Dynamic chart adaptive sampling and denoising display algorithm based on time series data

Abstract

The application discloses a dynamic chart adaptive sampling and noise reduction display algorithm based on time series data, comprising the following steps: obtaining time series data to be displayed and dividing local time series segments according to display range, pixel width and zoom level; calculating segment display necessity indexes and determining segment types; generating display necessary sampling point sets according to segment types and executing hierarchical noise reduction; realizing cross-zoom consistent display in combination with historical forced retention point records, and outputting dynamic chart display data through historical segment freezing and front segment updating.

Description

Technical Field

[0001] This invention relates to the fields of data visualization and time-series data processing technology. More specifically, this invention relates to adaptive sampling, anomaly fidelity, hierarchical noise reduction, consistent display across scaling, and real-time refresh control of time-series data in the process of dynamic chart display. In particular, it relates to a dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data. Background Technology

[0002] Existing dynamic charts are typically used in scenarios such as equipment operation monitoring, financial market display, energy consumption trend analysis, medical data observation, and business indicator dashboards. The time-series data in these scenarios are often characterized by high collection frequency, large data volume, and fast refresh rate. However, the display width and number of pixels that can be drawn on the terminal screen are limited. Therefore, the system usually reduces the number of data points that need to be displayed by means of fixed interval sampling, mean sampling, or smoothing and noise reduction, so that the chart can maintain a good refresh rate and smooth display.

[0003] The above processing methods can reduce the rendering burden when data changes are relatively stable. However, in actual business data, abnormal information often does not always manifest as long-term continuous changes, but may manifest as short-term limit exceedances, instantaneous spikes, local peaks and troughs, rapid changes, or trend inflection points. These information have a short duration and occupy few points in the original time series data, but are directly related to equipment failure early warning, market fluctuation judgment, energy consumption anomaly identification, vital sign monitoring, and business risk review. If data points are still extracted at fixed intervals, key limit exceedance points may fall exactly in the unextracted positions. If mean or ordinary smoothing noise reduction is used, instantaneous spikes and real changes may be weakened. If resampling is performed in the global view, local zoomed-in view, and real-time refresh view, different curve shapes may be presented in the same time period, leading to inconsistent judgments for users under different display states.

[0004] Therefore, the main shortcoming of existing technology is not just that dynamic charts are under great pressure to display high-frequency data, but that in the process of compressing data and reducing noise, it lacks the ability to distinguish the necessary information for display, and easily compresses or smooths out key points that affect trend judgment and anomaly judgment.

[0005] Based on the above, the core problem addressed by this invention can be summarized as follows: existing dynamic chart sampling and noise reduction display methods based on time-series data are difficult to balance data compression, short-term anomaly fidelity, consistent display across scaling, and historical curve stability under limited display space and real-time refresh conditions, which can easily lead to the hiding of key anomalies and misleading trend judgments. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data. By dividing the time-series data into local segments according to the current display pixel capacity, and performing compressed sampling, trend preservation, forced anomaly fidelity preservation, cross-scaling inheritance, and historical segment freezing respectively according to whether each segment contains key information that affects trend judgment or anomaly judgment, the algorithm solves the problems of existing dynamic chart sampling and noise reduction methods proposed in the background art, which easily hide short-term anomalies, weaken peak and valley amplitudes, change trend judgment, cause inconsistencies in display at different scaling levels, and cause repeated jumps in historical curves under limited display space and real-time refresh conditions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data, comprising the following steps: S1. Obtain the time series data to be displayed, the current chart display time range, the current chart pixel width, the current zoom level, and the real-time refresh timestamp. Extract the target time series data segment from the time series data to be displayed according to the current chart display time range, and divide the target time series data segment into multiple local time series segments according to the time span corresponding to the pixel width. S2. Calculate the segment display necessity index for each local time series segment. The segment display necessity index includes at least the maximum value, minimum value, beginning and end trend change, local slope change, limit-breaking state, peak and valley state, sudden change state and short-term spike state within the segment. Based on the segment display necessity index, the corresponding local time series segment is determined as a stationary compressible segment, a trend-preserving segment or an abnormal forced fidelity segment. S3. Generate a set of sampling retention points based on the determination results of local time series segments. For stationary compressible segments, retain the segment start point, segment end point, and representative point. For trend retention segments, retain the segment start point, segment end point, and trend point with the largest slope change. For abnormal forced fidelity segments, retain the limit-breaking point, peak point, valley point, abrupt change point, or short-term spike point and its adjacent points. This results in a set of necessary sampling points for display corresponding to the current chart display time range. S4. Perform hierarchical noise reduction on the set of necessary sampling points for display. Specifically, representative points in stable and compressible segments are displayed smoothly according to the average trend of the segments. Trend points in trend-preserving segments are restricted from changing their original beginning and end trends after noise reduction. Points exceeding the limit, peak points, valley points, abrupt change points, or short-term spikes in abnormal forced fidelity segments are prohibited from amplitude reduction and their original timestamps and original amplitudes are preserved. This results in the set of display points after noise reduction. S5. Based on the current zoom level, read the historical forced retention point record that overlaps with the target time series data segment, and incorporate the over-limit points, peak points, valley points, abrupt change points and short-term spike points already identified in the global view, local zoom view or playback view into the noise-reduced display point set; when a new abnormal point not existing in the historical forced retention point record is identified in the current zoom level, write the abnormal point and its adjacent points before and after it into the historical forced retention point record to maintain the display consistency of the same abnormal position under different zoom levels; S6. Based on the real-time refresh timestamp, the target time series data segment is divided into a stable historical segment and an updatable leading edge segment. The stable historical segment is a time period that exceeds the current refresh time window and whose data integrity meets the preset integrity conditions. The generated set of display points and display format are locked for the stable historical segment, and only anomaly markers are allowed to be added. For the updatable leading edge segment, the adaptive sampling and noise reduction processing from S2 to S5 is performed again to obtain the dynamic chart display data corresponding to this refresh. S7. Output the dynamic chart display data to the dynamic chart rendering end in chronological order, and synchronously output the display source mark of each display point. The display source mark includes the stable compression point, trend retention point, anomaly forced fidelity point, cross-scale inheritance point and history freeze point, so that the dynamic chart can compress stable data, retain short-term anomalies, maintain cross-scale consistency and stably display historical curves within the limited display space.

[0008] By adopting the above technical solution, the present invention does not reduce the number of display points by sampling at fixed intervals or uniform smoothing and noise reduction, but determines the sampling and noise reduction method according to the degree of influence of local time segments on the user's judgment, so that stable segments are compressed, trend segments maintain their direction, and abnormal segments retain their original amplitude and occurrence time, thereby reducing the interference of the chart compression process on the judgment of anomalies and trends.

