Cloud monitoring system based on artificial intelligence analysis

Abstract

The invention discloses a cloud monitoring system based on artificial intelligence analysis, particularly relates to the field of cloud monitoring, is used for solving the problem of data integrity damage caused by cross-channel serial number conflicts in a heterogeneous network switching scene, and comprises the following steps of: embedding real-time state parameters of a transmission path into a fragment packaging process; a multi-dimensional verification system covering time sequence jitter and spatial motion characteristics is constructed, so that detection and repair of conflict fragments can accurately accord with network fluctuation characteristics and physical scene change rules; through dual optimization of transmission window cardinal number reverse compensation and priority self-feedback suppression, while multi-path serial number dislocation is eliminated, a conflict regeneration path of a high-jitter channel is actively blocked, and a forward cycle of transmission stability enhancement and data integrity maintenance is formed. The space-time alignment precision of the video stream and the sensor data in a complex network environment is improved, a high-reliability real-time data basis is provided for an intelligent security scene, and a monitoring blind area caused by a mechanical verification rule in a traditional scheme is effectively avoided.

Description

Technical Field

[0001] The present invention relates to the field of cloud monitoring, and more specifically, to a cloud monitoring system based on artificial intelligence analysis. Background Art

[0002] Modern cloud monitoring systems generally adopt multi-path parallel transmission and adaptive compression technologies to cope with complex network environments. The video stream is segmented into shards with independent serial numbers through the Z-standard algorithm, and the transmission throughput is improved by using a dynamic window adjustment mechanism. This system can effectively reduce the stuttering rate in cross-regional deployment scenarios and rely on continuous serial number verification to ensure the frame integrity of the cloud unpacking engine, thus supporting the accurate analysis of security events by the artificial intelligence module.

[0003] However, in the mixed working conditions of multi-path link fluctuations (such as satellite and 4G handover) and ACK loss, the shards generate cross-channel serial number conflicts due to retransmission and window expansion, triggering the strong continuity verification rule of the unpacking engine, resulting in the misabandonment of key frames. This problem directly destroys the spatio-temporal alignment of the video stream and sensor data, rendering the anomaly recognition and compliance forensics functions of the cloud monitoring system ineffective and forming a monitoring blind spot.

[0004] To solve the above problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a cloud monitoring system based on artificial intelligence analysis. Through the dynamic perception and closed-loop collaboration mechanism of the multi-path transmission state, the problem of data integrity damage caused by cross-channel serial number conflicts in heterogeneous network handover scenarios is fundamentally solved. Embedding the real-time state parameters of the transmission path into the shard encapsulation process, a multi-dimensional verification system covering timing jitter and spatial motion characteristics is constructed, enabling the detection and repair of conflicting shards to accurately fit the characteristics of network fluctuations and the laws of physical scene changes; through the dual optimization of reverse compensation of the transmission window base number and self-feedback suppression of priority, while eliminating multi-path serial number misalignment, the conflict regeneration path of the high-jitter channel is actively blocked, forming a positive cycle of enhanced transmission stability and data integrity maintenance, improving the spatio-temporal alignment accuracy of video streams and sensor data in complex network environments, providing a highly reliable real-time data basis for intelligent security scenarios, and effectively avoiding the monitoring blind spots caused by mechanical verification rules in traditional solutions to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A cloud monitoring system based on artificial intelligence analysis, comprising: a path marking unit, a conflict perception unit, a spatio-temporal screening unit, a cross-path correction unit, and a window adjustment unit;

[0008] Path marking unit: When performing fragmentation encapsulation, append a channel identifier and the current transmission window base number to each fragment, and generate a fragment transmission priority label based on the instantaneous delay jitter at the moment of path switching;

[0009] Collision awareness unit: When receiving a fragment, extract the historical sequence number offset of the current channel according to the channel identifier, dynamically adjust the collision determination threshold in combination with the delay jitter, and if the difference between the fragment sequence number and the offset exceeds the threshold, mark it as a collision fragment and temporarily store it;

[0010] Space-time screening unit: For collision fragments, by calculating the deviation degree of its actual transmission delay from the path switching jitter, and analyzing the spatial motion vector continuity of adjacent fragment sensor data, screen out the repairable fragments that meet both delay tolerance and spatial consistency;

[0011] Cross-path correction unit: Based on the difference in the transmission window base numbers of the repairable fragments, inversely deduce the sequence number misalignment caused by multi-path switching, dynamically correct the fragment sequence number and verify its continuity with the current unpacking window, and inject it into the unpacking queue after passing the verification;

[0012] Window adjustment unit: According to the correction result of the fragment, freeze the window base number expansion for the channels with continuous successful corrections, and roll back the window base number of the channels with correction failures to the state before the collision and reduce the transmission priority of subsequent path switching.

[0013] In a preferred embodiment, the path marking unit includes the following:

[0014] In the fragmentation encapsulation stage, append a channel identifier, a transmission window base number, and a priority label to each video fragment, where the channel identifier is defined as the unique identifier of the transmission channel, the transmission window base number is defined as the starting sequence number of the current transmission window, and the priority label is determined based on the comparison result of the instantaneous delay jitter and a preset threshold, and the instantaneous delay jitter is defined as the absolute value of the difference in the transmission delays of adjacent video fragments.

[0015] In a preferred embodiment, the collision awareness unit includes the following:

[0016] Query the historical sequence number offset according to the channel identifier, dynamically adjust the collision determination threshold according to the priority label, calculate the sequence number offset between the fragment sequence number and the expected sequence number, and the collision determination difference between the sequence number offset and the average value of the historical offsets in the sequence number offset history. When the collision determination difference is greater than the collision determination threshold, mark the video fragment as a collision fragment and temporarily store it. When the collision determination difference is not greater than the collision determination threshold, put the video fragment into the unpacking queue.

[0017] In a preferred embodiment, the space-time screening unit includes the following:

[0018] By calculating the deviation degree between the actual transmission delay and the path switching jitter, and analyzing the spatial motion vector continuity of adjacent fragmented sensor data, the repairable fragments that meet both the delay tolerance and spatial consistency are screened out. When the deviation degree is less than the deviation degree threshold and the spatial motion vector continuity is less than the continuity metric threshold, the conflicting fragments are marked as repairable fragments.

