Power bank advertisement coverage verification system based on end-cloud cooperation

Abstract

The present application relates to the field of calculation estimation, specifically to a power bank advertisement coverage verification system based on end-cloud cooperation. A cloud server stores advertisement metadata and expected advertisement feature codes; a cabinet end reads the current advertisement identifier of a plugged-in power bank and uploads it; the cloud end matches the expected advertisement based on the position and issues the expected advertisement feature code; the cabinet end compares the current advertisement feature with the expected advertisement feature code, and if they are inconsistent, obtains target advertisement data and transmits it to the power bank via a universal serial bus contact; the power bank calculates a target feature value and reports it to the cloud server for secondary verification; if the verification fails, a retransmission mechanism or an alarm instruction is triggered, and the cloud server records coverage logs. The present application avoids redundant writing, reduces data transmission volume and time consumption, and excludes coverage failure caused by transmission interruption through secondary verification, thus ensuring that the actual stored advertisement is consistent with the expected advertisement.

Description

Technical Field

[0001] This invention relates to the field of computational estimation, specifically to a power bank advertising coverage verification system based on edge-cloud collaboration. Background Technology

[0002] Currently, the conventional technical solution for placing advertisements in shared power bank cabinets typically employs a local full-coverage approach. The cabinet locally stores the advertisement video files to be updated. When a power bank is inserted into a cabinet slot, the cabinet does not distinguish whether the currently stored advertisement content in the power bank matches the advertisement to be displayed; it directly rewrites the complete advertisement video file into the power bank's storage chip via a universal serial bus contact. This solution relies solely on the write operation on one side of the cabinet. After the write is completed, the cabinet does not perform data readback or cloud verification; it only determines whether the write process is finished.

[0003] The core problem with the above-mentioned conventional technical solutions is that the cabinet does not perform feature comparison on the existing advertising content in the power bank before covering. The full write operation leads to frequent redundant overwriting of power banks that have correctly stored the same advertisement, resulting in wasted communication bandwidth and writing time of the universal serial bus contacts. After the overlay is completed, there is no cloud-based secondary verification mechanism. When the universal serial bus contacts are not in good contact or the transmission is interrupted, resulting in incomplete written data, the cabinet cannot recognize the overlay failure, resulting in the actual advertisement stored in the power bank being inconsistent with the expected advertisement. Summary of the Invention

[0004] The purpose of this invention is to provide a power bank advertising coverage verification system based on end-to-end cloud collaboration, which can effectively solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] The power bank advertising coverage verification system based on end-to-end cloud collaboration includes a cloud server and a cabinet terminal. The cloud server stores the metadata, version information and expected advertising feature codes of the advertising videos.

[0007] The cabinet has a built-in communication module and includes a compartment for inserting a power bank.

[0008] When the power bank is inserted into the compartment, the cabinet reads the current advertising identifier stored in the power bank and uploads the current advertising identifier and the compartment number to the cloud server through the communication module.

[0009] The cloud server matches the expected advertisement based on the warehouse number and sends the expected advertisement feature code of the expected advertisement to the rack terminal;

[0010] The server rack compares the current advertising feature corresponding to the current advertising identifier with the expected advertising feature code;

[0011] If they are inconsistent, the rack terminal downloads the target advertising data from the cloud server or reads the target advertising data cached locally, and transmits the target advertising data to the power bank through the USB contact of the rack.

[0012] After the transmission is completed, the power bank calculates the target feature value of the target advertisement data and reports it to the cloud server;

[0013] The cloud server performs a second verification between the target feature value and the expected advertising feature code. If the verification fails, a retransmission mechanism or alarm command is triggered. If the verification is successful, the overlay is confirmed to be successful.

[0014] The cloud server records an overlay log, which includes the position number, the expected advertising feature code, and the result of the secondary verification.

[0015] Preferably, the metadata of the advertising video stored on the cloud server includes a content hash value and a multimodal content feature code;

[0016] When the cloud server receives a newly uploaded advertising video source file, it extracts the keyframe image sequence and audio segment sequence from the advertising video source file.

[0017] Based on a deep feature extraction network, feature mapping is performed on the keyframe image sequence and the audio segment sequence to generate image feature vectors and audio feature vectors, respectively.

[0018] The image feature vector and the audio feature vector are fused and dimensionality reduced to generate the multimodal content feature code, and the multimodal content feature code is bound to the content hash value and stored as the expected advertisement feature code;

[0019] When the cloud server sends out the expected advertising feature code, it simultaneously sends the content hash value to the server rack.

[0020] Preferably, the step of reading the current advertising identifier stored in the power bank at the cabinet end includes: the power bank has a built-in secure storage chip, and the secure storage chip stores the current advertising identifier and historical coverage record index;

[0021] When the power bank is inserted into the slot, the cabinet sends an authentication request to the secure storage chip via the USB contact.

[0022] After authentication is successful, the rack terminal reads the current advertising identifier and the historical coverage record index;

[0023] The rack terminal combines the historical coverage record index with the current advertisement identifier to generate current advertisement upload data, and sends the current advertisement upload data and the warehouse number to the cloud server through the communication module;

[0024] The cloud server verifies the legitimacy of the power bank based on the warehouse number and the historical coverage record index.

[0025] Preferably, the step of comparing the current advertising feature corresponding to the current advertising identifier with the expected advertising feature code at the rack end includes: the rack end parses the current advertising identifier and extracts the current hash value and the current multimodal feature;

[0026] The expected advertising feature code issued by the cloud server is parsed to extract the expected hash value and expected multimodal features;

[0027] Calculate the first similarity between the current hash value and the expected hash value, and calculate the second similarity between the current multimodal feature and the expected multimodal feature;

[0028] The first similarity and the second similarity are weighted and fused to obtain a comprehensive similarity.

[0029] If the overall similarity is lower than a preset coverage threshold, it is determined that the current advertising feature is inconsistent with the expected advertising feature code, triggering the cabinet terminal to download the target advertising data from the cloud server.

[0030] Preferably, the step of transmitting the target advertising data to the power bank via the USB contact of the compartment at the cabinet end includes: the cabinet end splitting the target advertising data into multiple data segments and adding a serial number and a check digit to each data segment;

[0031] The rack end transmits the data fragments to the power bank sequentially according to the serial number via the USB contact;

[0032] During transmission, the power bank monitors the voltage level of the USB contacts in real time. If an abnormal voltage level interruption is detected, the serial number of the interruption is recorded.

