A convenient service system and method for recycling materials
By performing local data processing and cloud verification on the user end, combined with blockchain technology, the problems of data authenticity and privacy leakage in internet recycling services have been solved, achieving efficient, reliable and low-carbon management of material recycling.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI JIUAO IND CO
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-05
AI Technical Summary
Existing internet recycling services suffer from problems such as difficulty in verifying data authenticity, high risk of privacy leaks, and wasteful carbon emissions from logistics. In particular, users are prone to cheating by using fake photos when declaring materials, and passive logistics response leads to resource waste.
The system employs a private domain-aware agent to process local video stream data, generating a declaration feature hash packet. This is combined with the reverse logistics entropy model of the community aggregation scheduling node and the dual handshake verification of the cloud-based intelligent command and control platform. Edge visual preprocessing and anti-fraud detection are utilized, and data verification and path optimization are achieved through blockchain token minting technology.
It effectively blocks photo copying cheating, reduces the risk of privacy data leaving the domain, improves the full load rate and identification accuracy of material recycling, reduces logistics carbon emissions, and forms credible carbon reduction certificates.
Smart Images
Figure CN122155565A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of Internet data processing and intelligent logistics scheduling technology, and in particular to a convenient service system and method for material recycling. Background Technology
[0002] With the acceleration of urbanization, recycling has become an important part of the carbon neutrality strategy. However, existing internet-based recycling services mainly face the following technical bottlenecks: First, verifying the authenticity of the data is difficult and the cost of trust is high. Current technology mainly relies on users filling out forms or uploading photos, which are then manually verified by recycling personnel. Due to the lack of a closed-loop verification mechanism at the technical level, users can easily use fake photos or substitute inferior materials (such as iron cans being passed off as aluminum cans) to submit their applications. Furthermore, the manual verification data at the recycling end is difficult to verify against the declared data in an unalterable and traceable manner, resulting in a high risk of data falsification on the platform and making it impossible to form credible carbon reduction certificates.
[0003] Secondly, in order to identify the type of materials, existing applications usually require users to upload original images containing the background of their home environment to the cloud server. These images of non-target areas often contain private information such as faces and interior furnishings. Uploading them directly to the cloud poses a serious risk of data leakage and privacy breaches. In addition, existing recycling logistics mostly adopts an order-triggered passive response, that is, users place orders and the platform dispatches orders. This often ignores the clustering effect of materials in the community and the differences in the physical perishability of different materials (such as cardboard and perishable biomass). As a result, recycling vehicles often run empty or take detours. The carbon emissions of the logistics process itself offset the emission reduction benefits brought by recycling, and cannot solve the contradiction between fragmented demand and large-scale collection. Summary of the Invention
[0004] The purpose of this invention is to address the deficiencies in the existing technology by proposing a convenient material recycling service system and method.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: A convenient service system for recycling materials includes: The private domain awareness agent is deployed on the user-side edge device and is equipped with a lightweight visual recognition model. It is used to collect video stream data of the materials to be recycled locally, perform background removal and privacy desensitization preprocessing, and extract the multimodal physical features of the materials to be recycled to generate a declaration feature hash package containing spatiotemporal stamps. The community aggregation scheduling node communicates with multiple private domain awareness agent terminals to receive the declaration feature hash packet, construct a reverse logistics entropy model based on the cumulative volume and time decay factor of materials in the region, and generate a clearance trigger signal when the entropy value exceeds a preset threshold. The recycling operation terminal is held by recycling personnel or mounted on a transport vehicle, and is used to collect physical data of materials at the physical receiving site and generate delivery feature hash packets. The cloud-based intelligent command and control platform is communicatively connected to both the community aggregation and scheduling node and the recycling operation terminal, including: The route planning module, in response to the collection trigger signal, uses a dynamic clustering algorithm to generate recycling route instructions for the target community; A dual handshake verification module is used to receive the delivered feature hash packet and calculate its feature vector similarity with the declared feature hash packet. The token minting module generates a digital verification token with corresponding carbon reduction weight and writes it to the private domain-aware agent account if and only if the similarity of the feature vector is higher than the consistency threshold.