[0009] In a preferred embodiment, in S1, the total display duration is obtained by subtracting the start and end times of the current chart display time range. The total display duration is then matched with the current chart pixel width to obtain the basic time span corresponding to a single pixel column, and corrected to the segment division time span according to the current scaling level. Subsequently, the target time series data segments are continuously divided according to the segment division time span. Time series points falling within the same segmented time range are grouped into the same local time series segment, and each local time series segment is bound to the segment start and end times, the pixel column range corresponding to the segment, and the original time series point list within the segment.

[0010] By adopting the above technical solution, the division of local time segments corresponds to the actual display pixel capability, which can avoid over-dense drawing in the global view or excessive compression in the local magnified view, so that the subsequent sampling density can be adjusted according to the display accuracy.

[0011] In a preferred embodiment, in S2, the segment amplitude index group is obtained by calculating the segment start value, segment end value, maximum value within the segment, minimum value within the segment, maximum value timestamp, and minimum value timestamp; the segment trend index group is obtained by calculating the beginning and end trend changes between the segment start value and the segment end value, as well as the direction and magnitude of numerical changes between adjacent time points within the segment; the over-limit state is obtained by comparing the time point values ​​within the segment with preset upper limit thresholds and preset lower limit thresholds.

[0012] By adopting the above technical solution, the necessity index for segment display can be refined from a general fluctuation judgment into three calculable dimensions: amplitude, trend, and limit exceedance, providing a clear data basis for subsequent segment type determination.

[0013] In a preferred embodiment, in S2, the peak state is obtained by comparing the maximum value within a segment with the boundary values ​​of the adjacent preceding and following segments; the valley state is obtained by comparing the minimum value within a segment with the boundary values ​​of the adjacent preceding and following segments; the abrupt change state is obtained by whether the magnitude of the numerical change between adjacent time points exceeds the abrupt change determination threshold; and the short-term spike state is obtained by judging the magnitude deviation, duration, and recovery time of the candidate spike point relative to the reference values ​​of the preceding and following neighborhoods.

[0014] By adopting the above technical solutions, peaks, valleys, abrupt changes, and spikes that are short in duration but important for risk assessment can be identified, avoiding the weakening or deletion of such information by ordinary mean sampling or smoothing and noise reduction.

[0015] In a preferred embodiment, in S2, when a local time series segment does not have a limit-breaking state, peak-valley state, abrupt change state, or short-term spike state, and the trend changes at the beginning and end and the local slope changes are all lower than the corresponding stability thresholds, it is determined to be a stationary compressible segment; when a local time series segment does not have an abnormal state but the trend changes at the beginning and end or the local slope changes reach the trend threshold, it is determined to be a trend-preserving segment; when a local time series segment has at least one of a limit-breaking state, peak-valley state, abrupt change state, or short-term spike state, it is determined to be an abnormal forced fidelity segment.

[0016] By adopting the above technical solution, a clear segment classification boundary can be established, allowing different segments to enter different sampling and noise reduction paths, thus avoiding the use of the same sampling or smoothing rule for all segments.

[0017] In a preferred embodiment, in S3, for a stationary compressible segment, the original time series point closest to the segment's average value and whose timestamp is located in the middle region of the segment is selected as a representative point, and together with the segment's start and end points, forms a stationary compression point group; for a trend-preserving segment, the original time series point with the largest slope change value within the segment is selected as a trend point, and together with the segment's start and end points, forms a trend-preserving point group; for anomaly forced fidelity segments, the limit-crossing points, peak points, valley points, abrupt change points, or short-term spike points, along with their preceding and following original time series points, form an anomaly forced fidelity point group.

[0018] By adopting the above technical solution, the retention of sampling points no longer depends on fixed intervals, but retains data points that can represent stable states, trend changes, or anomaly formation processes according to display necessity, thereby retaining key business information while reducing the number of drawing points.

[0019] In a preferred embodiment, in S4, the smoothed display value is calculated for the representative point of the stable compressible segment according to the average trend of the original time sequence points within the segment and the segment start and end point constraints; for the trend point of the trend-preserving segment, the start and end trend directions before and after noise reduction are compared. When the trend direction after noise reduction is inconsistent with the original start and end trend directions, the noise reduction amplitude of the trend point is canceled and restored to the original amplitude or trend constraint amplitude; for the abnormal points of the abnormal forced fidelity segment, the smoothing process of reducing peak values, raising valley values, or weakening the over-limit amplitude is prohibited.

[0020] By adopting the above technical solution, the noise reduction rules are matched with the segment type, which can reduce visual jitter in stable segments, without changing the direction judgment of trend segments, and without smoothing out the key risk points in abnormal segments.

[0021] In a preferred embodiment, in S5, the system queries the historical anomaly record in the historical mandatory retention point record that overlaps with the target time series data segment based on the current zoom level, the current chart display time range, and the start and end times of the target time series data segment. When the historical anomaly is not included in the denoised display point set, the historical anomaly and its adjacent points before and after it are incorporated into the denoised display point set. When a new anomaly is identified in the current zoom level, the original timestamp, original amplitude, anomaly type, current zoom level, source local time series segment, and adjacent points before and after it are written into the historical mandatory retention point record.

[0022] By adopting the above technical solution, the global view, the local zoom-in view, and the playback view can share the same anomaly record, avoiding the problem of the same anomaly shifting in position, changing in magnitude, or disappearing completely under different zoom states.

[0023] In a preferred embodiment, in S6, the historical freeze boundary time is obtained by backtracking the stable time window based on the real-time refresh timestamp. The portion of the target time series data segment that is earlier than or equal to the historical freeze boundary time and does not contain missing time intervals, out-of-order data points, or unconfirmed delayed data points is determined as a stable historical segment. The portion that is later than the historical freeze boundary time or whose data integrity does not meet the preset integrity condition is determined as an updatable leading edge segment. The set of display points, the order of display points, the display amplitude, and the curve connection shape are locked for the stable historical segments. Adaptive sampling and noise reduction processing are continued for the updatable leading edge segments.

[0024] By adopting the above technical solution, we can avoid the dynamic chart from repeatedly recalculating the already stable historical curves when refreshing in real time, thus keeping the shape of the historical segment stable, while retaining the real-time response capability of the leading segment to new data.

[0025] In a preferred embodiment, in S6, when a late abnormal data point is received within a stable historical segment during a subsequent refresh process, the stable historical segment is not resampled or re-denoised. Instead, an abnormal marker is added at the corresponding time position, and the late abnormal data point is written into the historical forced retention point record. Subsequently, in the playback view or when the time range is reloaded next time, the late abnormal data point is displayed as a forced retention point.

