A Power Grid Frequency Estimation Method and System Based on Event Cameras

Abstract

This invention belongs to the field of power technology and discloses a power grid frequency estimation method and system based on an event camera. The invention first records the event stream corresponding to light source flickering in a lighting environment using an event camera; then, it resamples the event stream along the time dimension to obtain several uniformly sampled event blocks; next, it performs polarity normalization on all event blocks to obtain the polarity of the lighting intensity change for each event block within its corresponding time period; it combines the polarities of the lighting intensity change of all event blocks chronologically to obtain a complete lighting polarity sequence; finally, it performs time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment. This invention can obtain good power grid frequency estimation results.

Description

Technical Field

[0001] This invention belongs to the field of power technology, and more specifically, relates to a power grid frequency estimation method and system based on event cameras. Background Technology

[0002] The power grid frequency, or Electric Network Frequency (ENF), is the frequency used in urban power distribution networks. Power distribution networks worldwide use transmission system standards based on a nominal frequency of 50 or 60 Hz, meaning the nominal ENF frequency is 50 / 60 Hz globally. Because the total instantaneous power consumption in cities is difficult to estimate, power supply departments struggle to match supply and demand. Therefore, the ENF frequency cannot remain constant at its nominal value and fluctuates. However, if the ENF frequency fluctuates too much from its nominal value, it can lead to unstable operation of power supply equipment and appliances, causing damage and significant economic losses. Therefore, power supply departments strive to control this frequency fluctuation within a small range around its nominal value to ensure power safety and stability. This results in the ENF frequency exhibiting random fluctuations within a small range around its nominal value. Furthermore, since the intensity of a light source is proportional to its instantaneous power supply, the intensity of the light source changes with the ENF frequency. Therefore, by recording these changes in light intensity using images or videos, the ENF frequency variation can be estimated. However, due to the low dynamic range and insufficient sampling rate of traditional cameras, images or videos often fail to clearly record changes in lighting, resulting in poor or inaccurate estimations of the power grid frequency. Summary of the Invention

[0003] This invention provides a power grid frequency estimation method and system based on event cameras, which solves the problem of poor performance in power grid frequency estimation using traditional cameras in existing technologies.

[0004] This invention provides a power grid frequency estimation method based on event cameras, comprising the following steps:

[0005] Step 1: Record the event stream corresponding to the flickering of light sources in the lighting environment using an event camera;

[0006] Step 2: Resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled;

[0007] Step 3: Perform polarity normalization on all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete lighting polarity sequence.

[0008] Step 4: Perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

[0009] Preferably, in step 2, the resampling method is as follows:

[0010] e i =e×g τ (t0+i×Δt)

[0011] Where e is the raw event stream recorded by the event camera, e i It is the i-th event block obtained by resampling the original event stream e, g τ (t) is a rectangular function, t0 is the time when the first event in the event stream occurs, and Δt is the sampling time interval;

[0012] g τ (t0+i×Δt)=u(t0+i×Δt)-u(t0+(i-1)×Δt)

[0013] Where u(t) is the step function.

[0014] Preferably, in step 3, the polarity normalization of the event block is performed in the following way:

[0015]

[0016] Where E(i) is the polarity of the illumination intensity change of the i-th event block within its corresponding time period; sgn(n) is the sign function, which returns +1 or -1; It is the number of positive events in the i-th event block. It represents the number of negative events in the i-th event block.

[0017] Preferably, in step 3, the illumination polarity sequence is represented as follows:

[0018] E(n)=(E(1),E(2),…,E(i),…)

[0019] Where E(n) is the illumination polarity sequence and n is the total number of event blocks.

[0020] Preferably, step 4 includes the following sub-steps:

[0021] Step 401: Bandpass filter the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence;

[0022] Step 402: Perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum results within each time window, take the frequency component with the strongest energy as the instantaneous power grid frequency estimation result at the corresponding time, and thus obtain the power grid frequency estimation result at the second harmonic within the entire event flow time period.

[0023] Step 403: Normalize the power grid frequency estimation result at the second harmonic to the nominal power grid frequency value to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment, expressed as:

[0024] f ENF =f 2ENF / 2

[0025] Among them, f 2ENF This is the power grid frequency estimation result at the second harmonic, f ENF It is the standard power grid frequency corresponding to the harmonics.