[0033] When the voltage level signal is restored, the power bank sends a resume transmission request carrying the serial number to the cabinet.

[0034] The rack end continues transmission from the data fragment corresponding to the sequence number based on the resume request until all data fragments have been transmitted.

[0035] Preferably, the step of the cloud server performing a secondary verification between the target feature value and the expected advertising feature code includes: the cloud server receiving the target feature value and the device identification code of the power bank;

[0036] Extract the expected advertising feature code corresponding to the device identification code as the benchmark feature value;

[0037] Calculate the Hamming distance between the target feature value and the reference feature value;

[0038] If the Hamming distance is greater than the preset fault tolerance distance, the secondary verification is determined to have failed, and the cloud server generates a verification failure instruction and sends the verification failure instruction to the rack terminal.

[0039] In response to the verification failure command, the rack terminal counts the number of consecutive verification failures. If the number of consecutive verification failures reaches a preset retransmission threshold, the alarm command is triggered, and the rack number of the rack is marked as faulty.

[0040] Preferably, the step of performing feature mapping on the keyframe image sequence and the audio segment sequence based on the deep feature extraction network includes: performing spatial Gaussian filtering noise reduction on the keyframe image sequence, and inputting the noise-reduced keyframe image sequence into a three-dimensional convolutional network to extract spatiotemporal feature maps.

[0041] The audio segment sequence is subjected to time-frequency transformation to obtain Mel spectrogram, and the Mel spectrogram is input into a residual network to extract high-dimensional audio features;

[0042] The spatiotemporal feature map and the high-dimensional audio feature are cross-attention fused, the spatial attention weight of the spatiotemporal feature map relative to the high-dimensional audio feature is calculated, the spatiotemporal feature map is weighted pooled based on the spatial attention weight, and the dimensionality-reduced multimodal content feature code is output.

[0043] Preferably, the step of the rack terminal sending an authentication request to the secure storage chip via the USB contact includes: the rack terminal generating an authentication challenge message carrying a timestamp and a random number, and sending the authentication challenge message to the secure storage chip via the USB contact;

[0044] The secure storage chip uses its built-in private key to digitally sign the authentication challenge message, generates a signed response message, and returns it to the rack.

[0045] The rack terminal uses the public key corresponding to the built-in private key to verify the signature response message.

[0046] If the signature verification is successful, the rack terminal is allowed to read the current advertising identifier and the historical coverage record index;

[0047] The cabinet also uploads the signature response message to the cloud server, and the cloud server verifies the physical uniqueness of the power bank and the tamper-proof nature of the historical overwrite record index based on the signature response message.

[0048] Preferably, the step of the cabinet end transmitting the data fragments to the power bank sequentially according to the serial number via the USB contact includes: the cabinet end detecting the current supply current value of the current compartment and the transmission load rate of the USB contact in real time;

[0049] When the power supply current value is greater than a preset current threshold and the transmission load rate is greater than a preset load threshold, the rack end dynamically reduces the transmission frequency of the data fragment and increases the transmission interval between adjacent data fragments.

[0050] When the power supply current value is less than the preset current threshold and the transmission load rate is less than the preset load threshold, the rack end dynamically increases the transmission frequency and transmits multiple data fragments in batches based on the sliding window mechanism.

[0051] The power bank performs cyclic redundancy checks on the received data segments in real time and accumulates the check results into the coverage progress table.

[0052] Preferably, the step of the cloud server recording the coverage log includes: after confirming successful coverage, the cloud server extracts the device identifier, the advertisement number corresponding to the expected advertisement feature code, the coverage completion timestamp, and the warehouse number;

[0053] The device identifier, the advertisement number, the coverage completion timestamp, and the warehouse number are bound and written into the blockchain distributed ledger;

[0054] Based on smart contracts, records in the blockchain distributed ledger are aggregated to generate an immutable billing bill.

[0055] When an external audit request is received, the cloud server retrieves the target block data from the blockchain distributed ledger, extracts the Merkle tree path from the target block data for hash verification, outputs the integrity verification result covering the log, and synchronously feeds back the integrity verification result with the billing bill.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] 1. This system uses a cloud server to match the expected advertisement based on the warehouse number and issues the expected advertisement feature code. The cabinet compares the current advertisement feature of the power bank with the expected advertisement feature code. Only when there is a discrepancy will the advertisement data be downloaded and transmitted. This avoids repeatedly writing data to power banks that have already stored the correct advertisement, reducing the amount of data transmission and writing time of the universal serial bus contacts. After the transmission is completed, the power bank calculates the target feature value and reports it to the cloud server for secondary verification. If the verification fails, a retransmission mechanism or alarm command is triggered, eliminating the risk of data incompleteness caused by poor contact or transmission interruption, and ensuring the consistency between the actual advertisement stored in the power bank and the expected advertisement.

[0058] 2. Combining multimodal feature extraction and Hamming distance verification mechanisms, the cloud server uses a deep feature extraction network to extract multimodal content feature codes from keyframe image sequences and audio segment sequences of advertising videos. The server rack calculates the weighted comprehensive similarity between the current multimodal features and the expected multimodal features. The increase in feature dimensions avoids misjudgments caused by single hash collisions. The cloud server calculates the Hamming distance based on the target feature value and the benchmark feature value, and compares it with the preset fault tolerance distance to determine whether the verification has failed. This distance measurement method allows for minor transmission disturbances, avoids meaningless retransmissions triggered by slight bit flips, and improves the verification fault tolerance rate. Attached Figure Description

[0059] Figure 1 This is the main flowchart of the power bank advertising coverage verification based on end-to-cloud collaboration of the present invention;

[0060] Figure 2 This is a flowchart of the cloud-based multimodal advertising feature extraction process of the present invention;

[0061] Figure 3 This is a flowchart illustrating the secure identity authentication process for the server rack and power bank according to the present invention.

[0062] Figure 4 This is a flowchart of the advertising feature weighted comparison and decision-making process of the present invention;

[0063] Figure 5 This is a flowchart illustrating the advertising data transmission process that supports resuming interrupted downloads according to the present invention.

