Intelligent grading method and system for quality of heterogeneous video equipment

Abstract

The invention provides an intelligent grading method and system for the quality of heterogeneous video equipment, and relates to the technical field of equipment quality grading, and the method comprises the steps: S1, collecting various performance parameters of a plurality of different video equipment in an equipment group, and collecting various dynamic performance indexes of each video equipment in real time in the operation process of the equipment group; standardizing the performance parameters and the dynamic performance indexes to obtain a standardized performance parameter set; and inputting the standardized performance parameter set into a pre-trained weight model for comprehensive scoring to obtain a scoring result. The method has the beneficial effects that the one-sidedness of scoring only depending on a single index in a traditional method is avoided, and the actual performance of the equipment in different application scenes can be reflected more comprehensively; according to the step, scoring deviation caused by equipment bottom layer difference is eliminated, so that heterogeneous video equipment of different brands, models and coding standards can be compared under the same evaluation benchmark, and the scoring fairness and objectivity are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of device quality scoring, and in particular to an intelligent scoring method and system for heterogeneous video device quality. Background Art

[0002] In the field of video equipment quality assessment in the security industry, traditional methods are mainly divided into:

[0003] 1. Manual inspection and scoring: This method relies on operations and maintenance personnel to regularly inspect and score metrics such as video quality, online status, and installation location. This method offers significant flexibility, but is limited by labor costs, subjective experience differences, and inspection cycles, making it difficult to achieve high-frequency, comprehensive evaluations. However, manual scoring or automated scoring based on a single metric can be highly subjective and inaccurate, leading to significant subjectivity.

[0004] 2. Single-metric automatic scoring: This method uses automated tools to quantify scores based on a single metric, such as device online rate, video fluency, or image clarity. While this method improves efficiency, it ignores key metrics like frame rate, latency, packet loss rate, and color reproduction, failing to fully reflect device performance. It also fails to consider device hardware performance, network quality, and environmental factors, resulting in a single, one-sided metric.

[0005] Therefore, in the current security industry, the quality assessment methods for heterogeneous video devices with multiple brands, models, and different encoding standards have the following main problems:

[0006] 1. Existing methods primarily base scoring on static standards (e.g., 1080P / 30fps), often focusing solely on resolution or bitrate, and fail to comprehensively consider multiple key metrics such as frame rate, latency, packet loss rate, color reproduction, and night vision. This single-metric evaluation approach fails to fully reflect a device's actual performance in different application scenarios and lacks the ability to monitor and comprehensively model its dynamic performance (e.g., low-light performance and adaptability to network fluctuations).

[0007] 2. Currently, most scoring schemes do not incorporate a device characteristic normalization model and are unable to automatically adjust evaluation criteria based on the underlying characteristics of the device. Scoring methods typically assume device homogeneity and ignore the essential differences in image quality performance across different encoding protocols (such as H.264, H.265, and MPEG-4) and different hardware platforms (such as CMOS sensors vs. CCD sensors). As a result, a unified scoring standard cannot reasonably evaluate heterogeneous devices.

[0008] 3. Existing detection methods mostly rely on manual inspections or periodic detection, and are mostly based on post-detection or regular spot checks. They are unable to achieve real-time perception and instant scoring of changes in equipment operating status (such as lens contamination, focal length offset, and worsening network packet loss), resulting in delayed detection of equipment anomalies. There is a lack of continuous quality monitoring and automatic evaluation mechanisms based on real-time data streams (such as video streams and network packet streams), and there are insufficient real-time and dynamic features, which affects the overall security effectiveness of the system.

[0009] 4. There is a lack of machine learning or statistical learning mechanisms to train and optimize the scoring criteria. Most current scoring systems are fixed rule engines, which make it difficult to dynamically adjust the scoring weights based on environmental changes (such as day and night changes, seasonal light changes) or historical usage data. This leads to rigid scoring models, poor adaptability, and scoring results that deviate from the actual usage experience.

[0010] Therefore, the above existing technologies cannot cope with large-scale equipment when dealing with heterogeneous equipment scenarios, and cannot effectively meet the management needs of diverse heterogeneous equipment at the city level. Summary of the Invention

[0011] In response to the problems existing in the prior art, the present invention provides an intelligent scoring method for the quality of heterogeneous video devices, comprising:

[0012] Step S1, collecting various performance parameters of a plurality of different video devices in a device group, and collecting various dynamic performance indicators of each of the video devices in real time during the operation of the device group;

[0013] Step S2, normalizing the performance parameters and the dynamic performance indicators to obtain a standardized performance parameter set;

[0014] Step S3: Input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

[0015] Preferably, the step S1 includes:

[0016] Step S11, collecting basic characteristic data of each of the video devices and adding it to the performance parameters;

[0017] Step S12, collecting the operational effectiveness index of each of the video devices and adding it to the performance parameters;

[0018] Step S13: collecting multiple dynamic performance indicators of each of the video devices in real time during the operation of the device group.