[0026] Preferably, in step 402, when performing short-time Fourier transform analysis on the filtered sequence, the time window is set to 16*1 / Δt, the step size is set to 1 / Δt, and Δt is the sampling time interval set in step 2 when resampling.

[0027] On the other hand, the present invention provides a power grid frequency estimation system based on an event camera, comprising:

[0028] An event camera is used to record and capture the event stream corresponding to the flickering of light sources in a lighting environment.

[0029] The resampling unit is used to resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled.

[0030] The polarity normalization unit is used to normalize the polarity of all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; and to combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete recording of the illumination polarity sequence.

[0031] The time-frequency analysis unit is used to perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

[0032] Preferably, the resampling unit performs resampling in the following manner:

[0033] e i =e×g τ (t0+i×Δt)

[0034] Where e is the raw event stream recorded by the event camera, e iIt is the i-th event block obtained by resampling the original event stream e, g τ (t) is a rectangular function, t0 is the time when the first event in the event stream occurs, and Δt is the sampling time interval;

[0035] g τ (t0+i×Δt)=u(t0+i×Δt)-u(t0+(i-1)×Δt)

[0036] Where u(t) is the step function.

[0037] Preferably, the polarity normalization unit performs polarity normalization on the event block in the following manner:

[0038]

[0039] Where E(i) is the polarity of the illumination intensity change of the i-th event block within its corresponding time period; sgn(n) is the sign function, which returns +1 or -1; It is the number of positive events in the i-th event block. It represents the number of negative events in the i-th event block.

[0040] Preferably, the time-frequency analysis unit includes:

[0041] A bandpass filter unit is used to perform bandpass filtering on the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence.

[0042] The transformation unit is used to perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum results within each time window, take the frequency component with the strongest energy as the instantaneous power grid frequency estimation result at the corresponding time, and thereby obtain the power grid frequency estimation result at the second harmonic within the entire event flow time period.

[0043] The normalization unit is used to normalize the power grid frequency estimation result at the second harmonic to the nominal value of the power grid frequency, so as to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment.

[0044] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0045] This invention addresses the problem that traditional cameras are easily affected by the low dynamic range and low sampling rate of the sensor itself when estimating power grid frequency information in light source illumination changes. Furthermore, pixel intensity changes in the recorded video are easily interfered with by factors other than motion content related to illumination changes, leading to poor power grid frequency estimation results using traditional cameras. This invention employs an event camera with high dynamic range and high temporal resolution to collect illumination intensity changes. Specifically, the event camera records the event stream corresponding to light source flickering in the lighting environment. Then, the event stream recorded by the event camera is resampled in the temporal dimension to obtain uniformly sampled event blocks. Next, the event polarity in the event blocks is normalized to obtain the polarity sequence of the corresponding temporal ambient illumination changes, i.e., the illumination polarity sequence. Finally, based on the understanding that the power grid frequency fluctuates only within a small range around the nominal value and that the illumination change frequency is twice the power grid frequency, bandpass filtering is applied to the illumination polarity sequence at twice the nominal value. The power grid frequency estimation result at the second harmonic is obtained through short-time Fourier transform and spectral peak search, and then standardized, thus realizing power grid frequency estimation based on the event camera. This invention not only proposes for the first time the use of event cameras to estimate power grid frequency, but also addresses the significant difference between the data patterns output by event cameras and those of traditional visual sensors, which prevents traditional camera-based methods from being applied to the event streams generated by event cameras. It proposes a specific solution that fully utilizes the high dynamic range and high temporal resolution of event cameras to convert discrete event streams into uniformly sampled illumination polarity sequences. This solves the problem of traditional cameras struggling to handle scenarios with weak lighting and motion interference, resulting in excellent estimation results. Attached Figure Description

[0046] Figure 1 This is a comparison of the data recording formats of traditional cameras and event cameras;

[0047] Figure 2 This is a schematic diagram illustrating the principle of an event camera recording changes in lighting.

[0048] Figure 3 This is a flowchart illustrating a power grid frequency estimation method based on an event camera provided in an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram illustrating the process of resampling the event stream in the time dimension in a power grid frequency estimation method based on an event camera provided in an embodiment of the present invention.