[0026] By adopting the above technical solution, we can balance the stability of historical curves and the visibility of late arrival anomalies, avoid the overall fluctuation of historical curves caused by late arrival data, and also avoid the complete ignoring of late arrival anomalies.

[0027] In a preferred embodiment, in S7, the display time, display amplitude, original timestamp source, and display source marker of each display point are written into the rendering point sequence, and the corresponding connection instruction is output to the dynamic chart rendering end according to the display source marker; wherein, a smooth connection instruction is output for stable compression points, a trend connection instruction is output for trend preservation points, an original amplitude display instruction is output for abnormal forced fidelity points, a forced display instruction is output for cross-scaling inheritance points, and a resampling prohibition instruction is output for historical frozen points.

[0028] By adopting the above technical solution, the algorithm calculation results can be directly converted into display rules that can be executed on the dynamic chart rendering end, so that sampling, noise reduction, anomaly fidelity, cross-scaling inheritance and history freezing are not just background judgment results, but can form a consistent visual output on the chart display end.

[0029] In a preferred embodiment, the historical mandatory retention point record is updated after each cross-scaling consistency process is completed; when a newly identified outlier does not exist in the historical mandatory retention point record, it is written into the record; when a newly identified outlier already exists in the historical mandatory retention point record and belongs to the same time neighborhood, the current scaling level and the source local time segment are added to the existing record; when a higher scaling level confirms a more accurate original timestamp or original amplitude, the historical mandatory retention point record is updated with the original timestamp and original amplitude confirmed by the higher scaling level.

[0030] By adopting the above technical solution, outlier records can be gradually corrected and improved during zooming operations, and used as the basis for mandatory retention in subsequent global views, local zoomed-in views, playback views, and real-time refresh views, forming a cross-view... Figure 1 The resulting abnormality is displayed in a closed loop.

[0031] In summary, this invention, through pixel matching and segmentation, calculation of display necessity index, segment type determination, type-based sampling and retention, hierarchical noise reduction, forced inheritance across scaling, freezing of historical segments, and output of rendering source markers, enables dynamic charts to compress stable data within a limited display space, retain the original amplitude and time at the location of anomalies, maintain consistent display of the same anomaly point under different scaling levels, and keep historical curves stable during real-time refresh. This solves the problem that existing dynamic chart sampling and noise reduction methods easily lead to the hiding of key anomalies and misleading trend judgments.

[0032] The present invention has at least the following technical effects and advantages: 1. By dividing local time segments according to the current display range, pixel width and scaling level, and deciding whether to compress, retain or force fidelity based on the necessity of segment display, the present invention can retain short-term over-limit, peaks and valleys, abrupt changes and spikes under limited display space and real-time refresh conditions, and avoid key anomalies being hidden by the sampling or noise reduction process.

[0033] 2. By calculating the maximum value, minimum value, beginning and end trend changes, local slope changes, limit-breaking state, peak and valley state, sudden change state and short-term peak state, this invention changes the sampling basis of dynamic charts from fixed intervals to display necessity judgment, making the sampling results more in line with the needs of trend recognition and anomaly recognition.

[0034] 3. By retaining the start point, end point, and representative point for stable compressible segments, retaining the trend point for trend-preserving segments, and retaining the abnormal point and its adjacent points for abnormal forced fidelity segments, this invention can reduce invalid display points while preserving the change process before and after the formation of anomalies, thereby improving the interpretability of the compressed chart.

[0035] 4. By performing graded noise reduction on display points from different sources, the present invention can suppress random jitter in stable segments, maintain the original trend direction in trending segments, and prohibit the weakening of the original amplitude in abnormal segments, thereby avoiding ordinary smoothing processing from misjudging the true peak, valley or out-of-limit amplitude as noise.

[0036] 5. By establishing historical mandatory retention point records and inheriting over-limit points, peak-valley points, abrupt change points, and short-term spike points between the global view, local zoom-in view, and playback view, this invention can maintain the consistency of the position and amplitude of the same anomaly under different zoom levels, reducing judgment bias during cross-view analysis.

[0037] 6. By dividing the target time series data segment into a stable historical segment and an updatable leading edge segment, and locking the display point set and curve shape of the stable historical segment, this invention can prevent the historical curve from jumping repeatedly during real-time refresh, while retaining the real-time update capability of the leading edge data, thereby improving the stability of dynamic chart review and monitoring judgment. Attached Figure Description

[0038] Figure 1 This is a flowchart of the dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to the present invention. Detailed Implementation

[0039] The following embodiments provide a clear and complete description of the dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data provided by the present invention. It should be understood that the following embodiments are used to illustrate the technical solution of the present invention and are not intended to limit the scope of protection of the present invention; equivalent substitutions made to data sources, chart types, refresh rates, anomaly thresholds, display terminals, and rendering methods without departing from the technical concept of the present invention should all fall within the scope of protection of the present invention.

[0040] This embodiment uses a dynamic chart of equipment monitoring as an example for illustration. The time-series data to be displayed can be a data sequence formed by collecting equipment current, temperature, pressure, vibration, energy consumption, transaction price, heart rate, blood oxygen, business indicators, etc. in chronological order. The dynamic chart can be a line chart, area chart, trend chart, real-time curve chart, or replay curve chart. The core of this algorithm is not to simply reduce the number of display points, nor to uniformly smooth the entire curve, but to determine whether each segment of time-series data contains trend changes or abnormal changes that may affect the user's judgment within the current display range and current pixel width limit, and to decide whether the data segment should be compressed, retained, or forcibly displayed with fidelity based on this.

[0041] Reference Figure 1 The details are as follows: S1. Obtain the timing data to be displayed and divide it into local timing segments. In this embodiment, the system first acquires the time series data to be displayed, the current chart display time range, the current chart pixel width, the current zoom level, and the real-time refresh timestamp. The time series data to be displayed includes multiple data points arranged by timestamps, and each data point includes at least the original timestamp and the original amplitude. The current chart display time range includes the display start time and the display end time. The current chart pixel width represents the number of pixel columns that can be used to draw curves in the horizontal direction of the dynamic chart. The current zoom level is used to indicate whether the current view is a global view, a local zoom-in view, or a playback view. The real-time refresh timestamp is used to distinguish between historical data segments and the latest leading-edge data segments.

[0042] In practice, the system first extracts data points from the time series data to be displayed, based on the current time range of the chart display, between the start and end times of the display, to form a target time series data segment. Then, the system sorts the data points within the target time series data segment in ascending order of timestamps and removes data points with missing timestamps, duplicate timestamps, or identical values. For data points with duplicate timestamps but different values, the system prioritizes retaining data points with higher reliability of the data source or those collected at a later time, in order to form a valid target time series data segment.