[0064] Figure 6 This is a flowchart of the cloud-based secondary verification and blockchain log recording process of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please refer to Figure 1 This embodiment provides a power bank advertising coverage verification system based on edge-cloud collaboration, including a cloud server and a cabinet terminal. The cloud server is deployed on a distributed cloud platform and establishes communication connections with multiple cabinet terminals via a wide area network. The storage module of the cloud server pre-stores the metadata, version information, and expected advertising feature codes of the advertising videos. Each advertising video corresponds to a unique advertising number, and the metadata and version information are bound to this advertising number for storage. The version information includes the advertising release time, validity period, placement area, and placement cabinet group identifier. The cabinet terminal has a built-in communication module that uses cellular network communication and supports 4G or 5G network standards. The cabinet terminal includes multiple slots for power banks to be inserted, and each slot has an independent USB contact group, which includes power contacts and data contacts. Each slot corresponds to a unique slot number, which is combined with the device identifier code of the cabinet terminal to generate a globally unique slot identifier.

[0067] When a power bank is inserted into a charging slot, a mechanical detection switch inside the slot is triggered, sending an insertion detection signal to the main control module at the rack level. Upon receiving the insertion detection signal, the main control module at the rack level sends a handshake request message to the power bank via the USB data contact. The power bank's microcontroller, upon receiving the handshake request message, returns a handshake response message, establishing a communication connection between the rack level and the power bank. After the communication connection is established, the rack level reads the current advertising identifier stored in the power bank. The current advertising identifier is a fixed-length string containing advertising number and version number information. The rack level combines the read current advertising identifier with the slot number to generate an upload data packet, which is then sent to the cloud server via the communication module.

[0068] After receiving the uploaded data packet, the cloud server's communication interface module parses the packet and extracts the rack number and the current advertisement identifier. The cloud server's advertisement matching module queries the pre-stored rack-advertisement mapping table based on the rack number to obtain the expected advertisement number corresponding to that rack. The rack-advertisement mapping table stores the correspondence between each rack number and the expected advertisement number, and this mapping table is updated by the cloud server's management module according to the advertisement delivery strategy. The advertisement matching module retrieves the corresponding expected advertisement feature code from the storage module based on the expected advertisement number. The cloud server's communication interface module then sends the expected advertisement feature code to the rack that sent the uploaded data packet.

[0069] After receiving the expected ad feature code from the cloud server, the server rack compares the current ad features corresponding to the current ad identifier with the expected ad feature code. The feature comparison module on the server rack first parses the current ad identifier and extracts the current ad's feature information. The feature comparison module then compares the current ad's feature information with the expected ad feature code bit by bit, calculating the matching degree. If the matching degree reaches a preset matching threshold, the current ad features are determined to match the expected ad feature code, and the server rack sends an overwrite completion notification to the cloud server, without needing to perform ad data transmission. If the matching degree is lower than the preset matching threshold, the current ad features are determined to be inconsistent with the expected ad feature code, triggering the ad data transmission process.

[0070] When it is determined that advertising data transmission is required, the server rack first checks whether the target advertising data is stored in its local cache module. The target advertising data is the complete advertising video file corresponding to the expected advertising feature code. If the target advertising data is stored in the local cache module, the server rack directly reads the target advertising data from the local cache module. If the target advertising data is not stored in the local cache module, the server rack sends an advertising download request to the cloud server through the communication module, carrying the expected advertising number in the request. After receiving the advertising download request, the cloud server retrieves the corresponding advertising video file from the storage module, divides it into multiple data blocks, and sends them to the server rack. The server rack receives all data blocks, performs concatenation verification, and stores the data in the local cache module after confirming its integrity.

[0071] The rack-side transmits target advertising data to the power bank via USB contacts in the storage compartment. The data transmission module on the rack-side encapsulates the target advertising data according to a preset transmission protocol, generating a transmission data packet. The transmission data packet includes a header, a data body, and a checksum. The header contains the data packet length, sequence number, and data type information. The rack-side sends the transmission data packets to the power bank sequentially according to their sequence numbers via the USB contacts. Upon receiving the transmission data packet, the power bank's microcontroller verifies the checksum. If the verification passes, the data body is written to the power bank's storage chip, and an acknowledgment message is sent to the rack-side. If the verification fails, a retransmission request is sent to the rack-side, requesting the retransmission of the data packet with the corresponding sequence number.

[0072] After transmission is complete, the power bank's feature calculation module calculates the target feature value of the target advertising data. The feature calculation module performs a hash operation on the complete target advertising data to generate a fixed-length hash value as the target feature value. The power bank reports the target feature value to the server rack via USB contacts. The server rack combines the target feature value with the power bank's device identification code to generate a verification request data packet, which is then sent to the cloud server via the communication module.

[0073] After receiving the verification request data packet, the secondary verification module of the cloud server extracts the target feature value and device identification code. The secondary verification module then queries the expected advertising feature code corresponding to the power bank based on the device identification code, using it as the baseline feature value. The secondary verification module compares the target feature value with the baseline feature value to perform a secondary verification. If the secondary verification fails, the cloud server triggers a retransmission mechanism or an alarm command. The retransmission mechanism involves the cloud server sending a retransmission command to the server rack, requesting the rack to retransmit the target advertising data to the power bank. The alarm command involves the cloud server sending an alarm message to the management platform, which includes the server rack device identification code, rack number, and the number of verification failures. If the secondary verification succeeds, the cloud server confirms successful overwriting.

[0074] The cloud server's logging module records overlay logs. These logs include the rack unit number, expected ad signature, secondary verification result, overlay start timestamp, and overlay completion timestamp. The logging module stores these overlay logs in the cloud server's log database and creates an index, supporting queries and retrieval by rack-side device identifier, rack unit number, ad number, or time range.

[0075] In this embodiment, communication between the server rack and the power bank uses the USB Mass Storage Protocol (USB-MTP) for data transmission. USB-MTP supports batch transmission, which improves data transmission efficiency. When transmitting target advertising data, the server rack uses batch transmission, transmitting data blocks of up to 512 bytes each time. When receiving data blocks, the power bank employs a dual-buffer mechanism: one buffer for receiving data and the other for writing to the storage chip, enabling parallel processing of receiving and writing.

[0076] In this embodiment, the communication between the cloud server and the server rack uses the Transmission Control Protocol (TCP) for data transmission. TCP provides reliable connection-oriented communication services, ensuring the integrity and order of data transmission. A persistent connection is established between the cloud server and the server rack to reduce the overhead of connection establishment and termination. When the persistent connection is interrupted, the server rack automatically attempts to reconnect until the connection is restored.

[0077] In this embodiment, the preset matching threshold is set to 0.95. When the matching degree between the current ad feature and the expected ad feature code is greater than or equal to 0.95, it is determined to be consistent; when the matching degree is less than 0.95, it is determined to be inconsistent. The preset matching threshold can be adjusted according to the actual application scenario, as shown in Table 1.