[0050] Figure 5 This is a schematic diagram illustrating the process of polarity normalization of event blocks in a power grid frequency estimation method based on an event camera provided in an embodiment of the present invention.

[0051] Figure 6This is a schematic diagram of power grid frequency estimation information obtained by a power grid frequency estimation method based on an event camera provided in an embodiment of the present invention. Detailed Implementation

[0052] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0053] In urban lighting environments, some light sources exhibit flickering that is invisible to the human eye but can be recorded by sensors such as cameras. Since the intensity of a light source is directly proportional to its power supply, the flicker frequency of the light source is twice the instantaneous frequency of the power grid. This means that the instantaneous power supply frequency in the power grid can be estimated from the changes in lighting recorded by a camera. However, traditional cameras are easily affected by the low dynamic range and low sampling rate of the sensors themselves when estimating power grid frequency information from changes in light source illumination. Furthermore, changes in pixel intensity in the recorded video are easily interfered with by factors other than moving content related to lighting changes, resulting in poor performance when using traditional cameras for power grid frequency estimation.

[0054] The event camera or dynamic vision sensor (DVS) used in this invention is a novel sensor with a higher dynamic range and sampling rate compared to common vision sensors. Unlike traditional sensors that use a fixed sampling rate to record the absolute values ​​of pixel intensity, the event camera simulates the imaging principle of the biological retina, responding to changes in intensity at each pixel by generating pulses. More specifically, the event camera is a discrete, asynchronous sensor; due to its unique imaging method that only records changes in pixel intensity, its output is a series of event stream data, rather than a standard image. If at time t... j At that time, pixel position u j =(x j ,y j If the intensity change at point () reaches the threshold ±C (C>0), then an event e will be generated at that location. j =(x j ,y j ,t j ,p j ), p jThe interval {+1,-1} represents the polarity of the event, where +1 indicates an increase in pixel intensity and -1 indicates a decrease in pixel intensity. Therefore, the event camera outputs an asynchronous event stream recording the direction of pixel intensity changes, rather than the absolute pixel intensity values ​​recorded in traditional images. Compared to traditional frame-based image sensors, event cameras can capture brightness changes at almost an infinite frame rate and record the time and pixel location of the event point. Furthermore, event cameras have a high dynamic range, enabling them to respond to and record even small pixel changes. Therefore, for scenes with weak changes in lighting intensity or where pixel changes are disturbed by motion, event cameras offer significant advantages in data generation principles, sampling rates, and dynamic range.

[0055] like Figure 1 The image shows a comparison of data recording formats for a rotating disk with dots captured simultaneously using a conventional camera and an event camera. The conventional camera outputs a series of images with pixel absolute intensities recorded at fixed time intervals, while the event camera outputs a discrete stream of events recorded at a high sampling rate. Because the data pattern output by the event camera differs significantly from that of traditional vision sensors, many existing methods based on conventional cameras cannot be applied to the event streams generated by event cameras.

[0056] Based on the above considerations, this invention not only pioneers power grid frequency estimation based on event cameras, but also develops a specific solution that utilizes the high dynamic range and high temporal resolution of event cameras. By comparing the difference in pixel intensity changes before and after illumination with a threshold, and by preserving the polarity of pixel changes, power grid frequency information is recorded. (See [link to relevant documentation]). Figure 2 This invention utilizes the fact that event cameras are exceptionally sensitive to changes in lighting, ultimately enabling the acquisition of excellent power grid frequency estimation results.

[0057] A flowchart illustrating a power grid frequency estimation method based on an event camera provided by this invention is shown below. Figure 3 As shown, the main components include recording the flickering of light in the lighting environment to obtain an event stream, and processing the event stream to achieve power grid frequency estimation.

[0058] The present invention will now be described in detail with reference to Embodiments 1 and 2.

[0059] Example 1:

[0060] Example 1 provides a power grid frequency estimation method based on event cameras, comprising the following steps:

[0061] Step 1: Record the event stream corresponding to the flickering of light sources in the lighting environment using an event camera.

