on the real-time total weight data of the garbage, mark each time the weight changes by more than 100 grams as an incremental throwing event; Step S14: Based on the time of the incremental throwing event, use the ultrasonic sensor array in the IoT front-end sensing network to obtain the echo time from each sensor to the garbage surface to obtain the original echo duration. Step S15: Convert the original echo duration into the corresponding spatial rectangular coordinates of the sensor according to the preset sound velocity value, thereby determining the surface height coordinates inside the trash can; Step S16: Analyze the structural voxel changes of the throwing event based on the surface height coordinates and the real-time total weight of the waste, and generate three-dimensional structural voxel data of the waste.

3. The implementation method of the smart city IoT device data acquisition and management platform according to claim 2, characterized in that, Step S16 involves analyzing the structure voxel changes of the throwing event based on the surface height coordinates and the real-time total weight of the waste filling data. Based on incremental throwing events, the ratio of volume increment to weight increment for each throwing event is calculated using surface height coordinates and real-time total weight data of the waste, to assess the density of different batches of waste and obtain stratified waste density values. Acquire historical surface height time-series data; Using historical surface height time series data and the three-dimensional geometric parameters of the trash can pre-set in the edge computing unit, coordinate mapping and basic voxel mesh calibration are performed in real time, and density weights are assigned to voxels at the corresponding depths using the layered trash density values ​​to generate a density layered voxel model; The density-layered voxel model is divided into lower-layer voxel meshes, and then the lower-layer voxel meshes are simulated and corrected for gravity settlement effect, that is, the layered waste density values ​​corresponding to the lower-layer voxel meshes are compressed and compensated by 1-5%, so as to obtain the three-dimensional structure voxel data of waste.

4. The implementation method of the smart city IoT device data collection management platform according to claim 1, characterized in that, Environmental risk compensation is performed based on dynamic thermal decomposition data using ambient humidity, and the urgency of the bin's condition is calculated based on the total waste filling rate, including: Gain compensation is applied to the dynamic thermal decomposition degree data based on ambient humidity to generate compensated thermal decomposition degree data. Real-time gas concentration values ​​are obtained using gas sensors in the front-end sensing network of the Internet of Things (IoT). Acquire local meteorological data and extract the current ambient wind speed as the local ambient wind speed value; Determine whether the real-time gas concentration value is greater than 50 ppm and the local ambient wind speed value is less than 1 m / s. If yes, set the influence coefficient to 1.4; otherwise, set it to 1.0 to obtain the odor diffusion influence coefficient. The public health sensitivity of the data on the degree of thermal decomposition was processed by using the odor diffusion influence coefficient, and the urgency of the state inside the bin was calculated based on the total waste filling rate, thus obtaining the dynamic urgency index.

5. The implementation method of the smart city IoT device data collection management platform according to claim 1, characterized in that, Step S3, which involves weighted processing of the urgency of waste disposal based on the dynamic urgency index, includes: Obtain the functional area category labels of trash cans; match regulatory weight coefficients based on the functional area category labels; where functional area category labels include commercial areas, residential areas, and public places; The urgency of waste disposal is calculated by weighting the regulatory weight coefficient and the dynamic urgency index, and the priority score is normalized to generate waste disposal score data. Set the collection and transportation start threshold to 75, filter out trash cans with a collection and transportation score data greater than 75 to obtain a high-priority site list; otherwise, obtain a low-priority site list. The site with the highest waste disposal score is selected from the list of high-priority sites as the initial center of the regional cluster and the cluster generation center point. Using the cluster generation center point as the center, a radius of 1 kilometer is defined on the map. All high-priority sites and some secondary-priority sites falling within this range are aggregated to obtain the potential cluster member set. Based on the three-dimensional structure voxel data of the waste, the estimated total volume and weight of the waste to be transported are calculated to obtain the estimated load of the cluster. Match the target collection vehicle that is closest to the cluster generation center point; The estimated load of the cluster is compared with the carrying capacity of the target collection vehicle. If the load exceeds the limit, the site with the lowest urgency is removed from the potential cluster member set until the condition is met, thus generating a cluster of sites to be processed.

6. The implementation method of the smart city IoT device data acquisition and management platform according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Take the trash can with the highest priority in the coordinate data of the site to be visited as the first target point of the target collection vehicle. Then, starting from the first target point, query the real-time driving time of all remaining points in the coordinate data of the site to be visited through the preset map service interface to obtain the time taken for the candidate points. Step S32: Select the point with the shortest time from the candidate points and determine it as the next optimal work point for the vehicle; Step S33: Add the next optimal work point to the task list of the current vehicle and remove the point from the coordinate data of the site to be visited to obtain the updated work path; Step S34: Repeat steps S31 to S33 until the capacity or shift duration of the target collection and transportation vehicles is met, and use the updated operation route as a feasible operation sequence. Step S35: Adjust the priority of the feasible operation sequence by secondary weighting based on the collection and transportation score data to obtain the theoretically optimal collection and transportation sequence.

7. The implementation method of the smart city IoT device data collection management platform according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Plan the navigation path for the theoretically optimal collection sequence and process the segmented navigation instructions to obtain segmented navigation instructions; Step S42: Based on each target station in the segmented navigation instructions, fill in the structural features through intelligent analysis of the three-dimensional structure voxel data of the garbage, and generate a single-point detailed operation annotation; Step S43: Combine the segmented navigation instructions with the single-point detailed operation annotations for each station and send them to the terminal of the target collection vehicle to manage the smart garbage bin collection operation.

8. A smart city IoT device data collection management platform, characterized in that, include: The platform comprises a data layer, a service layer, and an application layer. The data layer outputs to a service layer, and the service layer outputs to an application layer. The data layer collects and analyzes smart city information. The service layer provides various services to meet the access needs of the application layer. The application layer accesses smart terminal devices through an access interface. The data layer executes the implementation method of the smart city IoT device data collection and management platform as described in claim 1. The data layer in this smart city IoT device data collection and management platform includes: The IoT data acquisition module is used to deploy IoT sensing devices inside the trash can and build an IoT front-end sensing network. It collects data on incremental throwing events in the trash can through the IoT front-end sensing network, analyzes the changes in structural voxels in the trash can caused by incremental throwing events, and generates three-dimensional structural voxel data of the trash. The dynamic risk assessment module is used to calculate the total waste filling rate based on the three-dimensional structure voxel data of the waste; to assess the dynamic thermal decomposition degree data based on the three-dimensional structure voxel data of the waste; and to calculate the urgency of the state inside the bin based on the total waste filling rate, thereby obtaining a dynamic urgency index; to perform regulatory weighted collection urgency processing based on the dynamic urgency index, and then identify the cluster of sites to be processed. The collection and transportation route planning module is used to extract the coordinate data of the sites to be visited in the cluster of sites to be processed, and to dynamically plan the theoretically optimal collection and transportation sequence for the target collection and transportation vehicles. The intelligent operation management module is used to process the theoretically optimal collection and transportation sequence into segments and navigation instructions, and to perform single-point fine operation annotations through the three-dimensional structure voxel data of waste, so as to realize the intelligent management of waste collection and transportation in smart cities.

Images