[0019] Preferably, the training process of the weight model includes:

[0020] The manual performance benchmark score of each video device in the device group is collected and added to the manual initial score set, and the standardized performance parameter set and the manual initial score set collected from each video device in the device group are input into the initial machine learning model for model training. After completing multiple trainings, the weight model is obtained.

[0021] Preferably, the basic characteristic data includes at least one of resolution, encoding format, frame rate, bit rate, and delay.

[0022] Preferably, the operation effectiveness indicator includes at least one of call records, effective operation times, and response delay.

[0023] Preferably, the dynamic performance indicators include: video clarity, frame rate stability, image delay, network packet loss rate, color reproduction accuracy, and night vision imaging quality.

[0024] Preferably, the standardization processing includes: normalizing the performance parameters and the dynamic performance indicators, or performing dimensionality reduction processing on the performance parameters and the dynamic performance indicators using principal component analysis to obtain the standardized performance parameter set.

[0025] Preferably, before outputting the scoring result in step S3, a scoring result review process is also included, including: automatically modifying the score or prompting manual review when the scoring result exceeds a preset score range.

[0026] Preferably, different scenarios are configured, and each scenario is associated with a different weight ratio. The weight model adjusts the scoring result of the comprehensive score according to the weight ratio corresponding to the scenario configuration.

[0027] The present invention also provides an intelligent scoring method for the quality of heterogeneous video devices, which uses the above-mentioned intelligent scoring method, including:

[0028] A data acquisition module is used to acquire various performance parameters of a plurality of different video devices in the device group, and to acquire various dynamic performance indicators of each of the video devices in real time during the operation of the device group;

[0029] a data processing module, connected to the data acquisition module, for performing standardization processing on each of the performance parameters and the dynamic performance indicators to obtain a standardized performance parameter set;

[0030] The scoring module is connected to the data processing module and is used to input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

[0031] The above technical solution has the following advantages or beneficial effects:

[0032] 1. By collecting multiple performance parameters of each video device in step S1 and dynamic performance indicators collected in real time in step S2, multi-dimensional data collection of device performance is achieved, solving the problem of single-indicator evaluation and lack of multi-dimensional comprehensive analysis. This avoids the one-sidedness of traditional methods that rely solely on a single indicator for scoring, and can more comprehensively reflect the actual performance of the device in different application scenarios.

[0033] 2. In step S2, this method standardizes the collected characteristic data of different types, normalizing different device parameters to a unified evaluation benchmark. This step eliminates scoring bias caused by underlying device differences (such as different encoding protocols and hardware platforms), ensuring the comparability of subsequent scores. It enables heterogeneous video devices of different brands, models, and encoding standards to be compared under the same evaluation benchmark. The normalization of device characteristics solves the problem of lack of intelligent adaptation to heterogeneous device differences and improves the fairness and objectivity of scoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 1 is a flow chart of an intelligent scoring method for the quality of heterogeneous video devices in a preferred embodiment of the present invention;

[0035] Figure 2 Schematic diagram of a sub-flow chart of step S1 in a preferred embodiment of the present invention;

[0036] Figure 3 The figure is a schematic structural diagram of an intelligent scoring system for the quality of heterogeneous video devices in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0037] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment, and other embodiments may also fall within the scope of the present invention as long as they conform to the gist of the present invention.

[0038] In a preferred embodiment of the present invention, based on the above problems existing in the prior art, an intelligent scoring method for the quality of heterogeneous video devices is provided. Figure 1 Shown, including:

[0039] Step S1, collecting various performance parameters of multiple different video devices in the device group, and collecting various dynamic performance indicators of each video device in real time during the operation of the device group;

[0040] Step S2, normalizing each performance parameter and dynamic performance index to obtain a standardized performance parameter set;

[0041] Step S3: Input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

[0042] Specifically, most current device quality testing methods usually only use resolution or bit rate as the main scoring basis, and fail to comprehensively consider multiple key indicators such as frame rate, latency, packet loss rate, color reproduction, night vision effect, etc. This single-indicator evaluation method cannot fully reflect the actual performance of the device in different application scenarios. The main reason is that the existing methods mainly formulate scoring rules based on static standards (such as 1080P / 30fps), and lack the ability to monitor and comprehensively model the dynamic performance of the device (such as performance in low-light environments and adaptability to network fluctuations in real time).

[0043] Moreover, in the existing technology, most current scoring schemes do not introduce a device characteristic normalization processing model, cannot automatically adjust the evaluation criteria based on the underlying characteristics of the device, and lack intelligent adaptation to the differences between heterogeneous devices. The scoring method usually assumes device homogeneity and ignores the essential differences in image quality performance between different encoding protocols (such as H.264, H.265, MPEG-4) and different hardware platforms (such as CMOS sensors vs. CCD sensors). As a result, a unified scoring standard cannot reasonably evaluate heterogeneous devices.