[0062] Due to the unique imaging method of event cameras (each pixel of an event camera is independent and only outputs an event when the change in pixel intensity exceeds a threshold), the data format of the event stream is always discrete. That is, any pixel position at any time in the event stream may generate an event: at any time, the position where the event is generated is not fixed, and for a certain pixel, the time interval between the generation of the event is not fixed.

[0063] Step 2: Resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled.

[0064] The purpose of step 2 is to transform the discrete event stream obtained in step 1 into a uniformly sampled representation, see [link to step 2]. Figure 4 The sampling method is as follows:

[0065] e i =e×g τ (t0+i×Δt)

[0066] Where e is the raw event stream recorded by the event camera, e i It is the i-th event block obtained by resampling the original event stream e, g τ (t) is a rectangular function, t0 is the time when the first event in the event stream occurs, and Δt is the sampling time interval.

[0067] For example, Δt can be set to 1 / 1000s.

[0068] In the above process, the rectangular function g τ (t0+i×Δt) can be represented by a step function:

[0069] g τ (t0+i×Δt)=u(t0+i×Δt)-u(t0+(i-1)×Δt)

[0070] Where u(t) is the step function.

[0071] After the above-mentioned sampling at equal time intervals, the original event stream is divided into continuous event blocks e sampled at a fixed frequency. i Each event block contains all event points within time Δt:

[0072]

[0073] Step 3: Perform polarity normalization on all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete lighting polarity sequence.

[0074] Normalize the polarity of events in each event block to obtain the polarity of the light intensity change at the corresponding time for the event block.

[0075] In step 2, after sampling the time dimension, a complete event stream is converted into continuous event blocks with a time interval of Δt. However, since the event stream may contain many events unrelated to lighting, in addition to the event points generated by lighting changes, there may also be many events generated by noise or motion. This makes the event polarity in each event block not uniform.

[0076] After step 2, the complete event stream is divided into multiple event blocks according to chronological order. However, the discrete data format of the event stream results in uneven event recording within each event block: the pixel positions where events occur are not fixed. Furthermore, the polarity of events within an event block is inconsistent: in static scenes, due to the high sensitivity of events, many noise events are inevitably generated, and the polarity of these noise events is unpredictable; in dynamic scenes, the event camera is also very sensitive to pixel changes caused by motion, so the event block will contain a large number of motion events unrelated to lighting.

[0077] To determine the direction of lighting changes within a given time period, the polarity of all events within the event block is normalized.

[0078] Based on the understanding that the number of noise events is relatively small and that the movement of a closed object will simultaneously generate a similar number of positive and negative polarity events, in the event block polarity normalization stage, the "mode criterion" is used to determine the polarity of the event changes within the corresponding time period of each event block, resulting in the illumination polarity sequence E(n). (See [reference needed]). Figure 5 The polarity normalization method for the i-th event block is represented as follows:

[0079]

[0080] Where E(i) is the polarity of the illumination intensity change of the i-th event block within its corresponding time period; sgn(n) is the sign function, which returns +1 or -1; It is the number of positive events in the i-th event block. It represents the number of negative events in the i-th event block.

[0081] This invention uses the sgn(n) function to convert the polarity of numerous events in an event block into a single illumination polarity representation with a value of +1 / -1, which can remove the influence of the polarity of noise events and motion events.

[0082] The normalized illumination intensity change polarities of each event block are combined chronologically to obtain a sequence E(n) recording the complete illumination polarities:

[0083] E(n)=(E(1),E(2),…,E(i),…)

[0084] Where E(n) is the illumination polarity sequence and n is the total number of event blocks.

[0085] Step 4: Perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

[0086] Since the event stream is processed into an illumination polarity sequence E(n) with values ​​of +1 / -1, which includes information about light source flicker, a time-frequency analysis is performed on the illumination polarity sequence at twice the nominal grid frequency to obtain a final estimate of the grid frequency.

[0087] Specifically, step 4 includes the following sub-steps:

[0088] Step 401: Bandpass filter the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence.

[0089] Since the flicker frequency of the lighting source is twice the instantaneous value of the grid frequency, the lighting polarity sequence E(n) is first bandpass filtered at twice the nominal value of the grid frequency to obtain the corresponding filtered sequence, as shown below:

[0090] E(n)→E 2ENF (n)

[0091] Among them, E 2ENF E(n) is the filtered sequence, which is the sequence obtained after bandpass filtering E(n) at twice the nominal value of the power grid frequency.