[0019] In a preferred embodiment, the spatio-temporal screening unit further includes the following:

[0020] Wherein the actual transmission delay is defined as the difference between the received timestamp and the sent timestamp, the path switching jitter is determined based on the priority label, the deviation degree is calculated as the absolute value of the difference between the actual transmission delay and the path switching jitter divided by the path switching jitter, and the spatial motion vector continuity is obtained by calculating the minimum value of the relative change rate of the spatial motion vectors of the conflicting fragments and the adjacent fragments.

[0021] In a preferred embodiment, the path correction unit includes the following:

[0022] For the repairable fragment packets, by calculating the difference between the transmission window base number and the current unpacking window base number, and combining the non-linear adjustment of the channel historical misalignment amount and the priority coefficient, the sequence number misalignment amount is deduced, and the fragment sequence number is dynamically corrected according to the sequence number misalignment amount, and the continuity between the corrected fragment sequence number and the current unpacking window is verified, and the fragment packets that pass the verification are injected into the unpacking queue.

[0023] In a preferred embodiment, the path correction unit further includes the following:

[0024] The processing process of the corrected fragment sequence number is as follows: extract the original fragment sequence number from the header of the repairable fragment packet; then, subtract the sequence number misalignment amount from the original fragment sequence number to obtain the corrected fragment sequence number.

[0025] In a preferred embodiment, the path correction unit further includes the following:

[0026] The judgment process of the continuity verification is as follows: check whether the corrected fragment sequence number meets any of the following conditions: one is that it falls within the unpacking window range; the other is that it is equal to the current unpacking window base number minus one, that is, continuous with the window lower boundary; the third is that it is equal to the current unpacking window base number plus the window size, that is, continuous with the window upper boundary; if any of the conditions is met, it is considered to pass the verification; if none of the conditions is met, it is marked as unavailable and the corresponding fragment packet is discarded.

[0027] In a preferred embodiment, the path correction unit further includes the following:

[0028] The priority coefficient is obtained by mapping the fragment transmission priority label generated by the path marking unit.

[0029] In a preferred embodiment, the window adjustment unit includes the following:

[0030] According to the shard correction result, judge the stability of the transmission channel by counting the continuous successful correction count. Freeze the transmission window base expansion for the transmission channels whose continuous successful correction count reaches the stability threshold. Roll back the transmission window base of the shard correction failure channels to the window base before the conflict, and adjust the transmission priority through non-linear calculation.

[0031] Technical effects and advantages of the cloud monitoring system based on artificial intelligence analysis of the present invention:

[0032] Through the dynamic perception and closed-loop coordination mechanism of the multi-path transmission state, the present invention fundamentally solves the problem of data integrity damage caused by cross-channel sequence number conflicts in heterogeneous network handover scenarios. Embed the real-time state parameters of the transmission path into the shard encapsulation process, and construct a multi-dimensional verification system covering timing jitter and spatial motion characteristics, so that the detection and repair of conflict shards can accurately fit the network fluctuation characteristics and physical scene change rules; through the dual optimization of the transmission window base reverse compensation and priority self-feedback suppression, while eliminating the multi-path sequence number misalignment, actively block the conflict regeneration path of the high-jitter channel, form a positive cycle of enhanced transmission stability and data integrity maintenance, improve the spatio-temporal alignment accuracy of video streams and sensor data in complex network environments, provide a highly reliable real-time data basis for intelligent security scenarios, and effectively avoid monitoring blind spots caused by mechanical verification rules in traditional solutions. Description of the Drawings

[0033] Figure 1 It is a schematic flow chart of the cloud monitoring system based on artificial intelligence analysis of the present invention. Detailed Embodiments

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0035] Embodiment 1: Figure 1 The cloud monitoring system based on artificial intelligence analysis of the present invention is given, including: a path marking unit, a conflict perception unit, a spatio-temporal screening unit, a cross-path correction unit, and a window adjustment unit;

[0036] Path marking unit: When encapsulating shards, attach a channel identifier and the current transmission window base to each shard, and generate a shard transmission priority label based on the instantaneous delay jitter at the moment of path switching.

[0037] Collision Awareness Unit: When receiving a shard, it extracts the sequence number offset history of the current channel according to the channel identifier, dynamically adjusts the collision determination threshold in combination with the delay jitter. If the difference between the shard sequence number and the offset exceeds the threshold, it marks the shard as a collision shard and temporarily stores it.

[0038] Spatio-temporal Screening Unit: For collision shards, by calculating the deviation degree between its actual transmission delay and the path switching jitter, and analyzing the spatial motion vector continuity of adjacent shard sensor data, it screens out the repairable shards that meet both delay tolerance and spatial consistency.

[0039] Cross-path Correction Unit: Based on the difference in the transmission window base number of the repairable shards, it reversely deduces the sequence number misalignment caused by multi-path switching, dynamically corrects the shard sequence number and verifies its continuity with the current unpacking window. After passing the verification, it is injected into the unpacking queue.

[0040] Window Adjustment Unit: According to the correction result of the shards, it freezes the window base number expansion for the channels with continuous successful corrections, and rolls back the window base number of the channels with failed corrections to the state before the collision and reduces the transmission priority of subsequent path switching.

[0041] In the cloud monitoring system, the accurate transmission and spatio-temporal alignment of video streams and sensor data are the core guarantees for the anomaly recognition and compliance forensics functions. For the multi-path parallel transmission scenario in a complex network environment (such as satellite and 4G handover), the video stream is segmented into shards with independent sequence numbers through the Z-standard algorithm, and the throughput is optimized using a dynamic window adjustment mechanism. However, multi-path link fluctuations and ACK loss will cause shard retransmission and window expansion, triggering cross-channel sequence number conflicts, which in turn trigger the unpacking engine to discard key frames by mistake and damage data integrity. To solve this problem, the present invention attaches key metadata during the shard encapsulation stage, laying a foundation for subsequent conflict detection and repair. This abstract focuses on the specific processing technical logic of the path marking unit, aiming at the message processing during the encapsulation of video shards, ensuring the accuracy and availability of the metadata, so that the collision awareness unit can rely on these metadata to achieve dynamic correction of sequence number conflicts and data alignment.