[0044] In this embodiment, to address the problem that heterogeneous devices have differences in basic characteristics such as encoding protocols and sensor types, which leads to the lack of applicability of traditional scoring methods, the present invention designs a device characteristic normalization processing mechanism that can adaptively adjust the scoring criteria based on the underlying parameters of the device to achieve fair and accurate evaluation of multiple heterogeneous devices. Specifically, the device group in this embodiment includes video devices of multiple brands and models, and basic characteristic data of each device is collected, including but not limited to:

[0045] Encoding format (such as H.264, H.265, MPEG-4, etc.)

[0046] Resolution (e.g., 1920×1080, 1280×720, etc.)

[0047] Frames Per Second (FPS)

[0048] Bitrate

[0049] Sensor type (e.g., CMOS or CCD)

[0050] By collecting multiple performance parameters of each video device (such as resolution, frame rate, bit rate, encoding format, etc.) in step S1 and dynamic performance indicators (such as video clarity, frame rate stability, image delay, network packet loss rate, color reproduction accuracy, night vision imaging quality, etc.) collected in real time in step S2, multi-dimensional data collection of device performance is achieved, solving the problem of single indicator evaluation and lack of multi-dimensional comprehensive analysis. This avoids the one-sidedness of traditional methods that rely only on a single indicator (such as resolution or bit rate) for scoring, and can more comprehensively reflect the actual performance of the device in different application scenarios.

[0051] In step S2, this method normalizes the collected characteristic data of different types, normalizing different device parameters to a unified evaluation benchmark. This step eliminates scoring bias caused by underlying device differences (such as different encoding protocols and hardware platforms), ensuring the comparability of subsequent scores. This allows heterogeneous video devices of different brands, models, and encoding standards to be compared under the same evaluation benchmark. Normalizing device characteristics addresses the lack of intelligent adaptation to heterogeneous device differences and improves the fairness and objectivity of scoring.

[0052] Through real-time device data collection in steps S1 and S2, a dynamic indicator collection and instant scoring mechanism based on real-time data streams can complete detection and score adjustment within 3 minutes after a device status change (such as lens contamination, focal length offset, network jitter, etc.). This real-time data collection mechanism can promptly detect device anomalies and feedback score changes. Compared with traditional methods that rely on manual inspections or periodic detection (usually 24 hours or longer), the system response speed is approximately 480 times faster, providing a foundation for subsequent dynamic scoring, achieving real-time perception and instant scoring of changes in device operating status, greatly improving the overall system operation and maintenance efficiency, and solving the problems of insufficient real-time and dynamic performance.

[0053] In step S3, the calculation formula for the comprehensive scoring result in the weight model is as follows:

[0054]

[0055] Among them, w i is the weight of the i-th indicator, v i is the standardized value of the ith indicator, and n is the total number of indicators.

[0056] In a preferred embodiment of the present invention, Figure 2 As shown, step S1 includes:

[0057] Step S11, collecting basic characteristic data of each video device and adding performance parameters;

[0058] Step S12, collecting the operational effectiveness index of each video device and adding it to the performance parameters;

[0059] Step S13: collecting multiple dynamic performance indicators of each video device in real time during the operation of the device group.

[0060] In a preferred embodiment of the present invention, the basic characteristic data includes at least one of resolution, encoding format, frame rate, bit rate, and delay.

[0061] In a preferred embodiment of the present invention, the operation effectiveness indicator includes at least one of call records, effective operation times, and response delay.

[0062] In a preferred embodiment of the present invention, the dynamic performance indicators include: video clarity, frame rate stability, image delay, network packet loss rate, color reproduction accuracy, and night vision imaging quality.

[0063] Specifically, in this embodiment, the collected basic characteristic data, operational effectiveness indicators, and dynamic performance indicators each play an important role in the video device quality score, and their specific roles are as follows:

[0064] 1. The role of basic characteristic data:

[0065] Capture content: including resolution, encoding format, frame rate, bit rate, delay, etc.

[0066] Purpose: This data is the basis for evaluating the basic performance of video equipment. By normalizing different device parameters to a unified evaluation benchmark, we can eliminate scoring biases caused by underlying device differences (such as different compression algorithms and sensor differences), ensure the comparability of subsequent scores, and make the scoring results more fair and objective. For example, different brands and models of devices may use different encoding formats (such as H.264, H.265, MPEG-4, etc.). Through normalization, these devices with different encoding formats can be compared under the same evaluation benchmark.

[0067] 2. The role of operational effectiveness indicators:

[0068] Collection content: including call records, effective operation times, response delay, etc.