[0078] Table 1. Advertising Coverage Status Transition Table

[0079] Idle state Power bank inserted detected Handshake status Send handshake request message Handshake status Received handshake response message Read the identifier status Read the current ad identifier Read the identifier status Successfully read the current ad identifier Upload Identifier Status Upload the current ad identifier and position number. Upload Identifier Status Received expected ad signature Feature comparison status Compare current ad features with expected ad signatures Feature comparison status Feature matching consistent Idle state Send coverage completion notification Feature comparison status Inconsistent feature comparison Data preparation status Query local cache or download target ad data Data preparation status Target advertising data preparation complete Data transmission status Transmit target advertising data to power bank Data transmission status Data transmission completed Feature reporting status Receive target feature values ​​reported by the power bank Feature reporting status Received target feature value Secondary verification status Upload the target feature values ​​to the cloud server for secondary verification. Secondary verification status Secondary verification successful Idle state Record the overwrite log to confirm successful overwrite. Secondary verification status Secondary verification failed and the number of retransmissions did not reach the threshold. Data transmission status Trigger the retransmission mechanism to retransmit the target ad data. Secondary verification status Secondary verification failed and the number of retransmissions reached the threshold. Fault status Trigger an alarm command and mark the position as faulty.

[0080] Table 1 illustrates the state transition relationships during the ad coverage process. Each state corresponds to a specific execution operation; when the triggering condition is met, the system transitions from the current state to the target state. Managing the ad coverage process using a state machine ensures the standardization and traceability of the process.

[0081] In this embodiment, the cloud server matches the expected advertisement based on the storage location number and issues the expected advertisement feature code. The cabinet compares the current advertisement feature of the power bank with the expected advertisement feature code. Only when there is a discrepancy will the advertisement data download and transmission operation be performed. This avoids repeatedly writing data to power banks that have already stored the correct advertisement, reducing the data transmission volume and writing time of the USB contacts. After the transmission is completed, the power bank calculates the target feature value and reports it to the cloud server for secondary verification. If the verification fails, a retransmission mechanism or alarm command is triggered, eliminating the risk of data incompleteness caused by poor contact or transmission interruption, and ensuring the consistency between the actual advertisement stored in the power bank and the expected advertisement.

[0082] In a preferred embodiment, reference Figure 2The metadata of the advertising videos stored on the cloud server includes content hash values ​​and multimodal content feature codes. When the cloud server receives a newly uploaded advertising video source file, it first performs format and integrity checks on the file. Format checks confirm that the file's format meets preset requirements; supported formats include MP4, AVI, and MOV. Integrity checks confirm that the file was not corrupted during upload. After successful checks, the cloud server's video processing module extracts keyframe image sequences and audio segment sequences from the advertising video source file.

[0083] The video processing module extracts keyframes from the advertising video source file at preset time intervals. The preset time interval is set to 1 second, meaning one keyframe is extracted every second. For an advertising video with a duration of T seconds, the extracted keyframe image sequence contains T frames. The video processing module performs size normalization on the extracted keyframe images, adjusting the size of all keyframe images to 224×224 pixels. Size normalization ensures consistent input sizes for subsequent feature extraction.

[0084] The audio processing module extracts the audio track from the advertising video source file and divides it into multiple audio segments according to a preset duration. The preset duration is set to 1 second, meaning each audio segment is 1 second long. The audio processing module performs a sampling rate conversion on each audio segment, uniformly converting the sampling rate to 16kHz. This sampling rate conversion ensures a consistent input sampling rate for subsequent audio feature extraction.

[0085] The feature extraction module of the cloud server performs feature mapping on keyframe image sequences and audio segment sequences based on a deep feature extraction network, generating image feature vectors and audio feature vectors respectively. The deep feature extraction network includes an image feature extraction sub-network and an audio feature extraction sub-network. The image feature extraction sub-network adopts a three-dimensional convolutional network structure, capable of simultaneously extracting spatial and temporal features of the image. The audio feature extraction sub-network adopts a residual network structure, capable of extracting high-dimensional features of the audio.

[0086] The feature extraction module fuses and reduces the dimensionality of image and audio feature vectors to generate multimodal content feature codes. The fusion and dimensionality reduction process combines concatenation and principal component analysis. First, the image and audio feature vectors are concatenated to obtain a concatenated feature vector. Then, principal component analysis is performed on the concatenated feature vector to reduce its dimensionality to a preset level, generating the multimodal content feature code. The preset dimensionality is set to 256.

[0087] The cloud server's storage module binds the multimodal content signature code and the content hash value to store the expected ad signature code. The content hash value is generated by performing a SHA-256 hash operation on the complete ad video source file. The SHA-256 hash operation can generate a 256-bit hash value, which has high collision resistance. When the cloud server distributes the expected ad signature code, it simultaneously distributes the content hash value to the server rack.

[0088] refer to Figure 4 The steps for the server rack to compare the current advertisement features corresponding to the current advertisement identifier with the expected advertisement feature code include: the feature comparison module on the server rack parses the current advertisement identifier and extracts the current hash value and the current multimodal feature. The current hash value is the SHA-256 hash value of the current advertisement data stored in the power bank. The current multimodal feature is the 256-dimensional multimodal content feature code of the current advertisement data stored in the power bank. The feature comparison module parses the expected advertisement feature code issued by the cloud server and extracts the expected hash value and the expected multimodal feature.

[0089] The feature comparison module calculates the first similarity between the current hash value and the expected hash value. The first similarity is calculated using the Hamming distance metric. Hamming distance represents the number of different characters at corresponding positions in two strings of equal length. The formula for calculating the first similarity is:

[0090]

[0091] in, Indicates the first similarity. Indicates the current hash value Compared with the expected hash value Hamming distance between them This indicates the length of the hash value. For SHA-256 hash values, The value is 256.

[0092] The feature comparison module calculates the second similarity between the current multimodal features and the expected multimodal features. The second similarity is calculated using cosine similarity. Cosine similarity measures the angle between two vectors; the smaller the angle, the higher the cosine similarity, indicating greater similarity between the two vectors. The formula for calculating the second similarity is:

[0093]

[0094] in, Indicates the second similarity. This represents the current multimodal feature vector. This represents the expected multimodal feature vector. Represents the dot product of two vectors. and These represent the magnitudes of the two vectors, respectively.