[0042] Shard encapsulation refers to the process of dividing the original data stream (such as a video stream) into multiple independent small data blocks (i.e., shards) during data transmission, and attaching necessary metadata and control information to each shard to ensure that the receiving end can correctly reconstruct and process these shards. It aims to optimize the efficiency and reliability of data transmission by decomposing large data streams into small blocks, especially suitable for scenarios with unstable network environments or limited bandwidth.

[0043] Each shard usually contains the following key elements:

[0044] Data content, which is a part of the original data stream, such as a fragment of a video frame;

[0045] A sequence number, used to identify the order of the shards in the original data stream to ensure correctness during recombination;

[0046] Metadata, such as timestamps or channel identifiers, describing the attributes and transmission paths of the shards;

[0047] And control information, such as checksums or priority tags, used to ensure data integrity and optimize transmission strategies.

[0048] Through the attached metadata and control information, shard encapsulation can achieve intelligent transmission management, such as dynamically adjusting the transmission path or priority scheduling, thereby significantly improving the overall performance of the system and the user experience.

[0049] The path marking unit includes the following:

[0050] Attachment of channel identifier:

[0051] Attach the unique identifier of the transmission channel to which each video shard belongs, called the channel identifier. The channel identifier is an integer value generated by the multipath transmission protocol during path allocation and is used to distinguish different transmission channels, such as satellite channels or mobile communication channels. The way to attach the channel identifier is to reserve a fixed number of bytes, such as four bytes, in the video shard header and write the channel identifier of the current transmission channel into this reserved field. The purpose of attaching the channel identifier is to ensure that at the receiving end, the sequence number offset history of the corresponding transmission channel can be extracted according to the channel identifier, so as to accurately identify the source channel of the video shard.

[0052] Attachment of the transmission window base number:

[0053] Attach the starting sequence number of the current transmission window to each video shard, called the transmission window base number. The transmission window base number represents the reference value of the video shard sequence numbers within the current transmission window and is updated in real time by the dynamic window adjustment mechanism according to the transmission status. The way to attach the transmission window base number is to append a fixed number of bytes, such as four bytes, to the video shard header and write the transmission window base number of the current transmission window into this appended field. The purpose of attaching the transmission window base number is to provide reference data for subsequent steps, used to reverse-derive the sequence number misalignment caused by multipath switching, so as to achieve dynamic correction of video shard sequence numbers.

[0054] Calculation of instantaneous delay jitter:

[0055] At the moment of transmission channel switching, calculate the instantaneous delay jitter of the current transmission channel, which is called instantaneous delay jitter. The instantaneous delay jitter is defined as the absolute value of the difference in transmission delays between two adjacent video segments. The specific calculation method is as follows: Record the time when each video segment is sent by the sender and the time when the receiver returns the acknowledgment signal, calculate the transmission delay of the current video segment, that is, the difference between the sending time and the time when the acknowledgment signal returns; then, subtract the transmission delay of the previous video segment from the transmission delay of the current video segment, and take the absolute value of the difference as the instantaneous delay jitter. The purpose of calculating the instantaneous delay jitter is to quantify the impact of transmission channel switching on transmission stability and provide a basis for generating priority tags. The sender maintains a delay record table for each transmission channel and calculates the instantaneous delay jitter based on the delay records of the latest two video segments when the path is switched.

[0056] Generation of priority tags:

[0057] Based on the instantaneous delay jitter, generate a transmission priority tag for the video segment, which is called a priority tag. The generation rules of the priority tag are as follows: If the instantaneous delay jitter is less than the preset low delay jitter threshold, set the priority tag to high, indicating high transmission channel stability; if the instantaneous delay jitter is greater than or equal to the low delay jitter threshold and less than the preset high delay jitter threshold, set the priority tag to medium, indicating medium transmission channel stability; if the instantaneous delay jitter is greater than or equal to the high delay jitter threshold, set the priority tag to low, indicating low transmission channel stability. The values of the priority tag are high, medium, and low, which are represented by two-bit binary codes respectively. For example, high is represented by zero zero, medium is represented by zero one, and low is represented by one zero. The processing method of generating the priority tag is to append a two-bit field to the video segment header and write the encoded value of the priority tag into this field. The purpose of generating the priority tag is to provide a dynamic adjustment basis for conflict determination in subsequent steps and a reference for transmission channel priority adjustment.

[0058] Integrate the above-generated metadata, including channel identifier, transmission window base, and priority tag, with the video segment data to generate a complete data packet. The structure of the data packet is defined as follows: First, write the channel identifier, which occupies four bytes; then write the transmission window base, which occupies four bytes; then write the priority tag, which occupies two bits; finally, append the video segment data. The processing method is to write the channel identifier, transmission window base, and priority tag into the data packet header in sequence according to the predefined format, and append the video segment data at the end to form a complete encapsulated data packet.

[0059] In the multi-path parallel transmission scenario under a complex network environment (such as satellite and 4G network handover), the video stream is segmented into shards with independent sequence numbers through a specific algorithm, and the transmission throughput is optimized through a dynamic window adjustment mechanism. However, fluctuations in the multi-path links and loss of acknowledgment messages (ACKs) may lead to shard retransmission and window expansion, causing cross-channel sequence number conflicts, and further causing the unpacking engine to erroneously discard key frames, damaging data integrity. Therefore, the encapsulated video shard data packets are processed to identify and mark the conflicting shards.

[0060] The conflict awareness unit includes the following:

[0061] Extract the historical sequence number offset of the current channel according to the channel identifier:

[0062] Extract the channel identifier from the header of the received encapsulated video shard data packet. The channel identifier is an integer value that uniquely identifies the transmission channel to which the data packet belongs. The way to extract the channel identifier is to parse the reserved field in the data packet header, such as the first four bytes, and read the channel identifier value therein. For each transmission channel, maintain a historical record of sequence number offsets. The historical sequence number offset is a list that records the sequence number offsets of the video shards received by this channel in the past period of time. The offset is defined as the difference between the video shard sequence number and the expected sequence number, and is used to reflect the sequence number change trend of this channel. According to the extracted channel identifier, query and extract the corresponding historical sequence number offset from the channel management module. The channel management module maintains a mapping table, with the channel identifier as the key and the historical sequence number offset of this channel as the value. The way to extract the historical sequence number offset is to find the mapping table through the channel identifier and obtain the list of historical sequence number offsets of this channel.