[0006] Furthermore, the private domain awareness agent terminal is equipped with an edge visual preprocessing module; the edge visual preprocessing module is specifically configured as follows: using a semantic segmentation algorithm to identify the region of interest (ROI) in the video stream, performing Gaussian blurring or pixel block obfuscation on the background area outside the ROI; using a multi-frame difference method to extract key frames containing different perspectives of the material from the video stream, performing 3D point cloud sparse reconstruction on the key frames to obtain the volume estimate of the material, and encapsulating the volume estimate into the declaration feature hash packet.
[0007] Furthermore, the private domain awareness agent is also configured with anti-cheating liveness detection logic, which is configured as follows: During the acquisition of video streams, the correspondence between the camera's movement trajectory and the optical flow characteristics of the image is calculated in real time. Only when an object in the image is detected to have a depth parallax change that conforms to physical laws, and the light reflection characteristics change continuously with the viewing angle, is it determined to be a valid entity acquisition and the feature extraction process is initiated. Otherwise, lock and refuse to generate the declared feature hash packet.
[0008] Furthermore, the community aggregation and scheduling node runs a material spoilage risk assessment algorithm; the reverse logistics entropy model The calculation formula is configured as follows: in, For the first The estimated volume in each declaration feature hash packet As a weight for space occupancy, This refers to the length of time the supplies have been held since they were declared. The decay attenuation coefficient is based on the type of material used in the materials. Weighted by urgency; When the calculated real-time entropy value Greater than the preset critical scheduling threshold At that time, the community aggregation scheduling node sends an immediate cleanup request to the cloud-based intelligent command and control platform through a high-priority channel.
[0009] Furthermore, the path planning module of the cloud-based intelligent command and control platform adopts a genetic evolution strategy with time window constraints. The specific steps of the genetic evolution strategy include: The community node that issues the collection trigger signal is defined as the gene locus in the gene code, and the upper limit of the load capacity of the recycling vehicle and the current remaining load capacity are defined as constraints; the objective function is set as minimizing the sum of carbon emissions from total transportation mileage and carbon emissions from vehicle start-stop times. During the iteration process, a congestion index penalty factor is introduced to mutate the path population, output a dynamic recovery cruise path, and send the path instruction to the navigation unit of the recovery operation terminal.
[0010] Furthermore, the dual handshake verification module employs consistency comparison logic based on orthogonal feature projection, and the comparison logic specifically includes: The texture histogram features extracted from the declaration feature hash package are compared with the texture histogram features in the delivery feature hash package for the first comparison. A second comparison is performed between the color distribution features extracted from the declaration feature hash packet and the color distribution features extracted from the delivery feature hash packet; Ignoring differences in geometric features, the cosine similarity between the two feature vectors is calculated. If the cosine similarity value is greater than the preset consistency threshold, the physically delivered materials and the digitally declared materials are determined to be from the same batch.
[0011] Furthermore, the system also includes a dynamic credit anchoring and grading module, which is configured as follows: Maintain a dynamic confidence score for each private domain awareness agent; For agents with confidence scores higher than the first-level threshold, reduce the sampling frequency of the dual handshake verification module and trigger token minting by comparing only volume data. For agents with confidence scores below the second-level threshold, the full feature comparison mode is forcibly enabled, and the recycling client is required to upload a signature for manual review. Only after both algorithm verification and manual review are passed will the lock on the generation of digital verification tokens be released.
[0012] Furthermore, the token minting module is built based on consortium blockchain smart contract technology, and the specific configuration of the token minting module is as follows: A digital verification token containing carbon emission reduction weight values and a traceability chain for the entire life cycle of materials is generated. The traceability chain includes: the original feature fingerprint at the time of declaration, the vehicle trajectory hash at the time of disposal, and the processing hash at the final entry into the recycling plant. The traceability chain is packaged into a block and broadcast to a distributed ledger jointly maintained by the cloud and community nodes.