[0043] Furthermore, the system matches the total display duration of the current chart's time range with the current chart's pixel width, calculating the base time span corresponding to a single pixel column. If the current zoom level is high, indicating the user is viewing local details, the segment's time span is smaller than or close to the base time span. Conversely, if the current zoom level is low, indicating the user is viewing a global trend over a longer time range, the segment's time span is greater than the base time span, reducing meaningless high-density rendering. The system uses the display start time as the segment starting point, continuously dividing the effective target time-series data into segments according to the segment's time span, and grouping data points falling within the same segment's time range into the same local time-series segment.

[0044] For each local time series segment, the system counts the number of data points within the segment, the segment's coverage duration, and the corresponding pixel column range. When no original data points exist within a segment, the system uses the endpoints of adjacent valid segments as display boundary occupancy points to ensure the continuity of the curve within that time range. When the number of data points within a segment exceeds the capacity for clear display by a single pixel or a small number of pixels, the system marks that local time series segment as a segment to be compressed and evaluated. Finally, the system binds each local time series segment with its start and end times, the corresponding pixel column range, the list of original time series points within the segment, and the compression and evaluation marker to generate a set of local time series segments, which serves as the basis for subsequent calculations of the segment display necessity index.

[0045] Through the above processing, S1 transforms the original high-frequency timing data into local timing segments that match the display pixel capabilities, avoiding over-drawing in the global view and also avoiding loss of details in the local view due to excessively coarse slices.

[0046] S2. Calculate the necessity index for displaying segments and determine the segment type. After completing the division of local time series segments, the system calculates the segment display necessity index for each local time series segment. The segment display necessity index is used to determine whether the segment will affect the user's judgment of trends, anomalies, limits, or peaks and troughs in the current chart.

[0047] Specifically, the system first reads the list of original time points within each local time segment and calculates the segment start value, segment end value, maximum value within the segment, minimum value within the segment, maximum value timestamp, and minimum value timestamp to form a segment amplitude index group. The segment start value and segment end value are used to reflect the overall change between the beginning and end of the segment, while the maximum value and minimum value within the segment are used to identify local peaks and valleys and amplitude ranges.

[0048] Subsequently, the system calculates the trend changes at the beginning and end of the segment based on the segment's starting and ending values, and calculates the local slope changes based on the direction and magnitude of numerical changes between adjacent time points within the segment, forming a segment trend index group. If most adjacent data points within a segment change in the same direction, it indicates that the segment may have a trend change; if the direction of change within the segment is frequent and repetitive with small amplitude, it is more likely to be random jitter or noise.

[0049] For cases of exceeding limits, the system compares the value of each time point within a segment with preset upper and lower threshold values. When any time point exceeds the preset upper threshold or falls below the preset lower threshold, the local time segment is marked as exceeding limits, and the original timestamp, original amplitude, and exceedance magnitude of the exceeding point are recorded. The preset upper and lower threshold values ​​can be determined based on device safety limits, service thresholds, monitoring indicator standard ranges, or user configuration.

[0050] For peak-valley conditions, the system compares the maximum value within a segment with the corresponding boundary values ​​of the preceding and following segments. When the maximum value is higher than the neighboring segments and the difference exceeds the peak value threshold, the time point corresponding to the maximum value is marked as a peak point. Similarly, the system compares the minimum value within a segment with the corresponding boundary values ​​of the preceding and following segments. When the minimum value is lower than the neighboring segments and the difference exceeds the valley value threshold, the time point corresponding to the minimum value is marked as a valley value. This allows the system to avoid skipping local peaks or valleys during fixed-interval sampling.

[0051] For abrupt changes, the system calculates the magnitude of numerical change per unit time between adjacent time points. When the magnitude of change per unit time for a pair of adjacent points exceeds the abrupt change threshold, the corresponding time point is marked as an abrupt change point, and the local time segment is marked as having an abrupt change state. This judgment is used to identify rapid increases, rapid decreases, step changes, or instantaneous jumps.

[0052] For short-term spikes, the system uses each original time-series point within a segment as a candidate spike, establishing a forward neighborhood before and a backward neighborhood after each candidate spike. The duration of both the forward and backward neighborhoods does not exceed a preset short-term window. The system calculates the amplitude deviation of the candidate spike relative to the reference values ​​of the forward and backward neighborhoods. When the amplitude deviation exceeds a short-term spike amplitude threshold, and the time-series value after the candidate spike returns to the neighborhood reference range within a preset recovery time, while the duration of the candidate spike is shorter than a preset short-term duration threshold, it is identified as a short-term spike. This rule is used to retain transient anomalies that are short in duration but significant for business operations.

[0053] After calculating the above indicators, the system classifies local time series segments into three categories based on the necessity index for segment display. When a local time series segment does not have any out-of-limit, peak-valley, abrupt change, or short-term spike states, and the trend changes at the beginning and end and the local slope changes are all below the corresponding stability thresholds, the local time series segment is determined to be a stationary compressible segment. When a local time series segment does not have any abnormal states, but the trend changes at the beginning and end or the local slope changes reach the trend threshold, it is determined to be a trend-preserving segment. When a local time series segment has at least one of the out-of-limit, peak-valley, abrupt change, or short-term spike states, it is determined to be an abnormal forced fidelity segment.

[0054] Through the above processing, S2 further distinguishes local time segments from ordinary data segments into three types: compressible, trend-preserving, and display-fidelity-required, providing a clear basis for subsequent sampling and noise reduction.

[0055] S3. Generate a set of necessary sampling points for display based on the segment type. After obtaining the determination result of the local time segment, the system generates a set of sampling retention points according to different segment types.

[0056] For a stationary compressible segment, the system reads the segment's start point, end point, and average value. It then selects the original time series point closest to the average value and with its timestamp located in the middle region of the segment as a representative point. This representative point, along with the segment's start and end points, forms a group of stationary compressible points. This representative point is used to maintain the overall horizontal position of the segment, while the segment's start and end points are used to maintain the continuity of the curve in the time direction.

[0057] For trend-preserving segments, the system reads the slope change values ​​between adjacent time series points within the segment and selects the original time series point with the largest slope change value as the trend point. This trend point, along with the segment's start and end points, forms a trend-preserving point group. Trend points are used to express the location within the segment that most significantly influences the curve's bending direction or trend judgment, allowing users to still see the trend changes after zooming or compressing the display.

[0058] For anomaly-forced fidelity segments, the system reads the limit-crossing points, peak points, valley points, abrupt change points, and short-term spike points within the segment, and synchronously retains the preceding and following original time series points of each anomaly point as adjacent points, forming anomaly-forced fidelity point group. The purpose of retaining the preceding and following adjacent points of anomalies is to demonstrate the changes before the anomaly occurs, when the anomaly reaches its peak or exceeds the limit, and after the anomaly recovers, rather than simply displaying an isolated anomaly point.