[0063] Dynamically adjust the conflict determination threshold in combination with delay jitter:

[0064] Extract the priority label from the header of the received encapsulated video fragment data packet. The priority label is generated by the sender based on the instantaneous delay jitter at the moment of path switching in the path marking unit, and its values are high, medium, or low, respectively representing the real-time stability of the transmission channel. The way to extract the priority label is to parse the reserved field in the data packet header, such as the two bytes after the channel identifier and the transmission window base number, and read the priority label encoding value therein. According to the extracted priority label, dynamically adjust the conflict determination threshold to adapt to the network conditions of the transmission channel. The conflict determination threshold is a numerical value used to determine whether the offset of the video fragment sequence number is abnormal. For example, the dynamic adjustment method can be: if the priority label is high, set the conflict determination threshold to 1.0 times the basic conflict determination threshold; if the priority label is medium, set the conflict determination threshold to 1.5 times the basic conflict determination threshold; if the priority label is low, set the conflict determination threshold to 2.0 times the basic conflict determination threshold. The basic conflict determination threshold is a preset constant, such as 5, representing the sequence number offset tolerance under ideal network conditions. The dynamic adjustment process is to select the corresponding multiplication factor according to the priority label and multiply it by the basic conflict determination threshold to obtain the current conflict determination threshold.

[0065] Determine whether the difference between the fragment sequence number and the offset exceeds the threshold:

[0066] Extract the fragment sequence number from the header of the received encapsulated video fragment data packet. The fragment sequence number is an integer value representing the sequential number of the video fragment in the video stream, and the fragment sequence number is appended by the sender in the path marking unit. The way to extract the fragment sequence number is to parse the reserved field in the data packet header, such as the four bytes after the priority label, and read the fragment sequence number value therein. According to the starting sequence number of the current unpacking window and the status of the received video fragments, calculate the currently expected video fragment sequence number, called the expected sequence number. The expected sequence number reflects the next video fragment sequence number that the receiver expects to receive, and the calculation method is to determine the next video fragment sequence number to be received according to the current status of the unpacking window. Calculate the sequence number offset of the current video fragment. The sequence number offset is defined as the difference between the video fragment sequence number and the expected sequence number. The calculation method is to subtract the calculated expected sequence number from the extracted fragment sequence number to obtain the offset. Compare the sequence number offset of the current video fragment with the historical offset in the sequence number offset history of this channel to calculate the conflict determination difference.

[0067] The conflict determination difference is defined as the absolute difference between the current sequence number offset and the average value of historical offsets. The calculation method is as follows: First, calculate the average value of all historical offsets in the sequence number offset history, then subtract this average value from the current sequence number offset, and take the absolute value to obtain the conflict determination difference. Compare the conflict determination difference with the conflict determination threshold after dynamic adjustment: If the conflict determination difference is greater than the conflict determination threshold, then determine that this video segment is a conflict segment, mark it as a conflict and temporarily store it in the conflict segment buffer; if the conflict determination difference is less than or equal to the conflict determination threshold, then determine that this video segment has no conflict and directly put it into the unpacking queue for subsequent unpacking processing.

[0068] By extracting the channel identifier, priority label, and segment sequence number for each encapsulated video segment data packet during the receiving segment stage, extracting the sequence number offset history according to the channel identifier, dynamically adjusting the conflict determination threshold in combination with the priority label, calculating the sequence number offset and the conflict determination difference, and finally determining whether the video segment is a conflict segment and performing corresponding processing. The channel identifier ensures the targeted extraction of historical data, the dynamically adjusted conflict determination threshold enhances the adaptability to network fluctuations, and the determination of the sequence number difference accurately locates abnormal video segments. These processing results provide data packets marked as conflict segments and temporarily stored for the spatio-temporal screening unit, ensuring that the subsequent repair process can perform targeted analysis and processing based on the marking information of the conflict awareness unit, thereby maintaining the integrity and spatio-temporal alignment ability of the video stream and sensor data in the cloud monitoring system under complex network environments.

[0069] The video stream is segmented into segments carrying independent sequence numbers, transmitted through multiple paths and the window is dynamically adjusted to optimize performance. However, fluctuations in the network link and loss of acknowledgment messages may trigger segment retransmission or window expansion, resulting in cross-channel sequence number conflicts, causing the unpacking engine to accidentally discard key frames, and thus destroying data integrity. To solve this problem, the spatio-temporal screening unit focuses on the screening and repair of conflict segments. By analyzing the transmission delay and the spatial consistency of sensor data, it screens out repairable segments, providing reliable input for subsequent sequence number correction and injection into the unpacking queue, thereby ensuring the integrity and spatio-temporal alignment ability of data transmission.

[0070] The spatio-temporal screening unit includes the following:

[0071] Obtain the deviation degree of the actual transmission delay from the path switching jitter:

[0072] For each conflict segment data packet, evaluate its transmission delay characteristics to determine whether the delay anomaly is within the tolerable range.

[0073] Calculation of the actual transmission delay:

[0074] The actual transmission delay is defined as the time difference between when a conflict shard packet is sent from the sender and received at the receiver. The way to calculate the actual transmission delay is as follows: extract the sending timestamp and the receiving timestamp from the header of the conflict shard packet. The sending timestamp is the sending time recorded when the shard is encapsulated at the sender, and the receiving timestamp is the receiving time recorded when the shard is received at the receiver. Subtract the sending timestamp from the receiving timestamp to obtain the actual transmission delay.