[0095] The feature comparison module performs a weighted fusion of the first and second similarity scores to obtain a comprehensive similarity score. The formula for calculating the weighted fusion is as follows:

[0096]

[0097] in, Indicates the overall similarity. The weight coefficient representing the first similarity. This represents the weighting coefficient for the second similarity score. (Weighting coefficient) The value range is between 0 and 1. In this embodiment, The value is set to 0.6.

[0098] The feature comparison module compares the overall similarity with a preset coverage threshold, which is set to 0.9. If the overall similarity is lower than the preset coverage threshold, the current ad feature is determined to be inconsistent with the expected ad feature code, triggering the server to download the target ad data from the cloud server. If the overall similarity is greater than or equal to the preset coverage threshold, the current ad feature is determined to be consistent with the expected ad feature code, and no ad data transmission operation is required.

[0099] In a preferred embodiment, the step of feature mapping the keyframe image sequence and audio segment sequence based on a deep feature extraction network includes: performing spatial domain Gaussian filtering noise reduction on the keyframe image sequence. Spatial domain Gaussian filtering noise reduction uses a two-dimensional Gaussian kernel function to perform a convolution operation on the image, which can effectively remove Gaussian noise in the image. The expression of the two-dimensional Gaussian kernel function is:

[0100]

[0101] in, This indicates that the Gaussian kernel function is in coordinates The value at that location, This represents the standard deviation of the Gaussian kernel. In this embodiment, the size of the Gaussian kernel is set to 5×5, and the standard deviation... The value is set to 1.0.

[0102] The denoised keyframe image sequence is input into a 3D convolutional network to extract spatiotemporal feature maps. The 3D convolutional network consists of multiple 3D convolutional layers, pooling layers, and fully connected layers. The convolutional kernels of the 3D convolutional layers perform convolution operations simultaneously in both spatial and temporal dimensions, enabling the simultaneous extraction of spatial and temporal features from the image. The output of the 3D convolutional layers is downsampled by pooling layers to reduce the dimensionality of the feature maps. Finally, the fully connected layers convert the feature maps into one-dimensional image feature vectors. The dimension of the image feature vectors is set to 512.

[0103] A time-frequency transform (TFD) is performed on an audio segment sequence to obtain the Mel spectrogram. The TFD uses a short-time Fourier transform (SFT). The SFT divides the audio signal into multiple short time windows, and a Fourier transform is performed on the signal within each time window to obtain the time-frequency representation of the signal. The Mel spectrogram is obtained by filtering the spectrogram obtained from the SFT through a Mel filter bank. The Mel filter bank simulates the auditory characteristics of the human ear and can better extract the perceptual features of the audio.

[0104] The Mel spectrogram is input into a residual network to extract high-dimensional audio features. The residual network contains multiple residual blocks, each consisting of two convolutional layers and a skip connection. Skip connections address the vanishing gradient problem during deep network training, improving both training depth and performance. The output of the residual network is then subjected to global average pooling for dimensionality reduction, resulting in a one-dimensional audio feature vector. The audio feature vector is set to 512 dimensions.

[0105] This involves cross-attention fusion of spatiotemporal feature maps and high-dimensional audio features. The cross-attention fusion mechanism calculates the importance weight of each spatial location in the spatiotemporal feature map relative to the high-dimensional audio features, thereby highlighting image regions relevant to the audio content. The formula for cross-attention fusion is as follows:

[0106]

[0107] in, Represents the spatial attention weight matrix. Represents a spatiotemporal feature map. Representing high-dimensional features of audio, and This represents the learnable weight matrix. Representing feature dimension, This represents the softmax activation function.

[0108] Weighted pooling of the spatiotemporal feature map is performed based on spatial attention weights to output a dimensionality-reduced multimodal content feature code. The formula for weighted pooling is as follows:

[0109]

[0110] in, Represents multimodal content feature codes, and These represent the height and width of the spatiotemporal feature map, respectively. The spatial attention weight matrix represents the position. The value at that location, Indicates the spatiotemporal feature map at location The feature vector at the location. The multimodal feature extraction parameters are shown in Table 2.

[0111] Table 2 Multimodal Feature Extraction Parameter Configuration Table

[0112] Keyframe extraction Extraction interval 1 second One keyframe is extracted per second. Keyframe preprocessing Normalized size 224×224 pixels Adjust all keyframes to a uniform size Audio segmentation Fragment length 1 second The audio track was divided into 1-second segments. Audio preprocessing Sampling rate 16kHz Convert the audio sampling rate to 16kHz. Image feature extraction Network Structure 3D-ResNet18 An 18-layer three-dimensional residual network is used. Image feature extraction Output Dimension 512-dimensional Dimension of image feature vectors Audio feature extraction Network Structure ResNet34 A 34-layer two-dimensional residual network is used. Audio feature extraction Output Dimension 512-dimensional Dimension of audio feature vectors Feature fusion Fusion method Cross-attention fusion Multimodal feature fusion using cross-attention mechanism Feature fusion Output Dimension 256 dimensions Dimensions of multimodal content feature codes

[0113] Table 2 shows the parameter configurations during the multimodal feature extraction process. By configuring the parameters appropriately, it is possible to control computational complexity and storage overhead while ensuring the accuracy of feature extraction.

[0114] In this embodiment, the cloud server utilizes a deep feature extraction network to extract multimodal content feature codes from the keyframe image sequence and audio segment sequence of the advertising video. The server rack calculates the weighted comprehensive similarity between the current multimodal features and the expected multimodal features. The increased feature dimensions avoid misjudgments caused by single hash collisions. Multimodal features simultaneously encompass both visual and auditory information of the advertising video, enabling a more comprehensive description of the advertising content and improving the accuracy of feature comparison.

[0115] In a preferred embodiment, reference Figure 3 The power bank has a built-in secure storage chip. This chip uses a dedicated chip with hardware encryption capabilities, supporting both symmetric and asymmetric encryption algorithms. The secure storage chip stores the current ad identifier and a historical coverage record index. The historical coverage record index includes the timestamps, ad numbers, and coverage results of the N most recent ad coverages. The value of N is set to 10. The secure storage chip also contains a unique device private key and a corresponding public key certificate. The public key certificate is issued by a certificate authority on the cloud server and is used to verify the identity of the secure storage chip.

[0116] When a power bank is inserted into the storage compartment, the rack sends an authentication request to the secure storage chip via a USB contact. The authentication request process includes: the rack's security authentication module generating an authentication challenge message carrying a timestamp and a random number. The timestamp is the rack's current system time, used to prevent replay attacks. The random number is a 128-bit pseudo-random number used to increase the randomness of the authentication challenge. The security authentication module then sends the authentication challenge message to the secure storage chip via the USB contact.