[0013] Furthermore, the system also includes a federated learning model calibration module; the module is configured such that when the dual handshake verification module determines that the declaration and delivery are inconsistent, the cloud will retrieve the delivery image collected by the job terminal as a calibration sample. The calibration samples are used to train and generate model parameter update packages in the cloud, and the model parameter update packages are sent to the corresponding private domain awareness agent. The private domain awareness agent uses the model parameter update package to incrementally update the local lightweight visual recognition model.
[0014] The present invention also provides a convenient service method for recycling materials, based on the above-described system implementation.
[0015] A convenient method for recycling materials includes the following steps: S1: The private domain awareness agent wakes up the edge computing unit, collects video stream data of the materials to be recycled, performs background removal and privacy desensitization preprocessing, and extracts multimodal physical features to generate a declaration feature hash package; S2: The community aggregation scheduling node aggregates the declaration feature hash packets in the area in real time, calculates the reverse logistics entropy value based on the accumulated volume of materials and the time decay factor, and generates a clearance trigger signal when the entropy value exceeds the preset threshold. S3: The cloud-based intelligent command and control platform responds to the collection trigger signal, uses a dynamic clustering algorithm to generate recycling path instructions and sends them to the recycling operation terminal; S4: The recycling operation terminal collects physical data of materials at the physical receiving site and generates a delivery feature hash packet; S5: The cloud-based intelligent command and control platform receives the delivery feature hash packet, calculates its feature vector similarity with the declaration feature hash packet, and if and only if the similarity is higher than the consistency threshold, it forges a digital verification token through a smart contract and writes it into the private domain awareness agent account.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: By using a dual handshake verification of the declared feature hash and the delivered feature hash, and by comparing the orthogonal projection of texture and color distribution, cheating behaviors such as photo copying and passing off inferior goods as superior ones are effectively blocked. In addition, by combining blockchain token minting technology, the physical world material recycling behavior is transformed into an immutable digital verification token, which solves the technical problem of falsifying recycling data. By utilizing the edge vision preprocessing module, ROI extraction and background Gaussian blurring are completed locally on the user end. Only the desensitized feature hash package is uploaded instead of the original image, which not only completely blocks the risk of family privacy data leaving the domain, but also significantly reduces the network transmission bandwidth usage. By introducing a reverse logistics entropy model, the volume of materials is correlated with the decay time factor (…). By combining these technologies, the system can dynamically sense the urgency of community supplies. With a genetic evolution strategy constrained by time windows, it can shift from passively accepting orders to actively eliminating entropy, maximizing the full load rate of each clearing operation and significantly reducing logistics carbon emissions. In addition, through the federated learning calibration module, the edge model is incrementally updated using adversarial examples of verification failures, enabling the system to adapt to constantly changing cheating methods and the form of supplies, while maintaining recognition accuracy. Attached Figure Description
[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.
[0018] Figure 1 This is a schematic diagram of the overall architecture and module connection of a convenient material recycling service system provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the calculation logic of the reverse logistics entropy model running by the community aggregation scheduling node in this embodiment of the invention. Figure 3 This is a flowchart illustrating the logic of the dual handshake verification module performing feature consistency comparison and anomaly interception in this embodiment of the invention. Figure 4 This is a full-process interactive sequence diagram of a convenient material recycling service method provided by an embodiment of the present invention. Detailed Implementation
[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0020] Example 1 like Figure 1 As shown in the figure, a convenient service system for recycling materials in this embodiment includes a private domain perception agent terminal, a community aggregation and scheduling node, a recycling operation terminal, and a cloud-based intelligent command and control platform.
[0021] The private domain awareness proxy is deployed on a user-side edge device, such as a smartphone, a smart home gateway, or an embedded IoT terminal. The private domain awareness proxy integrates an edge visual preprocessing module, an anti-fraud detection unit, and a feature hash generator. The edge visual preprocessing module is configured to run a lightweight convolutional neural network model, such as MobileNet or ShuffleNet, to perform semantic segmentation and region of interest extraction on the acquired video stream data. The feature hash generator is configured to extract multimodal physical features within the region of interest and encapsulate them into an encrypted declaration feature hash packet.