[0069] Purpose: These metrics reflect the device's actual usage and performance during operation. By collecting these metrics, we can understand the device's stability and reliability in real-world use, providing data on the device's actual operating conditions for the device quality score, making the score more accurate to the device's actual operational status. For example, if a device has frequent call records, a high number of valid runs, and low response latency, it indicates good performance and high stability in actual operation.

[0070] 3. The role of dynamic performance indicators:

[0071] Collection content: including video clarity, frame rate stability, image delay, network packet loss rate, color reproduction accuracy, night vision imaging quality, etc.

[0072] Function: These indicators are used to dynamically perceive the performance of the device in the actual operating environment and promptly detect performance degradation or anomalies. By collecting these indicators in real time, the actual performance of the device in different scenarios can be comprehensively evaluated. Combining basic characteristic data and operational effectiveness indicators, a multi-dimensional and comprehensive evaluation of the device quality can be conducted, making the scoring results more accurate and reliable. At the same time, these dynamic performance indicators also provide real-time data support for the intelligent optimization learning module, enabling the scoring model to be dynamically adjusted and optimized based on actual operating data, improving the accuracy and adaptability of the scoring. For example, in night scenarios, night vision imaging quality becomes a key indicator. By collecting this indicator in real time, the monitoring capabilities of the device in night environments can be evaluated.

[0073] In a preferred embodiment of the present invention, the training process of the weight model includes:

[0074] The performance benchmark scores of each video device in the device group are collected manually and added to the manual initial score set. The standardized performance parameter set and the manual initial score set collected from each video device in the device group are input into the initial machine learning model for model training. After completing multiple trainings, a weighted model is obtained.

[0075] Specifically, the following is an embodiment for introducing the training process of the weight model in a preferred embodiment of the present invention:

[0076] Consider developing a city security surveillance system that deploys a variety of video devices of various brands and models for diverse scenarios, such as public safety and intelligent transportation. To ensure the stable and efficient operation of these devices, we need to accurately assess their quality. To this end, we propose a method for intelligently scoring the quality of heterogeneous video devices based on multiple sources, focusing on the training process of the weighting model.

[0077] The training process embodiment includes:

[0078] The first step is data collection:

[0079] Basic performance parameter collection: We first access video equipment from multiple brands and models and collect basic characteristic data for each device, including resolution, encoding format, frame rate, bit rate, latency, etc. This data is the basis for evaluating the basic performance of the device.

[0080] Device usage history data collection: We also record the device's call history, effective operation times, response latency, and other operational effectiveness indicators during actual operation. These indicators reflect the device's actual usage and performance.

[0081] Initial manual scoring sample collection: In the early stages of the system, we invited operations personnel to perform performance benchmark scores on some devices. These scores, based on the expertise and experience of the operations personnel, formed a training sample library, providing training samples for subsequent intelligent optimization modules.

[0082] Dynamic performance indicator collection: During device operation, we collect real-time dynamic performance indicators such as video clarity, frame rate stability, image latency, network packet loss rate, color reproduction accuracy, and night vision imaging quality. These indicators are used to dynamically perceive the performance of the device in the actual operating environment.

[0083] The second step is data standardization:

[0084] Collected basic performance parameters, equipment usage history data, initial manual scoring samples, and dynamic performance indicators are unified into performance standardization processing. A normalization algorithm (such as Min-Max Normalization) is used to process data of different dimensions and standards into a unified and comparable range, forming a standardized performance parameter set.

[0085] The third step is machine learning model training:

[0086] Select a machine learning algorithm: We chose a decision tree algorithm as the initial machine learning model. The decision tree algorithm can learn the relationship between various indicators and equipment quality from historical data, thereby dynamically optimizing the weight of each indicator.

[0087] Model training: A set of standardized performance parameters and a set of manual initial scores are fed into a decision tree algorithm for model training. Through multiple iterations of training, the model gradually learns the complex relationship between various indicators and equipment quality, and dynamically adjusts the weight of each indicator.

[0088] Model optimization: As data accumulates, we can regularly retrain the model to adapt to changes in device performance and usage environments. Furthermore, we can use online learning to dynamically update the model to reflect the latest device status in real time.

[0089] Step 4: Calculate the comprehensive score:

[0090] The optimized scoring model performs a weighted synthesis of the standardized performance parameter set to form a preliminary score, which reflects the comprehensive performance of the device in various indicators.

[0091] Step 5: Scoring, review and output:

[0092] Scoring review: The preliminary score is checked for anomalies. If an anomaly is detected (e.g., an abnormally high packet loss rate), the score is corrected or flagged for manual review to produce the final corrected score.

[0093] Output scoring results: The final score is pushed to the operation and maintenance management platform, providing operators with an intuitive understanding of the equipment's health status. When the equipment score falls below the set threshold, the system automatically triggers an alert or maintenance recommendation to ensure the stable operation of the security monitoring system.