[0075] Determination of path switching jitter:

[0076] Path switching jitter represents the degree of change in the channel delay at the moment of path switching. The receiver determines the path switching jitter based on the priority label extracted from the header of the conflict shard packet. The priority label is generated by the sender in the path marking unit based on the instantaneous delay jitter at the moment of path switching, and the values are high, medium, or low. According to the different priority labels, the path switching jitter is set to a predefined low jitter value, medium jitter value, or high jitter value, corresponding to high, medium, and low priority labels respectively. For example, the low jitter value can be set to 10 milliseconds, the medium jitter value to 30 milliseconds, and the high jitter value to 50 milliseconds.

[0077] The priority label reflects the network condition of the sender at the moment of path switching. By associating it with the path switching jitter, it can provide the receiver with an expectation of the network fluctuation during the transmission of the conflict shard packet.

[0078] Calculation of deviation:

[0079] Deviation is defined as the degree of difference between the actual transmission delay and the path switching jitter. The way to calculate deviation is as follows: first calculate the absolute value of the difference between the actual transmission delay and the path switching jitter, and then divide this absolute difference by the path switching jitter to obtain a dimensionless ratio, which is the deviation. The smaller the deviation, the closer the actual transmission delay is to the path switching jitter, indicating that the delay characteristics of the conflict shard packet conform to the expected fluctuation at the time of path switching.

[0080] Analyze the continuity of the spatial motion vectors of adjacent shard sensor data:

[0081] Use sensor data to evaluate the spatial consistency of conflict shard packets and ensure that they match the motion characteristics of the video stream context.

[0082] Extraction of spatial motion vectors:

[0083] Each conflicting shard packet contains sensor data associated with a video frame, recording spatial information. The sensor data in the conflicting shard packet is extracted, and the spatial motion vector is calculated. The spatial motion vector is a three-dimensional vector representing the motion speed of an object in space, corresponding to the components in the x, y, and z directions respectively. The sensor data is encapsulated with the video shard at the sending end and obtained by the receiving end through parsing the conflicting shard packet.

[0084] The spatial motion vector reflects the motion state of the object in the video frame and can be used to evaluate the continuity of the conflicting shard packet and adjacent shard packets in the spatial context.

[0085] Identification of adjacent shard packets:

[0086] According to the shard sequence number extracted from the header of the conflicting shard packet, the adjacent shard packets before and after it are identified. The shard sequence number is appended by the sending end in the path marking unit, indicating the sequential number of the shard in the video stream. The previous shard packet is the shard packet with a sequence number 1 less than that of the conflicting shard packet, and the next shard packet is the shard packet with a sequence number 1 greater than that of the conflicting shard packet. If the previous shard packet or the next shard packet is not received, this conflicting shard packet will not participate in the continuity analysis for the time being.

[0087] The spatial motion vectors of the previous shard packet and the next shard packet can provide the context information of the conflicting shard packet in the video stream for evaluating its spatial consistency.

[0088] Calculation of the continuity metric:

[0089] The continuity metric is defined as the relative change rate between the spatial motion vector of the conflicting shard packet and the spatial motion vectors of adjacent shard packets. The processing method for calculating the continuity metric is as follows: First, calculate the modulus of the difference between the spatial motion vector of the conflicting shard packet and the spatial motion vector of the previous shard packet, and then divide it by the modulus of the spatial motion vector of the previous shard packet to obtain a relative change rate; similarly, calculate the relative change rate between the spatial motion vectors of the conflicting shard packet and the next shard packet; finally, take the minimum value of these two relative change rates as the continuity metric. The modulus of the spatial motion vector is obtained by calculating the square root of the sum of the squares of its three components. The continuity metric is a dimensionless ratio, and the smaller the value, the higher the continuity of the conflicting shard packet and adjacent shard packets in spatial motion.

[0090] By calculating the continuity metric, the similarity of the conflicting shard packet and adjacent shard packets in motion characteristics can be quantified to ensure the consistency of the spatial context of the conflicting shard packet in the video stream.

[0091] Screening repairable shard packets:

[0092] Based on the above calculation results, set the screening conditions to screen out repairable shard data packets.

[0093] Judgment of the delay tolerance condition:

[0094] Judge whether the deviation degree is less than the preset deviation degree threshold, such as 0.2. If the deviation degree is less than the deviation degree threshold, it indicates that the difference between the actual transmission delay of the conflicting shard data packet and the path switching jitter is within an acceptable range, meeting the delay tolerance requirement. The delay tolerance condition ensures that the abnormal transmission delay of the conflicting shard data packet is within a controllable range, avoiding misjudging the conflicting shard data packet with too large or too small delay as a repairable shard data packet.

[0095] Judgment of the spatial consistency condition:

[0096] Judge whether the continuity metric is less than the preset continuity metric threshold, such as 0.1. If the continuity metric is less than the continuity metric threshold, it indicates that the conflicting shard data packet and the adjacent shard data packet have high continuity in the spatial motion vector, meeting the spatial consistency requirement. The spatial consistency condition ensures that the content of the conflicting shard data packet in the video stream matches that of the adjacent shard data packet, avoiding including the conflicting shard data packet with discontinuous content in the repair scope.

[0097] Application of the screening rule:

[0098] If the conflicting shard data packet meets both the delay tolerance condition and the spatial consistency condition, that is, the deviation degree is less than the deviation degree threshold and the continuity metric is less than the continuity metric threshold, then mark the conflicting shard data packet as a repairable shard data packet and transfer it to the path correction unit for sequence number correction; otherwise, mark the conflicting shard data packet as an irreparable shard data packet and discard it.

[0099] By comprehensively considering the transmission delay and spatial consistency, the repairability of the conflicting shard data packet can be comprehensively evaluated, ensuring that the screened conflicting shard data packet is both acceptable in terms of transmission characteristics and matches the video stream in terms of content.

[0100] By calculating the deviation degree between the actual transmission delay and the path switching jitter and the spatial motion vector continuity of the adjacent shard data packet sensor data in the spatio-temporal screening unit, the repairable shard data packets that meet both the delay tolerance and spatial consistency are screened out. The deviation degree quantifies the degree of coincidence between the transmission delay of the conflicting shard data packet and the path switching expectation, and the continuity metric ensures the spatial context continuity of the conflicting shard data packet in the video stream. These screening results provide reliable repairable shard data packets for the path correction unit, ensuring that the subsequent sequence number correction based on the transmission window base difference can accurately restore the shard order, thereby maintaining the integrity and spatio-temporal alignment ability of the video stream and sensor data in the cloud monitoring system in a complex network environment.