[0059] When the same original time series point belongs to both a trend point and an outlier point, the system treats it as an outlier point with forced fidelity and retains the original timestamp, original amplitude, and outlier type marker of the original time series point. When two adjacent local time series segments retain the same boundary time series point, the system retains only one boundary time series point and binds it to both the end point of the previous segment and the start point of the next segment to avoid the dynamic chart rendering end from repeatedly drawing the same time position.

[0060] Finally, the system merges the stable compression point group, the trend preservation point group, and the abnormal forced fidelity point group in ascending order of timestamps, and removes duplicate time series points to obtain the set of necessary sampling points for display corresponding to the current chart display time range.

[0061] Through the above processing, S3 achieves sampling based on display necessity. It reduces the number of drawing points in stable segments, retains key inflection points in trend segments, and preserves original anomaly points and their formation process in abnormal segments, thereby avoiding the loss of key points caused by fixed-interval sampling.

[0062] S4. Perform hierarchical noise reduction processing on the set of necessary sampling points for display. After obtaining the set of necessary sampling points for display, the system performs hierarchical noise reduction processing on this set. Unlike ordinary uniform smoothing, this embodiment sets different noise reduction rules according to the segment type to which the display points belong.

[0063] First, the system reads the source segment type and original amplitude of each display point in the set of necessary sampling points, and establishes noise reduction rules for stationary compressible segments, trend-preserving segments, and abnormal forced fidelity segments respectively.

[0064] For representative points derived from stationary compressible segments, the system calculates a smoothed display value based on the average trend of the original time series points within the segment and the segment's start and end point constraints, and uses this smoothed display value as the display amplitude of the representative point. This process reduces random jitter in stationary segments, making the curve visually more stable.

[0065] For trend points derived from trend-preserving segments, the system compares the original beginning and end trend directions of the segment with the denoised beginning and end trend directions. When the denoised trend direction is inconsistent with the original beginning and end trend direction, the system cancels the denoising amplitude of the trend point and restores it to the original amplitude or trend constraint amplitude, thereby ensuring that the denoising process does not display an upward trend as a downward trend, nor a downward trend as a stable trend.

[0066] For out-of-limit points, peak points, valley points, abrupt change points, or short-term spikes originating from abnormal forced fidelity segments, the system prohibits performing smoothing processes such as reducing peak values, increasing valley values, or weakening out-of-limit amplitudes on these out-of-limit points, and directly preserves the original timestamp and original amplitude of the out-of-limit point. For adjacent points before and after the out-of-limit point, the system preserves the temporal order and amplitude change direction of the adjacent points relative to the out-of-limit point, and only allows mild noise suppression within the range that does not change the anomaly formation process.

[0067] After completing the graded noise reduction, the system rearranges the stable compression points, trend preservation points, abnormal forced fidelity points and their adjacent points in chronological order to form a set of noise-reduced display points.

[0068] Through the above processing, S4 makes noise reduction no longer solely aimed at smoothing the curve, but rather reduces jitter in the stable range, maintains direction in the trend range, and preserves the true amplitude and occurrence time in the abnormal range.

[0069] S5. Read and update historical forced holdpoint records to maintain consistency across scaling. After obtaining the set of display points after noise reduction, the system reads the historical forced retention point records that overlap with the target time series data segment according to the current zoom level. The historical forced retention point records are used to save the over-limit points, peak points, valley points, abrupt change points, and short-term spike points that have been identified in the global view, local zoom-in view, or playback view.

[0070] In practice, the system queries historical outlier records that overlap with the target time series data segment from the historical mandatory retention point records, based on the current zoom level, the current chart display time range, and the start and end times of the target time series data segment. The system reads outlier points, peak points, valley points, abrupt change points, and short-term spikes from the historical outlier records and matches them with the timestamps of the denoised display point set. When a historical outlier is not included in the denoised display point set, the system incorporates that historical outlier and its preceding and following adjacent points into the denoised display point set.

[0071] When the current zoom level is a partial zoom view, the system sets historical outliers identified in the global view as mandatory retention points in the partial zoom view, prioritizing the preservation of their original amplitude and original occurrence time to prevent outlier position shifts due to resampling. When the current zoom level is a global view or a playback view, if a new outlier is identified in the partial zoom view, the system writes the outlier back to the historical mandatory retention point record and marks it as a cross-zoom inheritance point, ensuring its continued retention in subsequent global or playback views.

[0072] When different scaling levels identify multiple outliers in the same time neighborhood, the system determines the primary outlier in the following order: outlier takes precedence over mutation point, mutation point takes precedence over peak / valley point, and peak / valley point takes precedence over short-term spike point. The remaining outliers are then retained as associated outliers in the same mandatory historical retention point record. If a higher scaling level subsequently confirms a more accurate original timestamp or original amplitude, the system updates the mandatory historical retention point record with the original timestamp and original amplitude confirmed by the higher scaling level.

[0073] Finally, the system reorders the set of display points after incorporating historical outliers in ascending order of timestamps, and adds cross-scale inheritance point markers to the incorporated historical outliers, resulting in a set of display points that are consistent across scales.

[0074] Through the above processing, S5 enables the global view, local zoom-in view, and playback view to no longer sample independently, but instead share anomaly point records, avoiding the problem of the same anomaly disappearing, shifting, or having inconsistent amplitudes at different zoom levels.

[0075] S6. Divide the stable historical segment and the updatable emerging segment and generate the data to be displayed in this refresh. In real-time refresh scenarios, the system divides the target time-series data segment into a stable historical segment and an updatable frontier segment based on the real-time refresh timestamp.

[0076] Specifically, the system reads the real-time refresh timestamp, the start and end times of the target time series data segment, and the stable time window. Based on the real-time refresh timestamp, it backs the stable time window to obtain the historical freeze boundary time. The system divides the portion of the target time series data segment that is earlier than or equal to the historical freeze boundary time into candidate historical segments, and the portion that is later than the historical freeze boundary time into candidate leading-edge segments.

[0077] Subsequently, the system performs data integrity checks on the candidate historical segments. If a candidate historical segment contains no missing time intervals, out-of-order data points, or unconfirmed delayed data points, it is designated as a stable historical segment. If a candidate historical segment contains missing time intervals, out-of-order data points, or unconfirmed delayed data points, it continues to be treated as an updatable data segment. Updatable leading-edge segments include data segments later than the historical freeze boundary time and candidate historical segments that do not meet the integrity requirements.