[0094] The overall process is expressed as follows:

[0095] D: Device Group

[0096] Ps: Device performance parameter set (StaticPerformanceParameters)

[0097] Pr: Equipment running record parameter set (RunningRecordsParameters)

[0098] Sa: Artificial Score

[0099] Pd: Dynamic detection indicator set (DynamicPerformanceData)

[0100] N(P): Normalized Performance Set

[0101] M: ScoringModel (including weights)

[0102] S: Preliminary comprehensive score (Score)

[0103] S′: Final score after review and correction (FinalScore)

[0104] O: Output interface transmission results (Output)

[0105] Process formula description:

[0106] (Ps, Sa, Pr) = (D): indicates that the device performance parameter set Ps, the manual initial score Sa, and the device operation history parameter set Pr are collected from the device group D.

[0107] Pd=(D): indicates that the dynamic detection indicator set Pd is collected from the device group D.

[0108] N(P)=(Ps, Pr, Pd): represents the normalization of the equipment performance parameter set Ps, the equipment operation history parameter set Pr, and the dynamic detection indicator set Pd to obtain the normalized parameter set N(P).

[0109] M=(N(P), Sa): indicates that the scoring model M is trained using the standardized parameter set N(P) and the manual initial score Sa.

[0110] S = (N(P), M): indicates that the standardized parameter set N(P) is scored using the trained scoring model M to obtain a preliminary comprehensive score S.

[0111] S′=(S,N(P),Pd): indicates that the preliminary comprehensive score S is reviewed and revised. The final score S′ after review and revision is obtained by combining S, the standardized parameter set N(P), and the dynamic detection indicator set Pd.

[0112] O=(S′): indicates that the final score S′ after review and correction is transmitted through the output interface O.

[0113] This example demonstrates that the weight model training process is a data-driven, dynamic optimization process. It leverages the device's static performance parameters, dynamic indicator data, and device usage history data, dynamically optimizing the weights of each indicator through machine learning algorithms to accurately assess video device quality.

[0114] In a preferred embodiment of the present invention, the standardization process includes: normalizing the performance parameters and dynamic performance indicators, or performing dimensionality reduction processing on the performance parameters and dynamic performance indicators using principal component analysis to obtain a standardized performance parameter set.

[0115] Specifically, this embodiment provides an example of each of normalization processing and principal component analysis (PCA) to illustrate their application in the standardization processing of the present invention:

[0116] Example 1: Standardization based on normalization processing

[0117] Suppose we have a set of video devices whose performance needs to be evaluated. We collect performance parameters and dynamic performance indicators for these devices, including resolution (e.g., 1920×1080, 1280×720, etc., in pixels), frame rate (e.g., 30fps, 60fps, etc., in frames per second), bitrate (e.g., 2Mbps, 4Mbps, etc., in Mbps), and latency (e.g., 50ms, 100ms, etc., in milliseconds). These indicators have varying dimensions and value ranges, making it difficult to directly derive a comprehensive score.

[0118] To solve this problem, we use normalization to standardize the data. The specific steps are as follows:

[0119] Determine the normalization range: We choose to normalize the data to the interval [0,1].

[0120] Normalize each indicator:

[0121] For resolution, we can convert it to the total number of pixels and then normalize it. For example, the total number of pixels for 1920×1080 is 2073600, and the total number of pixels for 1280×720 is 921600. We can normalize it by dividing the total number of pixels by the maximum total number of pixels (2073600) to get a value in the [0,1] range.

[0122] For the frame rate, we can directly divide the frame rate value by the maximum frame rate value to normalize it.

[0123] For bitrate, we can normalize it by dividing the bitrate value by the maximum bitrate value.

[0124] For delay, since the smaller the delay, the better, we can use the inverse method to process it, that is, 1 / delay, and then normalize it.

[0125] Obtaining the normalized performance parameter set: After normalization, all indicators are converted to the [0,1] interval, forming a normalized performance parameter set.

[0126] Example 2: Standardization based on principal component analysis (PCA)

[0127] Using this set of video equipment as an example, we collected a large number of performance parameters and dynamic performance indicators, including resolution, frame rate, bit rate, latency, color reproduction, night vision effects, etc. Due to the large number of indicators and their potential correlation, directly providing a comprehensive score can be complex.

[0128] To simplify the scoring process, we use principal component analysis (PCA) to reduce the dimensionality of the data.

[0129] The specific steps are as follows:

[0130] Data preprocessing: Standardize the data so that the mean of each indicator is 0 and the variance is 1. This is a prerequisite for PCA analysis.

[0131] Calculate the covariance matrix: Calculate the covariance matrix of the standardized data to understand the correlation between the indicators.