[0101] The path correction unit addresses the sequence number misalignment problem caused by multi-path switching for the repairable shard data packets screened by the spatio-temporal screening unit.

[0102] The path correction unit includes the following:

[0103] Step 1. Calculate the difference in transmission window base numbers:

[0104] For each repairable shard data packet, calculate the difference between its transmission window base number and the current unpacking window base number to evaluate the degree of shard sequence number misalignment. The current unpacking window base number is defined as the starting sequence number of the window currently processed by the unpacking engine, which is maintained in real time by the unpacking engine and reflects the unpacking status of the receiving end. The calculation process of the transmission window base number difference is as follows: Extract the transmission window base number from the header of the repairable shard data packet. This value is attached by the sender during shard encapsulation in the path marking unit, indicating the shard window position of the sender. Then, subtract the current unpacking window base number from the extracted transmission window base number to obtain the transmission window base number difference.

[0105] The current unpacking window base number, as the benchmark for the processing progress of the receiving end, is compared with the transmission window base number of the repairable shard data packet, which can quantify the window offset between the two. The transmission window base number difference reflects the relative position of the window to which the shard belongs and the current unpacking window. A positive value indicates that the shard comes from a future window, and a negative value indicates that it comes from a past window, intuitively showing the direction and degree of sequence number misalignment.

[0106] Step 2. Reverse-derive the sequence number misalignment amount caused by multi-path switching:

[0107] Based on the transmission window base number difference and channel history information, reverse-derive the sequence number misalignment amount caused by multi-path switching, and calculate it using a non-linear adjustment strategy. The channel history misalignment amount is maintained by the channel management module, recording the sequence number offset accumulated by the channel during past multi-path switching, and is updated through the historical sequence number offset amount extracted by the conflict awareness unit. The calculation process of the sequence number misalignment amount is as follows: First, extract the priority label from the header of the repairable shard data packet; determine the priority coefficient according to the priority label. When the priority label is high, the priority coefficient takes twice the reference value; when it is medium, it takes the reference value; when it is low, it takes half of the reference value; then, calculate the absolute value of the transmission window base number difference and the absolute value of the channel history misalignment amount; then, multiply the transmission window base number difference by the priority coefficient and divide it by the sum of the absolute value of the transmission window base number difference plus one to obtain the first part of the result; multiply the channel history misalignment amount by the priority coefficient and divide it by the difference of the sum of the absolute value of the channel history misalignment amount plus one to obtain the second part of the result; finally, add the first part of the result and the second part of the result to obtain the sequence number misalignment amount.

[0108] For example, the calculation method of the sequence number misalignment amount can be as follows:

[0109]

[0110] Wherein:

[0111] PT coef : Priority coefficient, mapped by the shard transmission priority tag (PT) generated by the path marking unit:

[0112] PT = high, TP coef = 2.0;

[0113] PT = medium, TP coef = 1.0;

[0114] PT = low, PT coef = 0.5.

[0115] This coefficient reflects the impact of the transmission priority of the channel on the misalignment calculation.

[0116] |WBD| and |HDM|: Respectively, the absolute value of the difference in the transmission window base and the channel historical misalignment amount, used for non-linear normalization to avoid overfitting caused by a single weight factor. By and introducing non-linear attenuation to balance the contributions of the current window difference and the historical misalignment, ensuring that high-priority channels rely more on real-time data and low-priority channels rely more on historical trends.

[0117] Step 3. Dynamically correct the shard sequence number:

[0118] Based on the calculated sequence number misalignment amount, correct the shard sequence number of the repairable shard data packet. The calculation process of the corrected shard sequence number is as follows: Extract the original shard sequence number from the header of the repairable shard data packet. This sequence number is encapsulated by the sender in the path marking unit and represents the sequential number of the shard in the video stream; then, subtract the sequence number misalignment amount from the original shard sequence number to obtain the corrected shard sequence number.

[0119] By subtracting the sequence number misalignment amount, the shard sequence number is adjusted to its correct position in the video stream, which can eliminate the sequence number misalignment caused by multi-path switching and ensure that the shard sequence number is consistent with the actual transmission order.

[0120] Step 4. Verify the continuity with the current unpacking window:

[0121] Verify whether the corrected shard sequence number is continuous with the current unpacking window to determine whether to inject it into the unpacking queue. The range of the unpacking window is defined as the interval from the current unpacking window base number to the current unpacking window base number plus the window size minus one, and the window size is preset by the unpacking engine. The judgment process of the continuity verification is as follows: check whether the corrected shard sequence number meets any of the following conditions: one is that it falls within the unpacking window range; the second is that it is equal to the current unpacking window base number minus one, that is, continuous with the lower boundary of the window; the third is that it is equal to the current unpacking window base number plus the window size, that is, continuous with the upper boundary of the window. If any of the conditions are met, it is considered to pass the verification; if none of them are met, it is marked as unavailable and the shard data packet is discarded.

[0122] The range of the unpacking window defines the range of shard sequence numbers that the receiving end can currently accept, and the continuity verification ensures that the corrected shard sequence number matches the progress of the current unpacking window, avoiding unpacking errors caused by injecting discontinuous shards.

[0123] Step 5. Inject into the unpacking queue:

[0124] For the shard data packets that pass the continuity verification, insert them into the correct position in the unpacking queue according to the corrected shard sequence number to ensure that the unpacking engine processes them in order. The insertion process is as follows: according to the corrected shard sequence number, find the corresponding sequential position in the unpacking queue, place the shard data packet at this position, and maintain the sequence continuity of the queue. Inject the shard data packets into the unpacking queue in the order of the corrected shard sequence number to ensure that the shard data received by the unpacking engine is arranged in the correct order and maintain the spatio-temporal consistency of the video stream and sensor data.