[0117] Upon receiving the authentication challenge message, the secure storage chip digitally signs the message using its built-in private key. The digital signature employs the elliptic curve digital signature algorithm, which offers advantages such as short key length, fast computation speed, and high security. The secure storage chip then generates a signed response message, which includes the digital signature result and the chip's public key certificate. Finally, the secure storage chip returns the signed response message to the server rack.

[0118] The security authentication module at the server rack uses the public key corresponding to the built-in private key to verify the signed response message. The module first verifies the validity of the public key certificate, confirming that it was issued by the certificate authority on the cloud server and has not expired. After successful verification, the module uses the public key from the certificate to verify the digital signature result. If the verification passes, the server rack allows reading the current advertising identifier and historical overlay record index. If the verification fails, the server rack refuses to communicate with the power bank and sends an unauthorized device alarm to the cloud server.

[0119] The server rack combines the historical coverage record index with the current advertisement identifier to generate the current advertisement upload data, and sends the current advertisement upload data and rack number to the cloud server via the communication module. The cloud server verifies the legitimacy of the power bank based on the rack number and the historical coverage record index. The cloud server's legitimacy verification module first queries the historical coverage record corresponding to the rack number, and then compares the historical coverage record index reported by the power bank with the historical coverage record stored in the cloud. If they match, the power bank is deemed legitimate. If they do not match, the power bank may have been tampered with, and the cloud server sends an alarm command to the server rack.

[0120] The server rack also uploads the signed response message to the cloud server. The cloud server verifies the physical uniqueness of the power bank and the tamper-proof nature of the historical overlay record index based on the signed response message. The cloud server's security verification module verifies the digital signature in the signed response message, confirming that the signature was generated by a legitimate secure storage chip. Because each secure storage chip has a unique private key, the physical uniqueness of the power bank is guaranteed. Simultaneously, because the historical overlay record index is digitally signed by the secure storage chip before uploading, it is guaranteed that the historical overlay record index has not been tampered with during transmission.

[0121] In a preferred embodiment, reference Figure 5 The steps for transmitting target advertising data from the rack-side to the power bank via the USB contacts in the storage compartment include: The data transmission module on the rack-side splits the target advertising data into multiple data fragments. Each data fragment is set to 4KB in size. The data transmission module adds a sequence number and a checksum to each data fragment. The sequence number is a 32-bit unsigned integer used to identify the order of the data fragments. The checksum uses a cyclic redundancy check (CRC) code to detect whether errors have occurred during data fragment transmission. The generator polynomial for the CRC code is CRC-32.

[0122] The rack-side transmits data fragments sequentially to the power bank via USB contacts according to their serial numbers. The data transmission module maintains a sending window with a size of 8. The data transmission module continuously sends the data fragments within the sending window to the power bank without waiting for an acknowledgment message for each data fragment. Upon receiving a data fragment, the power bank verifies the checksum. If the verification passes, the power bank writes the data fragment to the storage chip and sends an acknowledgment message to the rack-side, containing the serial number of the correctly received data fragment. If the verification fails, a negative acknowledgment message is sent to the rack-side, containing the serial number of the erroneous data fragment.

[0123] During transmission, the power bank monitors the voltage levels of the USB contacts in real time. These voltage levels include both power and data levels. The power bank's voltage level monitoring module samples the USB contacts at a sampling frequency of 100Hz. If an abnormal voltage level interruption is detected, the sequence number of the interruption is recorded. Abnormal voltage level interruptions include a power level below a preset threshold or a data level remaining unchanged for more than a preset time. The preset power level threshold is set to 4.5V. The preset data level no-change time threshold is set to 1 second.

[0124] Once the signal level is restored, the power bank sends a resume request carrying a serial number to the server rack. The resume request message contains the serial number of the last correctly received data fragment. Based on the resume request, the server rack continues transmission from the next data fragment corresponding to that serial number until all data fragments have been transmitted. This resume mechanism avoids the need to retransmit all data due to temporary contact issues, improving data transmission efficiency.

[0125] In a preferred embodiment, the step of transmitting data segments sequentially to the power bank via USB contacts according to their serial numbers at the rack end includes: the rack-end transmission control module continuously monitors the current supply current value of the current compartment and the transmission load rate of the USB contacts. The current supply value is detected by a sampling resistor connected in series in the power circuit. The transmission load rate is the ratio of the current actual transmission rate to the maximum transmission rate of the USB interface. The USB interface adopts the USB 2.0 standard with a maximum transmission rate of 480Mbps.

[0126] When the supply current exceeds a preset current threshold and the transmission load rate exceeds a preset load threshold, the rack dynamically reduces the transmission frequency of data segments and increases the transmission interval between adjacent data segments. The preset current threshold is set to 1A. The preset load threshold is set to 80%. The transmission frequency adjustment range is 10Hz to 100Hz. The transmission interval adjustment range is 10ms to 100ms. By reducing the transmission frequency and increasing the transmission interval, the power consumption of the USB interface can be reduced, avoiding transmission interruptions due to insufficient power supply.

[0127] When the supply current is less than a preset current threshold and the transmission load rate is less than a preset load threshold, the rack dynamically increases the transmission frequency and continuously transmits multiple data fragments in batches based on a sliding window mechanism. The size of the sliding window is dynamically adjusted according to the current transmission load rate. When the transmission load rate is below 50%, the sliding window size is set to 16. When the transmission load rate is between 50% and 80%, the sliding window size is set to 8. By increasing the transmission frequency and the sliding window size, the transmission bandwidth of the USB interface can be fully utilized, thereby improving the data transmission speed.

[0128] The power bank performs cyclic redundancy check (CRC) on the received data fragments in real time and accumulates the check results into the coverage progress table. The coverage progress table contains the sequence number, reception status, and check result for each data fragment. The coverage progress table is stored in the power bank's random access memory (RAM), and its contents are not lost during transmission interruptions. Once all data fragments have been transmitted and checked successfully, the power bank writes the coverage progress table to the secure storage chip and sends a transmission completion message to the server rack. The dynamic adjustment parameters for data transmission are shown in Table 3.