[0022] The community aggregation and scheduling node is deployed on a community edge server or regional data center and communicates with multiple private domain awareness agents within its jurisdiction. The community aggregation and scheduling node internally runs a reverse logistics entropy model, used to calculate the cumulative state entropy of materials within the region in real time.
[0023] The recycling operation terminal is held by recycling personnel or mounted on a logistics transport vehicle, and is equipped with a handheld or vehicle-mounted scanning terminal for collecting physical data of materials at the physical receiving site and generating delivery feature hash packets.
[0024] The cloud-based intelligent command and control platform establishes encrypted communication connections with both the community aggregation and scheduling node and the recycling operation terminal. The platform includes a path planning module, a two-way handshake verification module, a federated learning calibration module, and a token minting module. The path planning module is configured to run a dynamic clustering algorithm and a genetic evolution strategy. The two-way handshake verification module is configured to calculate the feature vector similarity between the declaration feature hash packet and the delivery feature hash packet. The token minting module is built based on consortium blockchain smart contract technology to generate digital verification tokens containing carbon emission reduction weights and traceability chains, and writes them into the blockchain distributed ledger.
[0025] Example 2 like Figure 2 As shown, based on Embodiment 1, the specific process of the system performing reverse logistics entropy calculation and dynamic scheduling is as follows: The community aggregation scheduling node receives the declaration feature hash packets from each private domain awareness agent and parses them to obtain the estimated volume of each declaration unit. Including the declaration timestamp. The system calculates the duration of material retention based on the current system time. .
[0026] The reverse logistics entropy model matches the spoilage decay coefficient for each type of material based on a pre-set material database. For low-corrosion-risk materials, such as metals or dry plastics, set The value is set to low; for materials with high risk of spoilage, such as damp cardboard or biomass, a value is set. It is a high value.
[0027] The reverse logistics entropy model calculates the total regional entropy value based on the following formula. : in, For the first The volume of each application unit This is the space occupancy weighting coefficient. For the duration of stay, This represents the time urgency weighting coefficient. The entropy value is calculated in real-time by the community aggregation scheduling node. ,when Exceeding the preset critical scheduling threshold At that time, a collection trigger signal is generated and sent to the cloud-based intelligent command and control platform via a high-priority channel. The path planning module of the cloud-based intelligent command and control platform responds to the collection trigger signal by executing a genetic evolution strategy with time window constraints. This strategy defines the community node that issues the collection signal as a locus in the gene encoding, defines the upper limit of the recycling vehicle's load capacity as a constraint, and sets the objective function as minimizing the sum of carbon emissions from total transportation mileage and carbon emissions from vehicle start-stop cycles. Through iterative calculation, a dynamic recycling cruise path command is output and sent to the navigation unit at the recycling operation end.
[0028] Example 3 like Figure 3 and Figure 4 As shown, the specific data closed-loop process for the system to perform anti-cheating detection and dual handshake verification is as follows: During the application phase, the anti-cheating detection unit of the private domain perception agent terminal activates the liveness detection logic. During the video stream acquisition process, it calculates the correspondence between the camera movement trajectory and the optical flow characteristics of the image in real time. Only when the object in the image is detected to have a depth parallax change that conforms to physical laws and the light reflection characteristics change continuously with the viewing angle, it is determined to be a valid entity acquisition, and the feature extraction process is allowed to start. Otherwise, lock and refuse to generate the declaration feature hash packet.
[0029] After effective acquisition, the edge vision preprocessing module identifies the region of interest (ROI) in the video stream, performs Gaussian blur processing on the background pixels outside the ROI, and the feature hash generator extracts the texture histogram features and color distribution features of the ROI to generate the application feature hash package. And upload it.
[0030] During the delivery phase, the recycling terminal scans the physical items and generates a delivery feature hash packet. And upload it.