[0078] For a stable historical segment, the system locks the generated set of display points, the order of display points, the display amplitude, and the curve connection pattern, and prohibits subsequent real-time refresh processes from resampling or re-denoising the stable historical segment. If a late abnormal data point is received within the stable historical segment during a subsequent refresh process, the system does not change the original display pattern of the stable historical segment, but only adds an abnormal marker at the corresponding time position and writes the late abnormal data point into the historical mandatory retention point record. Subsequently, in the playback view or the next reload of this time range, the late abnormal data point participates in the display as a mandatory retention point.

[0079] For the updatable leading edge segment, the system continues to execute steps S2 to S5: calculating the segment display necessity index, determining the segment type, generating necessary display sampling points, performing hierarchical noise reduction, and cross-scaling consistency processing, to obtain the display point set for the updatable leading edge segment. Finally, the system concatenates the locked display point set of the stabilized historical segment with the display point set of the updatable leading edge segment in chronological order to obtain the dynamic chart display data corresponding to this refresh.

[0080] Through the above processing, S6 solves the problem of historical curves jumping due to resampling and noise reduction every time the real-time dynamic chart is refreshed, so that the historical facts that have been stabilized are displayed stably, while the latest data can still be updated in real time.

[0081] S7. Output dynamic chart display data and display source marker to the rendering end. After obtaining the dynamic chart display data corresponding to this refresh, the system outputs the dynamic chart display data to the dynamic chart rendering terminal in chronological order, and simultaneously outputs the display source marker for each display point.

[0082] In practice, the system reads the dynamic chart display data corresponding to this refresh and generates a sequence of rendering points in ascending order of timestamps. Each rendering point includes the display time, display amplitude, original timestamp source, and display source marker. The display source markers include stable compression points, trend preservation points, abnormal forced fidelity points, cross-scaling inheritance points, and historical freeze points.

[0083] Subsequently, the system outputs corresponding connection instructions to the dynamic chart rendering end based on the display source marker. For stable compression points, the system outputs a smooth connection instruction; for trend-preserving points, the system outputs a trend connection instruction; for abnormal forced fidelity points, the system outputs an original amplitude display instruction; for cross-scale inheritance points, the system outputs a forced display instruction; and for historical freeze points, the system outputs a disallow resampling instruction.

[0084] When the dynamic chart rendering client receives new real-time refreshed data, it prioritizes preserving the display position and connection pattern corresponding to historical freeze points, and only updates the rendering point sequence corresponding to the updatable leading edge segment. Therefore, the dynamic chart can compress stable data, retain short-term anomalies, maintain consistency across scaling, and stably display historical curves within a limited display space.

[0085] In summary, this embodiment divides the time-series data into local time-series segments according to the current display pixel capacity in S1; determines whether each segment is stationary and compressible, trend-preserving, or anomaly-forced fidelity in S2; generates necessary display sampling points according to segment type in S3; performs graded noise reduction on different types of display points in S4; maintains anomaly-forced preservation records consistent across scaling in S5; freezes stable historical segments and updates leading-edge segments in S6; and outputs display points with source markers to the rendering end in S7, thus forming a complete closed loop for dynamic chart adaptive sampling and noise reduction display. This method can reduce invalid drawing points under limited display space and real-time refresh conditions, while avoiding the hiding of short-term anomalies, changes in trend judgment, inconsistencies in display at different scaling levels, and repeated jumps in historical curves.

[0086] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data, characterized in that, Includes the following steps: S1. Obtain the time series data to be displayed, the current chart display time range, the current chart pixel width, the current zoom level, and the real-time refresh timestamp. Extract the target time series data segment from the time series data to be displayed according to the current chart display time range, and divide the target time series data segment into multiple local time series segments according to the time span corresponding to the pixel width. S2. Calculate the segment display necessity index for each local time series segment. The segment display necessity index includes at least the maximum value, minimum value, beginning and end trend change, local slope change, limit-breaking state, peak and valley state, sudden change state and short-term spike state within the segment. Based on the segment display necessity index, the corresponding local time series segment is determined as a stationary compressible segment, a trend-preserving segment or an abnormal forced fidelity segment. S3. Generate a set of sampling retention points based on the determination results of local time series segments. For stationary compressible segments, retain the segment start point, segment end point, and representative point. For trend retention segments, retain the segment start point, segment end point, and trend point with the largest slope change. For abnormal forced fidelity segments, retain the limit-breaking point, peak point, valley point, abrupt change point, or short-term spike point and its adjacent points. This results in a set of necessary sampling points for display corresponding to the current chart display time range. S4. Perform hierarchical noise reduction on the set of necessary sampling points for display. Specifically, representative points in stable and compressible segments are displayed smoothly according to the average trend of the segments. Trend points in trend-preserving segments are restricted from changing their original beginning and end trends after noise reduction. Points exceeding the limit, peak points, valley points, abrupt change points, or short-term spikes in abnormal forced fidelity segments are prohibited from amplitude reduction and their original timestamps and original amplitudes are preserved. This results in the set of display points after noise reduction. S5. Based on the current zoom level, read the historical forced retention point record that overlaps with the target time series data segment, and incorporate the over-limit points, peak points, valley points, abrupt change points and short-term spike points already identified in the global view, local zoom view or playback view into the noise-reduced display point set; when a new abnormal point not existing in the historical forced retention point record is identified in the current zoom level, write the abnormal point and its adjacent points before and after it into the historical forced retention point record to maintain the display consistency of the same abnormal position under different zoom levels; S6. Based on the real-time refresh timestamp, the target time series data segment is divided into a stable historical segment and an updatable leading edge segment. The stable historical segment is a time period that exceeds the current refresh time window and whose data integrity meets the preset integrity conditions. The generated set of display points and display format are locked for the stable historical segment, and only anomaly markers are allowed to be added. For the updatable leading edge segment, the adaptive sampling and noise reduction processing from S2 to S5 is performed again to obtain the dynamic chart display data corresponding to this refresh. S7. Output the dynamic chart display data to the dynamic chart rendering end in chronological order, and synchronously output the display source mark of each display point. The display source mark includes the stable compression point, trend retention point, anomaly forced fidelity point, cross-scale inheritance point and history freeze point, so that the dynamic chart can compress stable data, retain short-term anomalies, maintain cross-scale consistency and stably display historical curves within the limited display space.

2. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to claim 1, characterized in that: S1 divides the target time series data segment into multiple local time series segments, including the following steps: S11. Read the start time, end time, current chart pixel width, and current zoom level of the current chart display time range. Calculate the total display duration based on the start time and end time, and match the total display duration with the current chart pixel width to determine the basic time span corresponding to a single pixel column. S12. Extract the target time series data segment from the time series data to be displayed according to the current chart display time range, and sort the time series points in the target time series data segment in ascending order according to the timestamp. Remove data points with missing timestamps, duplicate timestamps and the same value to obtain the valid target time series data segment. S13. Determine the segment division time span based on the base time span and the current scaling level. The higher the current scaling level, the smaller the segment division time span; the lower the current scaling level, the larger the segment division time span, so that the segment division density matches the current display precision. S14. Using the start time of the current chart's displayed time range as the segment starting point, continuously divide the effective target time series data segments according to the segment's time span, and group the time series points falling into the same segmented time range into the same local time series segment. S15. For each local time segment, count the number of time points in the segment, the segment coverage duration, and the range of the corresponding pixel column. When the number of time points in the segment is empty, use the endpoints of adjacent valid segments as display boundary occupancy points. When the number of time points in the segment exceeds the display capacity of a single pixel, mark the local time segment as a segment to be compressed and evaluated. S16. Bind each local time segment to its start and end times, the range of the corresponding pixel column, the list of original time points within the segment, and the evaluation marker to be compressed, to generate a set of local time segments for calculating the necessity index for subsequent segment display.

3. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 or 2, characterized in that: S2 calculates the necessity index for displaying segments and determines the type of local time segments, including the following steps: S21. For each local time segment, read the list of original time points within the segment and calculate the segment start value, segment end value, maximum value within the segment, minimum value within the segment, maximum value timestamp, and minimum value timestamp to form a segment amplitude index group. S22. Calculate the trend change at the beginning and end of the segment based on the starting and ending values ​​of the segment, and calculate the local slope change based on the direction and magnitude of the numerical changes between adjacent time points within the segment, forming a segment trend index group. S23. Compare the value of each time point in the segment with the preset upper limit threshold and the preset lower limit threshold. When any time point exceeds the preset upper limit threshold or falls below the preset lower limit threshold, mark the local time segment as having an over-limit state and record the over-limit point timestamp and over-limit magnitude. S24. Compare the maximum value within a segment with the corresponding boundary values ​​of the preceding and following segments. When the maximum value is higher than its immediate neighbors and the difference exceeds the peak value determination threshold, mark the peak state. Compare the minimum value within a segment with the corresponding boundary values ​​of the preceding and following segments. When the minimum value is lower than its immediate neighbors and the difference exceeds the valley value determination threshold, mark the valley value state. S25. Calculate the mutation state based on the magnitude and duration of numerical changes between adjacent time points. When the magnitude of numerical changes per unit time exceeds the mutation judgment threshold, mark the corresponding time point as a mutation point and mark the local time segment as having a mutation state. S26. For time series points in the segment whose duration is shorter than the preset short-time window and whose amplitude deviates significantly from the previous and next stable intervals, short-time spike judgment is performed. When the time series point rises or falls rapidly within the short-time window and then recovers to the neighborhood reference range, the time series point is marked as a short-time spike point, and the local time series segment is marked as having a short-time spike state. S27. When a local time series segment does not have any out-of-limit state, peak-valley state, abrupt change state, or short-term spike state, and the trend changes at the beginning and end and the local slope changes are all lower than the corresponding stability thresholds, the local time series segment is determined to be a stationary compressible segment; when a local time series segment does not have any abnormal state but the trend changes at the beginning and end or the local slope changes reach the trend threshold, it is determined to be a trend-preserving segment; when a local time series segment has at least one of the out-of-limit state, peak-valley state, abrupt change state, or short-term spike state, it is determined to be an abnormal forced fidelity segment.

4. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 3, characterized in that: S3 generates a set of sample retention points based on the determination results of local time segments, including the following steps: S31. For a local time series segment that is determined to be a stationary compressible segment, read the segment start point, segment end point and segment average value of the segment, and select the original time series point that is closest to the segment average value and whose timestamp is located in the middle region of the segment as the representative point to generate a stationary compressible point group. S32. For a local time series segment that is determined to be a trend-preserving segment, read the slope change value between adjacent time series points within the segment, and select the original time series point with the largest slope change value as the trend point, and generate a trend-preserving point group together with the segment start point and segment end point. S33. For local time segments that are determined to be abnormal forced fidelity segments, read the over-limit points, peak points, valley points, mutation points and short-term spike points, and synchronously retain the previous original time segment and the next original time segment as adjacent points for each abnormal point to generate an abnormal forced fidelity point group. S34. When the same original time series point belongs to both a trend point and an anomaly point, the original time series point shall be treated as an anomaly forced fidelity point and its original timestamp, original amplitude and anomaly type mark shall be retained. S35. When two adjacent local time segments retain the same boundary time point, only one boundary time point is retained and it is simultaneously bound to the end of the previous segment and the start of the next segment to avoid the dynamic chart rendering end from repeatedly drawing the same time position. S36. Merge the stable compression point group, trend retention point group, and abnormal forced fidelity point group in ascending order of timestamp, and remove duplicate time series points to obtain the set of necessary sampling points for display corresponding to the current chart display time range.

5. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 4, characterized in that: S4 performs hierarchical noise reduction on the set of necessary sampling points for display, including the following steps: S41. Read the source segment type and original amplitude of each display point in the set of necessary sampling points for display, and establish noise reduction rules according to stationary compressible segments, trend-preserving segments and abnormal forced fidelity segments respectively; S42. For representative points derived from stationary compressible segments, calculate smoothed display values ​​according to the average trend of the original time series points within the segment and the segment start and end point constraints, and use the smoothed display values ​​as the display amplitude of the representative points to reduce random jitter in stationary segments at the display end. S43. For trend points derived from trend-preserving segments, compare the beginning and end trend directions before and after noise reduction. When the trend direction after noise reduction is inconsistent with the original beginning and end trend directions, cancel the noise reduction amplitude of the trend point and restore it to the original amplitude or trend constraint amplitude so that the noise reduction process does not change the original trend judgment. S44. For over-limit points, peak points, valley points, abrupt change points, or short-term spike points originating from abnormal forced fidelity segments, it is prohibited to perform smoothing processing such as reducing peak values, raising valley values, or weakening over-limit amplitude on the abnormal points, and the original timestamp and original amplitude of the abnormal points should be directly retained. S45. For adjacent points before and after the anomaly in the anomaly forced fidelity segment, retain the temporal order and amplitude change direction of the adjacent points relative to the anomaly, and only allow mild noise suppression within the range that does not change the anomaly formation process. S46. The stable compression points, trend retention points, abnormal forced fidelity points and their adjacent points that have completed noise reduction processing are rearranged in chronological order to form a set of display points after noise reduction.

6. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 5, characterized in that, Maintaining consistent display of the same abnormal location across different scaling levels in S5 includes the following steps: S51. Based on the current zoom level, the current chart display time range, and the start and end times of the target time series data segment, query the historical abnormal point records in the historical forced retention point records that overlap with the target time series data segment. S52. Read the over-limit points, peak points, valley points, abrupt change points and short-term spike points in the historical anomaly point record, and match them with the timestamp of the denoised display point set. When the historical anomaly point is not included in the denoised display point set, merge the historical anomaly point and its adjacent points into the denoised display point set. S53. When the current zoom level is a local zoom-in view, the historical anomaly points identified in the global view are set as the forced retention points of the local zoom-in view, and their original amplitude and original occurrence time are retained first to prevent the anomaly position from shifting due to resampling of the local zoom-in view. S54. When the current zoom level is a global view or a playback view, newly identified outliers in the local zoom-in view are written back to the history forced retention point record, and the outlier is marked as a cross-zoom inheritance point so that it continues to be retained in subsequent global views or playback views. S55. When different scaling levels identify multiple outliers in the same time neighborhood, the main outlier is determined in the order of outlier points taking priority over mutation points, mutation points taking priority over peak and valley points, and peak and valley points taking priority over short-term spikes. The remaining outliers are retained as associated outliers in the same historical forced retention point record. S56. Reorder the set of display points after incorporating historical outliers in ascending order of timestamps, and add cross-scale inheritance point markers to the incorporated historical outliers to obtain a set of display points that are consistent across scales.

7. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 6, characterized in that, S6 divides the stable historical segment and the updatable emerging segment based on the real-time refresh timestamp, including the following steps: S61. Read the real-time refresh timestamp, the start and end times of the target time series data segment, and the stable time window. Based on the real-time refresh timestamp, backtrack the stable time window to obtain the historical frozen boundary time. S62. The portion of the target time series data segment that is earlier than or equal to the historical freezing boundary time is divided into candidate historical segments, and the portion that is later than the historical freezing boundary time is divided into candidate frontier segments. S63. Check the data integrity of the candidate historical segment. If there are no missing time intervals, out-of-order data points and unconfirmed delayed data points in the candidate historical segment, the candidate historical segment shall be determined as a stable historical segment. S64. Lock the generated set of display points, display point order, display amplitude and curve connection form for the stable historical segment, and prohibit the subsequent real-time refresh process from resampling or re-denoising the stable historical segment; S65. When a late abnormal data point is received in the already stable historical segment during the subsequent refresh process, the original display form of the already stable historical segment is not changed. Instead, an abnormal mark is added at the corresponding time position, and the late abnormal data point is written into the history forced retention point record. S66. Continue to perform adaptive sampling, hierarchical noise reduction and cross-scaling consistency processing from S2 to S5 on the candidate leading edge segments and the candidate historical segments that do not meet the data integrity requirements to obtain the set of display points that can update the leading edge segments. S67. Combine the set of locked display points for the stable historical segment with the set of display points for the updatable frontier segment in chronological order to obtain the dynamic chart display data corresponding to this refresh.

8. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 7, characterized in that, In S7, outputting dynamic chart display data and display source markers to the dynamic chart rendering end includes the following steps: S71. Read the dynamic chart display data corresponding to this refresh, and generate a rendering point sequence in ascending order of timestamps, so that each rendering point contains the display time, display amplitude, original timestamp source, and display source marker; S72. Based on the displayed source markers, the rendered points are divided into stable compressed points, trend preserved points, abnormal forced fidelity points, cross-scale inherited points, and historical frozen points, and the different source markers are written into the point attribute field of the dynamic chart rendering end. S73. Output a smooth connection command for stable compression points, a trend connection command for trend preservation points, an original amplitude display command for abnormal forced fidelity points, a forced display command for cross-scaling inheritance points, and a disallow resampling command for historical freeze points. S74. When the dynamic chart rendering terminal receives new real-time refresh data, it prioritizes retaining the display position and connection form corresponding to the historical frozen point, and only performs updates on the rendering point sequence corresponding to the updatable leading edge segment. S75. Synchronously output the rendering point sequence, point attribute fields, and connection instructions to the dynamic chart rendering end, so that the dynamic chart can complete data compression, anomaly fidelity, cross-scaling inheritance, and stable historical display according to display necessity within the limited display space.

9. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to claim 3, characterized in that, The determination of short-term spike states includes the following steps: S261. Read the original time points within each local time segment, and establish a forward neighborhood and a backward neighborhood with each original time point as a candidate peak point. The time length of the forward neighborhood and the backward neighborhood shall not exceed the preset short time window. S262. Calculate the magnitude deviation of the candidate peak point relative to the forward neighborhood reference value and the backward neighborhood reference value respectively. When the magnitude deviation of the candidate peak point exceeds the short-term peak magnitude threshold, mark the candidate peak point as a peak candidate point. S263. Determine whether the backward neighborhood after the peak candidate point returns to the neighborhood reference range within the preset recovery time. When the time series value after the peak candidate point recovers to the neighborhood reference range, confirm that the peak candidate point has short-term recovery characteristics. S264. Determine whether the duration of the candidate peak is shorter than the preset short-term duration threshold. If the duration is shorter than the preset short-term duration threshold and has short-term recovery characteristics, determine the candidate peak as a short-term peak. S265. Write the short-time spike point, its preceding original time series point, and its following original time series point into the abnormal forced fidelity point group, and prohibit weakening the original amplitude of the short-time spike point in subsequent graded noise reduction processing.

10. The dynamic chart adaptive sampling and noise reduction display algorithm based on time-series data according to any one of claims 1 to 9, characterized in that, Updating historical mandatory retention records involves the following steps: S511. After completing the cross-scaling consistency processing of S5 each time, read the newly identified over-limit points, peak points, valley points, abrupt change points and short-term spike points within the current chart display time range, and determine whether each abnormal point already exists in the historical forced retention point record. S512. When a newly identified outlier does not exist in the historical mandatory retention point record, write the original timestamp, original amplitude, outlier type, current scaling level, source local time segment, and preceding and following adjacent points of the outlier into the historical mandatory retention point record. S513. When a newly identified outlier already exists in the historical mandatory retention point record, compare the original timestamp and original amplitude of the current outlier with the existing record. When the two belong to the same time neighborhood, supplement the existing record with the current scaling level and the source local time segment, and do not generate a new mandatory retention point repeatedly. S514. When an outlier in a historical mandatory retention point record is confirmed to have a more accurate original timestamp or original amplitude at a higher scaling level, the historical mandatory retention point record shall be updated with the original timestamp and original amplitude confirmed at the higher scaling level. S515. The updated historical forced retention point record is used as the forced retention basis for subsequent global view, local zoom view, playback view and real-time refresh view, so that the same exception is displayed consistently in different display states.

Images