[0129] Table 3 Data Transmission Dynamic Adjustment Parameters

[0130] Supply current > 1A and transmission load rate > 80% 10Hz 4 100ms 0.5A < supply current ≤ 1A and 50% < transmission load rate ≤ 80% 50Hz 8 50ms The power supply current is ≤0.5A and the transmission load rate is ≤50%. 100Hz 16 10ms

[0131] Table 3 shows the dynamically adjusted parameters during data transmission. By configuring different transmission parameters according to different power supply current and transmission load conditions, data transmission efficiency can be maximized while ensuring transmission stability.

[0132] In a preferred embodiment, reference Figure 6 The cloud server performs a secondary verification between the target feature value and the expected advertising feature code, including the following steps: The secondary verification module of the cloud server receives the target feature value and the power bank's device identification code. The device identification code is a unique hardware identifier built into the power bank, written into a secure storage chip during the manufacturing process and cannot be tampered with. The secondary verification module extracts the expected advertising feature code corresponding to the device identification code as the base feature value. The cloud server's storage module maintains a mapping table between the device identification code and the expected advertising feature code, which is updated after each successful advertising coverage.

[0133] The secondary verification module calculates the Hamming distance between the target feature value and the baseline feature value. The formula for calculating the Hamming distance is:

[0134]

[0135] in, Indicates Hamming distance, Indicates the length of the eigenvalues. The first eigenvalue represents the target eigenvalue. Bit, The first eigenvalue representing the baseline eigenvalue Bit.

[0136] The secondary verification module compares the Hamming distance with a preset fault tolerance distance, which is set to 3. If the Hamming distance is greater than the preset fault tolerance distance, the secondary verification is considered a failure. If the Hamming distance is less than or equal to the preset fault tolerance distance, the secondary verification is considered a success. This distance metric allows for minor transmission disturbances, avoids meaningless retransmissions triggered by slight bit flips, and improves the verification fault tolerance rate.

[0137] If the secondary verification fails, the cloud server generates a verification failure command and sends it to the server rack. The server rack responds to the verification failure command by counting the number of consecutive verification failures. The consecutive verification failure count is the number of times the same power bank continuously covers the same advertisement in the same rack and fails the secondary verification. The server rack compares the consecutive verification failure count with a preset retransmission threshold. The preset retransmission threshold is set to 3. If the consecutive verification failure count reaches the preset retransmission threshold, an alarm command is triggered, and the rack number is marked as faulty. If the consecutive verification failure count does not reach the preset retransmission threshold, a retransmission mechanism is triggered, and the target advertisement data is retransmitted to the power bank.

[0138] In a preferred embodiment, the step of the cloud server recording the coverage log includes: after confirming successful coverage, the cloud server's log recording module extracts the device identifier, the ad number corresponding to the expected ad feature code, the coverage completion timestamp, and the position number. The log recording module binds the device identifier, ad number, coverage completion timestamp, and position number and writes them into the blockchain distributed ledger. The blockchain distributed ledger adopts a consortium blockchain architecture and is jointly maintained by multiple ad placement parties and operators. Each node stores a complete copy of the ledger, ensuring the immutability and traceability of the ledger data.

[0139] The logging module generates new blocks, each containing multiple overwritten log records. The block generation interval is set to 10 minutes. Each block contains a block header and a block body. The block header includes the hash value of the previous block, the hash value of the current block, a timestamp, and a Merkle root. The block body contains multiple overwritten log records. The Merkle root is generated by hashing all overwritten log records within the block body and is used to verify the integrity of the data in the block body.

[0140] Smart contracts aggregate records from the blockchain's distributed ledger to generate immutable billing invoices. Deployed on the blockchain, these smart contracts automatically execute pre-defined logic. They aggregate coverage log records by ad ID and time range, counting the number of successful coverages for each ad within the specified time frame. Based on the number of successful coverages and a pre-defined billing rate, the smart contract calculates the advertising cost and generates a billing invoice. This billing invoice is then written to the blockchain's distributed ledger and is therefore immutable.

[0141] Upon receiving an external audit request, the audit module on the cloud server retrieves the target block data from the blockchain's distributed ledger. The audit request includes the advertisement ID and time range to be audited. The audit module extracts the Merkle tree path from the target block data for hash verification. The Merkle tree path consists of all hash values ​​along the path from the overlay log record to the Merkle root. By recalculating the Merkle root and comparing it with the Merkle root in the block header, the integrity of the overlay log record can be verified. The audit module outputs the integrity verification result of the overlay log and synchronously feeds back the integrity verification result along with the billing invoice to the audit requester.

[0142] In this embodiment, the cloud server writes the overlay logs into the blockchain distributed ledger, leveraging the blockchain's immutability and traceability to ensure the authenticity and integrity of the overlay logs. Billing invoices are automatically generated based on smart contracts, avoiding errors and fraudulent activities during manual billing. When auditing is required, Merkle tree path verification can quickly and accurately verify the integrity of the overlay logs, improving auditing efficiency.

Claims

1. A power bank advertisement coverage verification system based on end-cloud cooperation, comprising a cloud server and a cabinet end, characterized in that, The cloud server stores the metadata, version information, and expected advertising feature codes of the advertising videos. The cabinet has a built-in communication module and includes a compartment for inserting a power bank. When the power bank is inserted into the compartment, the cabinet reads the current advertising identifier stored in the power bank and uploads the current advertising identifier and the compartment number to the cloud server through the communication module. The cloud server matches the expected advertisement based on the warehouse number and sends the expected advertisement feature code of the expected advertisement to the rack terminal; The server rack compares the current advertising feature corresponding to the current advertising identifier with the expected advertising feature code; If they are inconsistent, the rack terminal downloads the target advertising data from the cloud server or reads the target advertising data cached locally, and transmits the target advertising data to the power bank through the USB contact of the rack. After the transmission is completed, the power bank calculates the target feature value of the target advertisement data and reports it to the cloud server; The cloud server performs a second verification between the target feature value and the expected advertising feature code. If the verification fails, a retransmission mechanism or alarm command is triggered. If the verification is successful, the overlay is confirmed to be successful. The cloud server records an overlay log, which includes the position number, the expected advertising feature code, and the result of the secondary verification. 2.The power bank advertisement coverage verification system based on end-cloud cooperation according to claim 1, wherein, The metadata of the advertising videos stored on the cloud server includes content hash values ​​and multimodal content feature codes; When the cloud server receives a newly uploaded advertising video source file, it extracts the keyframe image sequence and audio segment sequence from the advertising video source file. Based on a deep feature extraction network, feature mapping is performed on the keyframe image sequence and the audio segment sequence to generate image feature vectors and audio feature vectors, respectively. The image feature vector and the audio feature vector are fused and dimensionality reduced to generate the multimodal content feature code, and the multimodal content feature code is bound to the content hash value and stored as the expected advertisement feature code; When the cloud server sends out the expected advertising feature code, it simultaneously sends the content hash value to the server rack. 3.The power bank advertisement coverage verification system based on end-cloud cooperation according to claim 1, wherein, The step of reading the current advertising identifier stored in the power bank at the cabinet end includes: the power bank has a built-in security storage chip, which stores the current advertising identifier and historical coverage record index; When the power bank is inserted into the slot, the cabinet sends an authentication request to the secure storage chip via the USB contact. After authentication is successful, the rack terminal reads the current advertising identifier and the historical coverage record index; The rack terminal combines the historical coverage record index with the current advertisement identifier to generate current advertisement upload data, and sends the current advertisement upload data and the warehouse number to the cloud server through the communication module; The cloud server verifies the legitimacy of the power bank based on the warehouse number and the historical coverage record index.