[0031] The dual handshake verification module of the cloud-based intelligent command and control platform receives and Perform consistency comparison based on orthogonal feature projection. The specific steps include: Step 1, Extract Texture histogram features and The texture histogram features in the data are used for the first comparison; The second step is extraction. Color distribution characteristics and A second comparison is performed on the color distribution characteristics in the data; The third step is to ignore the differences in geometric shape features and calculate the weighted cosine similarity of the two feature vectors mentioned above.
[0032] The token minting module triggers a smart contract to generate a corresponding digital verification token and write it into the private domain awareness agent account, completing the data loop, only when the calculated cosine similarity value is greater than a preset consistency threshold.
[0033] If the cosine similarity value is lower than the consistency threshold, the system performs a blocking operation, refuses to generate a digital verification token, and... and The data pairs are labeled as negative samples. The federated learning calibration module uses these negative samples to generate model parameter update packages and sends them to the private domain-aware agent to incrementally update the local lightweight visual recognition model.
[0034] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A convenient service system for recycling materials, characterized in that, include: The private domain awareness agent is deployed on the user-side edge device and is equipped with a lightweight visual recognition model. It is used to collect video stream data of the materials to be recycled locally, perform background removal and privacy desensitization preprocessing, and extract the multimodal physical features of the materials to be recycled to generate a declaration feature hash package containing spatiotemporal stamps. The community aggregation scheduling node communicates with multiple private domain awareness agent terminals to receive the declaration feature hash packet, construct a reverse logistics entropy model based on the cumulative volume and time decay factor of materials in the region, and generate a clearance trigger signal when the entropy value exceeds a preset threshold. The recycling operation terminal is held by recycling personnel or mounted on a transport vehicle, and is used to collect physical data of materials at the physical receiving site and generate delivery feature hash packets. The cloud-based intelligent command and control platform is communicatively connected to both the community aggregation and scheduling node and the recycling operation terminal, including: The route planning module, in response to the collection trigger signal, uses a dynamic clustering algorithm to generate recycling route instructions for the target community; A dual handshake verification module is used to receive the delivered feature hash packet and calculate its feature vector similarity with the declared feature hash packet. The token minting module generates a digital verification token with corresponding carbon reduction weight and writes it to the private domain-aware agent account if and only if the similarity of the feature vector is higher than the consistency threshold.
2. The convenient material recycling service system according to claim 1, characterized in that, The private domain awareness agent is equipped with an edge visual preprocessing module; the edge visual preprocessing module is specifically configured to: use a semantic segmentation algorithm to identify the region of interest in the video stream, and perform Gaussian blur or pixel block obfuscation on the background area outside the region of interest. The multi-frame difference method is used to extract key frames containing different perspectives of the material from the video stream. The key frames are then subjected to 3D point cloud sparse reconstruction to obtain the volume estimate of the material, and the volume estimate is encapsulated into the declaration feature hash packet.
3. The convenient material recycling service system according to claim 1, characterized in that, The private domain awareness agent is also configured with anti-cheating liveness detection logic, which is configured as follows: During the acquisition of video streams, the correspondence between the camera's movement trajectory and the optical flow characteristics of the image is calculated in real time. Only when an object in the image is detected to have a depth parallax change that conforms to physical laws, and the light reflection characteristics change continuously with the viewing angle, is it determined to be a valid entity acquisition and the feature extraction process is initiated. Otherwise, lock and refuse to generate the declared feature hash packet.
4. The convenient material recycling service system according to claim 1, characterized in that, The community aggregation and scheduling node runs a material corruption risk assessment algorithm; the reverse logistics entropy model The calculation formula is configured as follows: in, For the first The estimated volume in each declaration feature hash packet As a weight for space occupancy, This refers to the length of time the supplies have been held since they were declared. The decay attenuation coefficient is based on the type of material used in the materials. Weighted by urgency; When the calculated real-time entropy value Greater than the preset critical scheduling threshold At that time, the community aggregation scheduling node sends an immediate cleanup request to the cloud-based intelligent command and control platform through a high-priority channel.