[0125] Based on the repairable shard data packets output by the spatio-temporal screening unit, the path correction unit calculates the difference in the transmission window base number, combines the historical channel misalignment amount and the non-linear adjustment of the priority coefficient to deduce the sequence number misalignment amount, then dynamically corrects the shard sequence number, verifies its continuity with the current unpacking window, and finally injects the shard data packet into the unpacking queue. This process addresses the sequence number misalignment problem in multi-path transmission in the cloud monitoring system, integrating real-time and historical data to ensure the integrity and spatio-temporal alignment of the video stream and sensor data.

[0126] In the cloud monitoring system, the reliable transmission of video streams and sensor data is a core requirement for anomaly recognition and compliance forensics. It needs to address the challenges of multi-path transmission in complex network environments, such as cross-channel sequence number conflicts caused by network fluctuations and fragment retransmission. The aforementioned path marking unit to cross-path correction unit has completed fragment encapsulation, conflict detection, repairable fragment screening, and sequence number correction, and successfully injected the corrected fragment data packets into the unpacking queue, ensuring the initial continuity of video streams and sensor data. However, the continuous fluctuations in the network environment may lead to changes in the stability of the transmission channels, and some channels may affect the overall transmission performance due to correction failures. Therefore, based on the correction results of the path correction unit, the window adjustment unit further optimizes the configuration of the transmission channels, and improves the transmission stability and data integrity by dynamically adjusting the window base and transmission priority, providing reliable guarantee for subsequent data processing.

[0127] The window adjustment unit includes the following:

[0128] Statistical stability of continuously corrected successful channels:

[0129] For each transmission channel, according to the correction results of the fragments in the path correction unit, calculate its stability index of continuously corrected success. The specific processing logic is to maintain a continuous correction success count for the identifier of each transmission channel, and its initial value is set to zero. When the fragment correction of a certain transmission channel is successful, increase the continuous correction success count of this transmission channel by one; when the fragment correction of this transmission channel in the path correction unit fails, reset the continuous correction success count of this transmission channel to zero. At the same time, a stability threshold is preset as the basis for judging whether the transmission channel is stable. The stability threshold is a fixed value used to compare with the continuous correction success count to determine the stability state of the transmission channel.

[0130] Freeze the expansion of the window base:

[0131] For the transmission channels with continuously corrected success, perform the freezing operation of the transmission window base expansion to maintain its stability. First, check whether the continuous correction success count of each transmission channel is greater than or equal to the preset stability threshold. If the continuous correction success count of a certain transmission channel reaches or exceeds the stability threshold, lock the transmission window base of this transmission channel to the current value, and prohibit any increase operation of the transmission window base within a preset freezing duration. The freezing duration is a predetermined time period used to ensure that the transmission window base remains unchanged for a period of time.

[0132] Roll back the window base to the state before the conflict:

[0133] For a transmission channel with failed fragment correction, its transmission window base is rolled back to the stable state before the conflict. Maintain a pre - conflict window base for each transmission channel, recording the transmission window base value of this transmission channel when the last fragment correction was successful. Each time the fragment correction of a certain transmission channel is successful, update the pre - conflict window base of this transmission channel to the current transmission window base; when the fragment correction fails, keep the pre - conflict window base unchanged. If the fragment correction of a certain transmission channel fails, adjust the current transmission window base of this transmission channel to the recorded pre - conflict window base.

[0134] The pre - conflict window base reflects the last stable state of the transmission channel before the conflict. Rolling back the transmission window base to the pre - conflict window base can eliminate the sequence number misalignment problem caused by window base expansion, thus restoring the stable transmission ability of the transmission channel.

[0135] Reduce the transmission priority for subsequent path switching:

[0136] For a transmission channel with failed fragment correction, adjust its transmission priority to reduce the traffic allocation to unstable transmission channels. Maintain a transmission priority for each transmission channel, with its initial value set to a reference value. When the fragment correction of a certain transmission channel fails, lower the transmission priority of this transmission channel. The specific adjustment method is to multiply the current transmission priority by a decay factor, which is determined by a non - linear calculation based on the size of the consecutive correction success count, ensuring that the reduction amplitude of the transmission priority is related to the instability degree of the transmission channel. When the fragment correction of a certain transmission channel is successful, slowly restore its transmission priority. The specific adjustment method is to add a recovery factor to the current transmission priority, which is calculated based on the size of the consecutive correction success count, ensuring that the transmission priority only gradually increases when the transmission channel remains stable. In subsequent path switching, allocate traffic proportions according to the transmission priorities of each transmission channel, and the transmission channels with lower transmission priorities receive less traffic allocation.

[0137] For example, the transmission priority can be dynamically adjusted in the following way:

[0138] Maintain a transmission priority TP for each channel, with an initial value of 1.0, representing the default priority.

[0139] When the fragment correction fails, lower TP, and the calculation formula is:

[0140]

[0141] Parameter explanation: CSC is the consecutive correction success count of the current channel (0 when it fails);

[0142] In the formula is the attenuation factor, ensuring that the priority adjustment is closely related to the channel stability. The priority of unstable channels is rapidly attenuated through an exponential function.

[0143] When the shard correction is successful, TP is slowly restored, and the calculation formula is:

[0144]

[0145] TP new represents the transmission priority of a certain transmission channel at the current moment. TP current represents the new value obtained after adjusting the transmission priority of a certain transmission channel, that is, the updated transmission priority.

[0146] In the formula is the recovery factor.

[0147] The recovery process is smooth and controlled by CSC, ensuring that the priority is gradually increased only when the channel remains stable.

[0148] By reducing the transmission priority of the transmission channels with failed shard corrections, the traffic allocation to unstable transmission channels can be reduced, so as to preferentially utilize the stable transmission channels with higher transmission priorities. This dynamic adjustment method can optimize the overall transmission strategy and improve the reliability of data transmission. The mechanism of slowly restoring the transmission priority avoids frequent adjustments caused by short-term fluctuations.

[0149] The window adjustment unit realizes the dynamic optimization of the transmission channels based on the correction results of the shards in the path correction unit. By freezing the expansion of the transmission window base for the continuously corrected successful transmission channels, rolling back the transmission window base of the transmission channels with failed shard corrections to the state before the conflict and reducing their transmission priorities, it ensures the stable transmission and spatio-temporal alignment of the video stream and sensor data in a complex network environment.