4. The power bank advertisement covering verification system based on end-cloud cooperation according to claim 1, characterized in that, The step of comparing the current advertising feature corresponding to the current advertising identifier with the expected advertising feature code at the rack end includes: the rack end parses the current advertising identifier and extracts the current hash value and the current multimodal feature; The expected advertising feature code issued by the cloud server is parsed to extract the expected hash value and expected multimodal features; Calculate the first similarity between the current hash value and the expected hash value, and calculate the second similarity between the current multimodal feature and the expected multimodal feature; The first similarity and the second similarity are weighted and fused to obtain a comprehensive similarity. If the overall similarity is lower than a preset coverage threshold, it is determined that the current advertising feature is inconsistent with the expected advertising feature code, triggering the cabinet terminal to download the target advertising data from the cloud server.

5. The power bank advertisement covering verification system based on end-cloud cooperation according to claim 1, characterized in that, The step of transmitting the target advertising data to the power bank via the USB contact of the compartment at the cabinet end includes: the cabinet end splitting the target advertising data into multiple data segments and adding a serial number and a check digit to each data segment; The rack end transmits the data fragments to the power bank sequentially according to the serial number via the USB contact; During transmission, the power bank monitors the voltage level of the USB contacts in real time. If an abnormal voltage level interruption is detected, the serial number of the interruption is recorded. When the voltage level signal is restored, the power bank sends a resume transmission request carrying the serial number to the cabinet. The rack end continues transmission from the data fragment corresponding to the sequence number based on the resume request until all data fragments have been transmitted.

6. The power bank advertising coverage verification system based on end-to-cloud collaboration according to claim 1, characterized in that, The step of the cloud server performing secondary verification between the target feature value and the expected advertising feature code includes: the cloud server receiving the target feature value and the device identification code of the power bank; Extract the expected advertising feature code corresponding to the device identification code as the benchmark feature value; Calculate the Hamming distance between the target feature value and the reference feature value; If the Hamming distance is greater than the preset fault tolerance distance, the secondary verification is determined to have failed, and the cloud server generates a verification failure instruction and sends the verification failure instruction to the rack terminal. In response to the verification failure command, the rack terminal counts the number of consecutive verification failures. If the number of consecutive verification failures reaches a preset retransmission threshold, the alarm command is triggered, and the rack number of the rack is marked as faulty.

7. The power bank advertising coverage verification system based on end-to-cloud collaboration according to claim 2, characterized in that, The step of performing feature mapping on the keyframe image sequence and the audio segment sequence based on the deep feature extraction network includes: performing spatial Gaussian filtering noise reduction on the keyframe image sequence, inputting the denoised keyframe image sequence into a three-dimensional convolutional network, and extracting spatiotemporal feature maps. The audio segment sequence is subjected to time-frequency transformation to obtain Mel spectrogram, and the Mel spectrogram is input into a residual network to extract high-dimensional audio features; The spatiotemporal feature map and the high-dimensional audio feature are cross-attention fused, the spatial attention weight of the spatiotemporal feature map relative to the high-dimensional audio feature is calculated, the spatiotemporal feature map is weighted pooled based on the spatial attention weight, and the dimensionality-reduced multimodal content feature code is output.

8. The power bank advertising coverage verification system based on end-to-cloud collaboration according to claim 3, characterized in that, The step of the rack terminal sending an authentication request to the secure storage chip through the USB contact includes: the rack terminal generating an authentication challenge message carrying a timestamp and a random number, and sending the authentication challenge message to the secure storage chip through the USB contact; The secure storage chip uses its built-in private key to digitally sign the authentication challenge message, generates a signed response message, and returns it to the rack. The rack terminal uses the public key corresponding to the built-in private key to verify the signature response message. If the signature verification is successful, the rack terminal is allowed to read the current advertising identifier and the historical coverage record index; The cabinet also uploads the signature response message to the cloud server, and the cloud server verifies the physical uniqueness of the power bank and the tamper-proof nature of the historical overwrite record index based on the signature response message.

9. The power bank advertising coverage verification system based on end-to-cloud collaboration according to claim 5, characterized in that, The step of transmitting the data segments sequentially to the power bank via the USB contact according to the serial number at the rack end includes: the rack end detecting the current supply current value of the current compartment and the transmission load rate of the USB contact in real time; When the power supply current value is greater than a preset current threshold and the transmission load rate is greater than a preset load threshold, the rack end dynamically reduces the transmission frequency of the data fragment and increases the transmission interval between adjacent data fragments. When the power supply current value is less than the preset current threshold and the transmission load rate is less than the preset load threshold, the rack end dynamically increases the transmission frequency and transmits multiple data fragments in batches based on the sliding window mechanism. The power bank performs cyclic redundancy checks on the received data segments in real time and accumulates the check results into the coverage progress table.

10. The power bank advertising coverage verification system based on end-to-cloud collaboration according to claim 6, characterized in that, The steps of the cloud server recording the coverage log include: after confirming successful coverage, the cloud server extracts the device identification code, the advertisement number corresponding to the expected advertisement feature code, the coverage completion timestamp, and the warehouse number. The device identifier, the advertisement number, the coverage completion timestamp, and the warehouse number are bound and written into the blockchain distributed ledger; Based on smart contracts, records in the blockchain distributed ledger are aggregated to generate an immutable billing bill. When an external audit request is received, the cloud server retrieves the target block data from the blockchain distributed ledger, extracts the Merkle tree path from the target block data for hash verification, outputs the integrity verification result covering the log, and synchronously feeds back the integrity verification result with the billing bill.

Images