[0132] Calculate eigenvalues ​​and eigenvectors: Perform eigendecomposition on the covariance matrix to obtain the eigenvalues ​​and corresponding eigenvectors.

[0133] Select principal components: Based on the size of the eigenvalues, select the eigenvectors corresponding to the first few largest eigenvalues ​​as principal components. These principal components can explain most of the variance in the data.

[0134] Data dimensionality reduction: Project the original data onto the selected principal components to obtain the reduced-dimensional data. This reduced-dimensional data is the standardized performance parameter set.

[0135] Through PCA dimensionality reduction, we can reduce the original multidimensional data to several principal components, thereby simplifying the scoring process while retaining the main information in the data.

[0136] In a preferred embodiment of the present invention, before outputting the scoring result in step S3, a scoring result review process is also included, including: automatically modifying the score or prompting manual review when the scoring result exceeds a preset score range.

[0137] Specifically, a specific embodiment is used to illustrate how to automatically modify the score when the scoring result exceeds the preset score range in a preferred embodiment of the present invention:

[0138] The system evaluated the quality of one of the video devices, initially scoring 85. However, during review, the system detected an abnormally high packet loss rate, exceeding the preset normal range. This could be a temporary anomaly caused by network fluctuations, but it could also indicate a potential performance issue with the device.

[0139] The automatic score modification process includes:

[0140] The first step is anomaly detection:

[0141] The scoring result review module performs an anomaly detection on the preliminary score (S) and finds that the packet loss rate indicator of the device exceeds the preset normal range.

[0142] The second step is to trigger the automatic score modification mechanism:

[0143] Since packet loss rate is one of the key indicators for evaluating video device quality, abnormalities can directly affect the device's overall score. Therefore, when the packet loss rate is detected to be outside the preset range, the system automatically triggers the automatic score adjustment mechanism.

[0144] The third step is to determine the modification rules:

[0145] According to the preset modification rules, when the packet loss rate indicator exceeds the normal range, the score of this indicator needs to be lowered and the overall score adjusted accordingly. The specific rules are as follows:

[0146] If the packet loss rate indicator is within a normal range (eg, 0% to 5%), the original score is kept unchanged.

[0147] If the packet loss rate indicator exceeds the normal range (for example, greater than 5%), the score of the packet loss rate indicator will be reduced by 2 points for every 1 percentage point exceeding the normal range, and the overall score will be adjusted accordingly (the reduction is determined by the weight of the packet loss rate indicator in the overall score).

[0148] Step 4: Automatically modify the score:

[0149] Assume that the packet loss rate of the device is 10%, which is 5 percentage points higher than the normal range. According to the modified rules, the packet loss rate score is reduced from 20 points to 10 points (a reduction of 10 points because the score exceeds the normal range by 5 percentage points, and the score is reduced by 2 points for every 1 percentage point higher than the normal range).

[0150] Assuming the packet loss rate metric weighs 20% in the overall score, the overall score needs to be adjusted down by 2 points (10 points * 20%). Therefore, the preliminary score is adjusted from 85 to 75 points.

[0151] Step 5: Output the final corrected score:

[0152] The scoring result review module pushes the final revised score (S'=75 points) to the operation and maintenance management platform so that the operation and maintenance personnel can intuitively understand the health status of the equipment.

[0153] Through this example, we can see that when the scoring result exceeds the preset score range, the scoring method proposed in this invention can automatically trigger a score modification mechanism, adjusting the score according to preset modification rules. This automatic score modification mechanism can effectively avoid score distortion caused by factors such as temporary network fluctuations and interference from abnormal indicators, thereby improving the accuracy and robustness of the scoring system. Furthermore, by promptly outputting the final, corrected score, this invention can help operation and maintenance personnel promptly identify and address abnormal devices, ensuring the stable operation of the security monitoring system.

[0154] In a preferred embodiment of the present invention, different scenarios are configured, each scenario is associated with a different weight ratio, and the weight model adjusts the scoring result of the comprehensive score according to the weight ratio corresponding to the scenario configuration.

[0155] More specifically, existing systems mostly use fixed rule engines without introducing machine learning or statistical learning mechanisms. Scoring models lack adaptive learning capabilities and are unable to dynamically adjust scoring weights based on environmental changes (such as seasonal lighting differences) or historical data. For example, the weight of color reproduction in night scenes should be higher than that in daytime scenes, but traditional models distort scoring due to fixed weights. To address the problem of scoring results lacking adaptive learning and optimization capabilities:

[0156] Specifically, in this embodiment, by continuously collecting and analyzing device performance data, a machine learning model can be trained to optimize the scoring weights, so that the scoring results can be dynamically adjusted according to environmental changes (such as day and night changes, seasonal light changes) or historical usage data, thereby more accurately reflecting the performance of the device in actual use.