[0092] Step 402: Perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum within each time window, and take the frequency component with the strongest energy as the instantaneous grid frequency estimation result for the corresponding time. This yields the grid frequency estimation result f at the second harmonic throughout the entire event flow time period. 2ENF .

[0093] The formula for the short-time Fourier transform is:

[0094]

[0095] Where w(n) is the window function; R is the step size; m is the number of steps of the window function; mR means stepping to the right m times; ω is the angular frequency.

[0096] For example, the time window is set to 16*1 / Δt, the step is set to 1 / Δt, and Δt is the sampling time interval set when resampling is performed in step 2.

[0097] Step 403: Normalize the power grid frequency estimation result at the second harmonic to the nominal power grid frequency value to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment, expressed as:

[0098] f ENF =f 2ENF / 2

[0099] Among them, f 2ENF This is the power grid frequency estimation result at the second harmonic, f ENF It is the standard power grid frequency corresponding to the harmonics.

[0100] Normalizing the grid frequency estimate at twice the nominal value back to the nominal grid frequency gives the standard grid frequency f corresponding to the harmonics. ENF Since the frequency of lighting flicker is proportional to the frequency of the power grid, the second harmonic estimation result is standardized by dividing by 2. Figure 6 An example of a standardized power grid frequency estimation result is given (the +0.02 in the figure is an additional bias added to avoid curve overlap and to make the comparison between the estimation result and the reference value clearer and more obvious). It can be seen that the estimation result is good and very consistent with the fluctuation pattern of the power grid frequency.

[0101] Example 2:

[0102] Example 2 provides a power grid frequency estimation system based on an event camera, comprising:

[0103] An event camera is used to record and capture the event stream corresponding to the flickering of light sources in a lighting environment.

[0104] The resampling unit is used to resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled.

[0105] The polarity normalization unit is used to normalize the polarity of all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; and to combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete recording of the illumination polarity sequence.

[0106] The time-frequency analysis unit is used to perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

[0107] Specifically, the resampling unit performs resampling in the following manner:

[0108] e i =e×g τ (t0+i×Δt)

[0109] Where e is the raw event stream recorded by the event camera, e i It is the i-th event block obtained by resampling the original event stream e, g τ (t) is a rectangular function, t0 is the time when the first event in the event stream occurs, and Δt is the sampling time interval;

[0110] g τ (t0+i×Δt)=u(t0+i×Δt)-u(t0+(i-1)×Δt)

[0111] Where u(t) is the step function.

[0112] The polarity normalization unit performs polarity normalization on the event block in the following manner:

[0113]

[0114] Where E(i) is the polarity of the illumination intensity change of the i-th event block within its corresponding time period; sgn(n) is the sign function, which returns +1 or -1; It is the number of positive events in the i-th event block. It represents the number of negative events in the i-th event block.

[0115] The time-frequency analysis unit includes:

[0116] A bandpass filter unit is used to perform bandpass filtering on the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence.

[0117] The transformation unit is used to perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum results within each time window, take the frequency component with the strongest energy as the instantaneous power grid frequency estimation result at the corresponding time, and thereby obtain the power grid frequency estimation result at the second harmonic within the entire event flow time period.

[0118] The normalization unit is used to normalize the power grid frequency estimation result at the second harmonic to the nominal value of the power grid frequency, so as to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment.

[0119] Example 2 provides a system corresponding to the method provided in Example 1. The system provided in Example 2 is used to execute the steps in the method of Example 1, so it will not be described again. The system can be understood by referring to the description of the method.

[0120] The power grid frequency estimation method and system based on event cameras provided in this invention have at least the following technical advantages:

[0121] Compared to traditional cameras, which are easily affected by scene factors such as motion when recording changes in light source illumination, and traditional visual sensors, which suffer from insufficient sampling and thus make it difficult to accurately estimate the power grid frequency in video recording, this invention utilizes the unique imaging principle of an event camera and its high temporal resolution and high dynamic range to achieve power grid frequency estimation in more complex situations with changes in lighting intensity. This invention is the first to use an event camera and has achieved good power grid frequency estimation results.