[0150] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0151] It should be noted that the system of the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminals with a user interface, so as to meet various hardware environments and usage requirements.

[0152] Only some exemplary embodiments of the present invention are described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

[0153] It should be noted that in this text, if there are relational terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent in such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the process, method, article or device comprising the element.

[0154] As described above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the said claims.

Claims

1. A cloud monitoring system based on artificial intelligence analysis, characterized in that, Including the steps: A path marking unit, a conflict awareness unit, a spatio-temporal screening unit, a cross-path correction unit, and a window adjustment unit; Path marking unit: When performing shard encapsulation, attach a channel identifier and the current transmission window base number to each shard, and generate a shard transmission priority label based on the instantaneous delay jitter at the moment of path switching; Conflict awareness unit: When receiving a shard, extract the historical sequence number offset of the current channel according to the channel identifier, dynamically adjust the conflict determination threshold in combination with the delay jitter. If the difference between the shard sequence number and the offset exceeds the threshold, mark it as a conflict shard and temporarily store it; Spatio-temporal screening unit: For conflict shards, by calculating the deviation degree between its actual transmission delay and the path switching jitter, and analyzing the spatial motion vector continuity of adjacent shard sensor data, screen out the repairable shards that meet both delay tolerance and spatial consistency; Cross-path correction unit: Based on the difference in the transmission window base numbers of the repairable shards, inversely deduce the sequence number misalignment caused by multi-path switching, dynamically correct the shard sequence number and verify its continuity with the current unpacking window. After passing the verification, inject it into the unpacking queue; Window adjustment unit: According to the correction results of the shards, freeze the window base number expansion for the channels with consecutive successful corrections, and roll back the window base number of the channels with failed corrections to the state before the conflict and reduce the transmission priority of subsequent path switching.

2. The cloud monitoring system based on artificial intelligence analysis according to claim 1, characterized in that The path marking unit includes the following: In the shard encapsulation stage, attach a channel identifier, a transmission window base number, and a priority label to each video shard. The channel identifier is defined as the unique identifier of the transmission channel, the transmission window base number is defined as the starting sequence number of the current transmission window, and the priority label is determined based on the comparison result between the instantaneous delay jitter and a preset threshold. The instantaneous delay jitter is defined as the absolute value of the difference in transmission delays between adjacent video shards.

3. The cloud monitoring system based on artificial intelligence analysis according to claim 2, wherein The conflict awareness unit includes the following: Query the historical sequence number offset according to the channel identifier, and dynamically adjust the conflict determination threshold according to the priority label. By calculating the sequence number offset between the shard sequence number and the expected sequence number, and the conflict determination difference between the sequence number offset and the average value of the historical offsets in the sequence number offset history. When the conflict determination difference is greater than the conflict determination threshold, mark the video shard as a conflict shard and temporarily store it. When the conflict determination difference is not greater than the conflict determination threshold, put the video shard into the unpacking queue.

4. The cloud monitoring system based on artificial intelligence analysis according to claim 3, wherein The spatio-temporal screening unit includes the following: By calculating the deviation degree between the actual transmission delay and the path switching jitter, and analyzing the spatial motion vector continuity of adjacent shard sensor data, screen out the repairable shards that meet both delay tolerance and spatial consistency. When the deviation degree is less than the deviation degree threshold and the spatial motion vector continuity is less than the continuity metric threshold, mark the conflict shard as a repairable shard.

5. The cloud monitoring system based on artificial intelligence analysis according to claim 4, characterized in that, The spatio-temporal screening unit also includes the following: Where the actual transmission delay is defined as the difference between the reception timestamp and the transmission timestamp, the path switching jitter is determined based on the priority label, the deviation degree is calculated as the absolute value of the difference between the actual transmission delay and the path switching jitter divided by the path switching jitter, and the spatial motion vector continuity is obtained by calculating the minimum value of the relative change rate of the spatial motion vectors between the conflict shard and the adjacent shards.

6. The cloud monitoring system based on artificial intelligence analysis according to claim 5, wherein, The path correction unit includes the following: For repairable fragmented data packets, by calculating the difference between the transmission window base number and the current unpacking window base number, combining the non-linear adjustment of the channel historical misalignment amount and the priority coefficient, the sequence number misalignment amount is deduced. The fragmented sequence number is dynamically corrected based on the sequence number misalignment amount, and the continuity between the corrected fragmented sequence number and the current unpacking window is verified. The fragmented data packet that passes the verification is injected into the unpacking queue.

7. The cloud monitoring system based on artificial intelligence analysis according to claim 6, wherein, The path correction unit further includes the following: The processing process of the corrected fragmented sequence number is as follows: Extract the original fragmented sequence number from the header of the repairable fragmented data packet; then, subtract the sequence number misalignment amount from the original fragmented sequence number to obtain the corrected fragmented sequence number.

8. The cloud monitoring system based on artificial intelligence analysis according to claim 7, wherein The path correction unit further includes the following: The judgment process of the continuity verification is as follows: Check whether the corrected fragmented sequence number meets any of the following conditions: One is that it falls within the unpacking window range; the other is that it is equal to the current unpacking window base number minus one, that is, continuous with the window lower boundary; the third is that it is equal to the current unpacking window base number plus the window size, that is, continuous with the window upper boundary; if any condition is met, it is considered to pass the verification; if none of them are met, it is marked as unavailable and the corresponding fragmented data packet is discarded.

9. The cloud monitoring system based on artificial intelligence analysis according to claim 8, characterized in that, The path correction unit further includes the following: The priority coefficient is obtained by mapping the fragmentation transmission priority label generated by the path marking unit.

10. The cloud monitoring system based on artificial intelligence analysis according to claim 6, characterized in that, The window adjustment unit includes the following: According to the fragmentation correction result, the stability of the transmission channel is judged by counting the continuous correction success count. For the transmission channel whose continuous correction success count reaches the stability threshold, the expansion of the transmission window base number is frozen. For the transmission channel with fragmented correction failure, the transmission window base number is rolled back to the window base number before the conflict, and the transmission priority is adjusted through non-linear calculation.

Images