[0157] More specifically, in this embodiment, different scenarios are configured, each scenario is associated with a different weight ratio, and the weight model adjusts the comprehensive score result according to the weight ratio corresponding to the scenario configuration. This embodiment solves the problems existing in the prior art in the following ways:

[0158] 1. Solve the problems of static weight model scoring method:

[0159] 1. Dynamic weight adjustment: The traditional static weight model uses fixed weights for multi-indicator weighted calculations. The weight parameters are solidified for a long time, and the dynamic performance of the equipment and scene differences are not taken into account. However, this embodiment sets different indicator weights according to the actual application scenarios, which means that the weights are not fixed, but can be dynamically adjusted according to actual operating conditions and historical data. For example, in public safety scenarios, higher weights are assigned to clarity and night vision effects; in intelligent traffic scenarios, higher weights are assigned to frame rate and latency. This dynamic weight adjustment mechanism solves the problem of the lack of dynamic adjustment mechanism in the existing technology, making the scoring results closer to the actual usage experience.

[0160] 2. Multi-dimensional Comprehensive Evaluation: This embodiment incorporates multiple key indicators, including clarity, frame rate stability, latency, packet loss rate, color reproduction, and night vision imaging, and combines them with device characteristics for normalization, enabling more comprehensive and accurate quality scoring of heterogeneous devices across multiple brands and models. This avoids the one-sidedness of traditional scoring methods that rely solely on a single indicator, resolving the problem of existing technologies that rely on a single indicator evaluation and lack multi-dimensional comprehensive analysis.

[0161] 2. Solve the problem of lack of adaptive learning and optimization capabilities in scoring results:

[0162] 1. Adaptive optimization capability: The intelligent optimization and self-learning mechanism introduced in this embodiment can automatically adjust the weight of each indicator or add / delete some indicators based on the correlation between the actual scoring results and the device usage performance to optimize the scoring accuracy. For example, the system can continuously optimize the scoring model based on historical scoring data through machine learning models (such as decision trees, random forests, or neural networks). This adaptive optimization capability enables the scoring system to continuously learn and improve, improves the accuracy and reliability of the scoring, and solves the problem that the scoring model in the existing technology lacks adaptive learning capabilities.

[0163] The following is an example of adjusting weights according to different scenarios:

[0164] Consider a city security monitoring system that deploys video equipment of various brands and models for different scenarios such as public safety and intelligent transportation.

[0165] In public safety scenarios:

[0166] Weighting: Higher weight is assigned to clarity and night vision. Public security surveillance requires high image clarity and nighttime monitoring capabilities, as these two indicators are directly related to the effectiveness and safety of surveillance.

[0167] Scoring results: The system will focus on evaluating the device's performance in terms of clarity and night vision effect based on their weighting, thereby generating a scoring result that better meets the needs of public safety scenarios.

[0168] In the intelligent transportation scenario:

[0169] Weighting: Frame rate and latency are given higher weights. Because intelligent traffic monitoring requires real-time capture of traffic conditions, high frame rate and latency are required to ensure smooth and real-time monitoring images.

[0170] Scoring results: The system will focus on evaluating the device's performance on these two indicators based on the weights of frame rate and latency, thereby generating scoring results that better meet the needs of intelligent traffic scenarios.

[0171] The following example illustrates how to dynamically adjust weights based on the environment:

[0172] Suppose there is a video device deployed at a traffic intersection. The device needs to dynamically adjust the scoring weight according to the lighting conditions at different time periods.

[0173] For daytime hours:

[0174] With sufficient lighting, clarity may not be a major issue, but frame rate and latency are very important for real-time monitoring of traffic conditions.

[0175] Weight adjustment: The system automatically adjusts weights to give higher weights to frame rate and latency to reflect the actual performance of the device in daytime environments.

[0176] For night time:

[0177] As lighting conditions deteriorate, clarity becomes a key indicator, and night vision effects also directly affect monitoring quality.

[0178] Weight adjustment: The system automatically adjusts the weights, assigning higher weights to clarity and night vision effects to reflect the actual performance of the device in a night environment.

[0179] By dynamically adjusting the weights according to the environment, the present invention can ensure that the scoring results are always consistent with the performance of the device in the actual usage scenario, thereby improving the accuracy and reliability of the scoring.

[0180] In response to the problem that the existing video equipment quality assessment method is single and cannot comprehensively reflect multi-dimensional performance indicators such as clarity, frame rate, delay, packet loss rate, and color reproduction, the present invention proposes a comprehensive scoring method based on the fusion of multi-dimensional indicators to achieve a more comprehensive and objective equipment quality evaluation.