[0122] Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A power grid frequency estimation method based on event cameras, characterized in that, Includes the following steps: Step 1: Record the event stream corresponding to the flickering of light sources in the lighting environment using an event camera; Step 2: Resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled; Step 3: Perform polarity normalization on all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete lighting polarity sequence. In step 3, the polarity normalization of the event block is performed in the following way: in, It is the first The polarity of the lighting intensity change of each event block within its corresponding time period; It is a sign function that returns +1 or -1. It is the first The number of positive events in each event block It is the first The number of negative events in each event block; Step 4: Perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

2. The power grid frequency estimation method based on event cameras according to claim 1, characterized in that, In step 2, the resampling method is as follows: in, It is the raw event stream recorded by the event camera. It is the original event stream. The first one obtained after resampling One event block, It is a rectangle function. It is the time when the first event in the event stream occurs. It is the sampling time interval; in, It is a step function.

3. The power grid frequency estimation method based on event cameras according to claim 1, characterized in that, In step 3, the illumination polarity sequence is represented as follows: in, It is the lighting polarity sequence, and n is the total number of event blocks.

4. The power grid frequency estimation method based on event cameras according to claim 1, characterized in that, Step 4 includes the following sub-steps: Step 401: Bandpass filter the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence; Step 402: Perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum results within each time window, take the frequency component with the strongest energy as the instantaneous power grid frequency estimation result at the corresponding time, and thus obtain the power grid frequency estimation result at the second harmonic within the entire event flow time period. Step 403: Normalize the power grid frequency estimation result at the second harmonic to the nominal power grid frequency value to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment, expressed as: in, This is the power grid frequency estimation result at the second harmonic. It is the standard power grid frequency corresponding to the harmonics.

5. The power grid frequency estimation method based on event cameras according to claim 4, characterized in that, In step 402, when performing short-time Fourier transform analysis on the filtered sequence, the time window is set to... Step settings , It is the sampling time interval set during resampling in step 2.

6. A power grid frequency estimation system based on an event camera, characterized in that, include: An event camera is used to record and capture the event stream corresponding to the flickering of light sources in a lighting environment. The resampling unit is used to resample the event stream in the time dimension to obtain several event blocks that are uniformly sampled. The polarity normalization unit is used to normalize the polarity of all event blocks to obtain the polarity of the illumination intensity change of each event block within its corresponding time period; and to combine the polarities of the illumination intensity change of all event blocks in chronological order to obtain a complete recording of the illumination polarity sequence. The polarity normalization unit performs polarity normalization on the event block in the following manner: in, It is the first The polarity of the lighting intensity change of each event block within its corresponding time period; It is a sign function that returns +1 or -1. It is the first The number of positive events in each event block It is the first The number of negative events in each event block; The time-frequency analysis unit is used to perform time-frequency analysis on the lighting polarity sequence to obtain the power grid frequency estimation information corresponding to the lighting environment.

7. The power grid frequency estimation system based on event cameras according to claim 6, characterized in that, The resampling unit performs resampling in the following manner: in, It is the raw event stream recorded by the event camera. It is the original event stream. The first one obtained after resampling One event block, It is a rectangle function. It is the time when the first event in the event stream occurs. It is the sampling time interval; in, It is a step function.

8. The power grid frequency estimation system based on event cameras according to claim 6, characterized in that, The time-frequency analysis unit includes: A bandpass filter unit is used to perform bandpass filtering on the lighting polarity sequence at twice the nominal value of the power grid frequency to obtain a filtered sequence. The transformation unit is used to perform short-time Fourier transform analysis on the filtered sequence, search for energy peaks in the estimated spectrum results within each time window, take the frequency component with the strongest energy as the instantaneous power grid frequency estimation result at the corresponding time, and thereby obtain the power grid frequency estimation result at the second harmonic within the entire event flow time period. The normalization unit is used to normalize the power grid frequency estimation result at the second harmonic to the nominal value of the power grid frequency, so as to obtain the standard power grid frequency corresponding to the harmonic, and use it as the power grid frequency estimation information corresponding to the lighting environment.

Images