5. The convenient material recycling service system according to claim 1, characterized in that, The path planning module of the cloud-based intelligent command and control platform adopts a genetic evolution strategy with time window constraints. The specific steps of the genetic evolution strategy include: The community node that issues the collection trigger signal is defined as the gene locus in the gene code, and the upper limit of the load capacity of the recycling vehicle and the current remaining load capacity are defined as constraints; the objective function is set as minimizing the sum of carbon emissions from total transportation mileage and carbon emissions from vehicle start-stop times. During the iteration process, a congestion index penalty factor is introduced to mutate the path population, output a dynamic recovery cruise path, and send the path instruction to the navigation unit of the recovery operation terminal.
6. The convenient material recycling service system according to claim 1, characterized in that, The dual handshake verification module employs consistency comparison logic based on orthogonal feature projection, which specifically includes: The texture histogram features extracted from the declaration feature hash package are compared with the texture histogram features in the delivery feature hash package for the first comparison. A second comparison is performed between the color distribution features extracted from the declaration feature hash packet and the color distribution features extracted from the delivery feature hash packet; Ignoring differences in geometric features, the cosine similarity between the two feature vectors is calculated. If the cosine similarity value is greater than the preset consistency threshold, the physically delivered materials and the digitally declared materials are determined to be from the same batch.
7. The convenient material recycling service system according to claim 1, characterized in that, The system also includes a dynamic credit anchoring and grading module, which is configured as follows: Maintain a dynamic confidence score for each private domain awareness agent; For agents with confidence scores higher than the first-level threshold, reduce the sampling frequency of the dual handshake verification module and trigger token minting by comparing only volume data. For agents with confidence scores below the second-level threshold, the full feature comparison mode is forcibly enabled, and the recycling client is required to upload a signature for manual review. Only after both algorithm verification and manual review are passed will the lock on the generation of digital verification tokens be released.
8. The convenient material recycling service system according to claim 1, characterized in that, The token minting module is built based on consortium blockchain smart contract technology, and the specific configuration of the token minting module is as follows: A digital verification token containing carbon emission reduction weight values and a traceability chain for the entire life cycle of materials is generated. The traceability chain includes: the original feature fingerprint at the time of declaration, the vehicle trajectory hash at the time of disposal, and the processing hash at the final entry into the recycling plant. The traceability chain is packaged into a block and broadcast to a distributed ledger jointly maintained by the cloud and community nodes.
9. A convenient material recycling service system according to claim 1, characterized in that, The system also includes a federated learning model calibration module; the module is configured such that when the dual handshake verification module determines that the declaration and delivery are inconsistent, the cloud will collect the delivery image collected by the work terminal as a calibration sample. The calibration samples are used to train and generate model parameter update packages in the cloud, and the model parameter update packages are sent to the corresponding private domain awareness agent. The private domain awareness agent uses the model parameter update package to incrementally update the local lightweight visual recognition model.
10. A method for providing convenient recycling services, implemented based on the system described in any one of claims 1-9, characterized in that, Includes the following steps: S1: The private domain awareness agent wakes up the edge computing unit, collects video stream data of the materials to be recycled, performs background removal and privacy desensitization preprocessing, and extracts multimodal physical features to generate a declaration feature hash package; S2: The community aggregation scheduling node aggregates the declaration feature hash packets in the area in real time, calculates the reverse logistics entropy value based on the accumulated volume of materials and the time decay factor, and generates a clearance trigger signal when the entropy value exceeds the preset threshold. S3: The cloud-based intelligent command and control platform responds to the collection trigger signal, uses a dynamic clustering algorithm to generate recycling path instructions and sends them to the recycling operation terminal; S4: The recycling operation terminal collects physical data of materials at the physical receiving site and generates a delivery feature hash packet; S5: The cloud-based intelligent command and control platform receives the delivery feature hash packet, calculates its feature vector similarity with the declaration feature hash packet, and if and only if the similarity is higher than the consistency threshold, it forges a digital verification token through a smart contract and writes it into the private domain awareness agent account.