[0181] To address the problem that traditional scoring methods are not applicable enough due to differences in basic characteristics such as encoding protocols and sensor types among heterogeneous devices, the present invention designs a device characteristic normalization processing mechanism that can adaptively adjust the scoring criteria according to the underlying parameters of the device, thereby achieving fair and accurate evaluation of multiple heterogeneous devices.

[0182] To address the problems of existing scoring methods' poor real-time and dynamic nature, and their inability to promptly reflect changes in equipment operating status, the present invention constructs a dynamic quality monitoring and instant scoring system based on real-time data streams, which can promptly detect equipment anomalies and provide feedback on scoring changes, thereby improving the overall operation and maintenance efficiency of the system.

[0183] To address the problem that existing scoring systems lack learning and optimization capabilities, this invention introduces an intelligent scoring mechanism based on adaptive optimization of historical data. It continuously adjusts and optimizes the scoring model through machine learning methods, so that the scoring results can adapt to the changing needs of different environments and different time periods.

[0184] In summary, the present invention realizes efficient, accurate and intelligent quality scoring of multiple heterogeneous video devices through technical means such as multi-dimensional comprehensive evaluation, adaptive normalization processing, real-time dynamic monitoring and intelligent optimization scoring, effectively improving the scientificity and timeliness of video equipment management and maintenance in security systems.

[0185] The present invention also provides an intelligent scoring method for the quality of heterogeneous video devices, which uses the above-mentioned intelligent scoring method. Figure 3 Shown, including:

[0186] Data acquisition module 1, used to collect various performance parameters of multiple different video devices in the device group, and to collect various dynamic performance indicators of each video device in real time during the operation of the device group;

[0187] The data processing module 2 is connected to the data acquisition module 1 and is used to standardize the performance parameters and dynamic performance indicators to obtain a standardized performance parameter set;

[0188] The scoring module 3 is connected to the data processing module 2 and is used to input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

[0189] The above are only preferred embodiments of the present invention and do not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the contents of this specification and illustrations should be included in the protection scope of the present invention.

Claims

1. An intelligent scoring method for heterogeneous video device quality, characterized in that: include: Step S1, collecting various performance parameters of a plurality of different video devices in a device group, and collecting various dynamic performance indicators of each of the video devices in real time during the operation of the device group; Step S2, normalizing the performance parameters and the dynamic performance indicators to obtain a standardized performance parameter set; Step S3: Input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

2. The intelligent scoring method according to claim 1, characterized in that: The step S1 comprises: Step S11, collecting basic characteristic data of each of the video devices and adding it to the performance parameters; Step S12, collecting the operational effectiveness index of each of the video devices and adding it to the performance parameters; Step S13: collecting multiple dynamic performance indicators of each of the video devices in real time during the operation of the device group.

3. The intelligent scoring method according to claim 1, characterized in that: The training process of the weight model includes: The manual performance benchmark score of each video device in the device group is collected and added to the manual initial score set, and the standardized performance parameter set and the manual initial score set collected from each video device in the device group are input into the initial machine learning model for model training. After completing multiple trainings, the weight model is obtained.

4. The intelligent scoring method according to claim 2, characterized in that: The basic characteristic data includes at least one of resolution, encoding format, frame rate, bit rate, and delay.

5. The intelligent scoring method according to claim 2, characterized in that: The operation effectiveness indicator includes at least one of call records, effective operation times, and response delay.

6. The intelligent scoring method according to claim 2, characterized in that: The dynamic performance indicators include: video clarity, frame rate stability, image delay, network packet loss rate, color reproduction accuracy, and night vision imaging quality.

7. The intelligent scoring method according to claim 1, characterized in that: The standardization processing includes: normalizing the performance parameters and the dynamic performance indicators, or performing dimensionality reduction processing on the performance parameters and the dynamic performance indicators using principal component analysis to obtain the standardized performance parameter set.

8. The intelligent scoring method according to claim 1, characterized in that: Before outputting the scoring result in step S3, a scoring result review process is also included, including: automatically modifying the score or prompting manual review when the scoring result exceeds a preset positive score range.

9. The intelligent scoring method according to claim 1, characterized in that: Different scenarios are also configured, and each scenario is associated with a different weight ratio. The weight model adjusts the scoring result of the comprehensive score according to the weight ratio corresponding to the scenario configuration.

10. An intelligent scoring method for heterogeneous video device quality, characterized in that: Applying the intelligent scoring method according to any one of claims 1 to 9, comprising: A data acquisition module is used to acquire various performance parameters of a plurality of different video devices in the device group, and to acquire various dynamic performance indicators of each of the video devices in real time during the operation of the device group; a data processing module, connected to the data acquisition module, for performing standardization processing on each of the performance parameters and the dynamic performance indicators to obtain a standardized performance parameter set; The scoring module is connected to the data processing module and is used to input the standardized performance parameter set into the pre-trained weight model for comprehensive scoring to obtain a scoring result.

Images