A method for acquiring hydrological data through multi-sensor fusion

Abstract

This invention relates to the field of hydrological data acquisition technology, specifically a multi-sensor fusion hydrological data acquisition method. The method includes: fixedly installing a camera and a frequency-modulated continuous wave radar level gauge on a shore base; calibrating the floating object detection threshold and the reference length of each velocity measurement line during the calm water period; comparing the echo intensity frame-by-frame with the threshold during the flood period: if the intensity is below the threshold, the velocity measurement line length is corrected in real time based on the current water surface elevation, and the cross-sectional average flow velocity is obtained through spatiotemporal image velocity measurement; if the intensity is not below the threshold, the preceding flow velocity is estimated based on the water surface elevation to determine the scope of floating object influence and the affected frames are marked as invalid. This invention, through data fusion from a shore-based camera and a radar level gauge, corrects scale errors in real time using water surface elevation, independently identifies floating objects based on echo intensity, and estimates the preceding flow velocity based on changes in water surface elevation. It then determines the duration of floating object passage and the scope of its influence time by comparing the characteristic size of the floating object with the estimated flow velocity.

Description

Technical Field

[0001] This invention relates to the field of hydrological data acquisition technology, specifically a multi-sensor fusion hydrological data acquisition method. Background Technology

[0002] The spatiotemporal image velocimetry method uses a camera fixed on the riverbank to continuously capture images of the water surface. Velocity measurement lines are set along the direction of water flow in the image sequence. The pixel brightness values ​​of each frame on the velocity measurement line are superimposed along the time axis to construct a spatiotemporal image. The water surface velocity is calculated by extracting the main direction angle of the brightness texture stripes in the spatiotemporal image and combining it with the actual length of the velocity measurement line and the length of the time window. Then, the cross-sectional average velocity is obtained by combining it with the cross-sectional morphology. This method does not require the instrument to be placed in the water and is suitable for use under conditions of high flow velocity and high sediment content during flood season. The conversion ratio between the actual length of the velocity measuring line and the pixel length in the image depends on the oblique distance from the optical center of the camera to the water surface. Since the camera is fixedly installed on the shore, this oblique distance changes with the water surface elevation. The higher the water surface elevation, the shorter the distance, and the smaller the physical length of the water surface corresponding to the same set of pixels. If the fixed conversion ratio calibrated during the calm water period is always used during the flood, the flow velocity calculated when the water surface elevation deviates from the calm water level will be inaccurate. Moreover, the magnitude of the deviation increases with the increase of the water level rise. The water level deviation and flow velocity deviation are the largest during the flood peak, which is precisely the time when accurate data is most needed for flood control decisions.

[0003] Furthermore, during flood season, the river surface carries a large amount of floating solids. When these solids pass through the velocity measurement section, they form strong bands in the spatiotemporal image that are different from the direction of the water surface velocity texture. This causes the main direction angle of the texture to deviate from the direction of the water surface velocity, and the relevant image frames must be excluded from the velocity calculation. The time range for exclusion is determined by the characteristic size of the floating object in the direction of water flow and the water surface velocity when it passes through. However, the relevant image frames during the passage of floating objects are excluded due to interference, and the water surface velocity cannot be directly obtained from the image at this time. If the historical velocity of the most recent valid frame is substituted, the velocity has changed with the water level during the rise and fall of the flood, and the old velocity cannot represent the current moment, resulting in a deviation in the estimation of the time range of influence, which in turn leads to the waste of valid data or the inclusion of contaminated frames in the calculation. Summary of the Invention

[0004] (1) Technical problem to be solved: The purpose of this invention is to provide a multi-sensor fusion hydrological data acquisition method to solve the impact of a large number of floating objects on data distortion during the hydrological measurement process when the flood peak arrives.

[0005] (2) Technical solution: To achieve the above objectives, the present invention provides a method for acquiring hydrological data through multi-sensor fusion, the method comprising: A camera and a frequency-modulated continuous wave radar water level gauge are fixedly installed on the bank foundation on one side of the river cross section. The frequency-modulated continuous wave radar water level gauge synchronously outputs the elevation of the water surface on the bank and the echo intensity value during each sampling, so that each frame of the image captured by the camera corresponds to the elevation of the water surface on the bank and the echo intensity value at the corresponding sampling time in time. During calm water conditions without floating debris, the floating debris determination threshold is obtained by repeatedly collecting echo intensity values ​​from the frequency-modulated continuous wave radar level gauge. The reference length of each velocity line is obtained through the camera installation position parameters and camera parameters. During flood season, the echo intensity value corresponding to the current image frame is compared with the floating debris determination threshold frame by frame. When the echo intensity value is lower than the floating debris determination threshold, each velocity line of the current image frame is marked as valid. When the echo intensity value is not lower than the floating debris determination threshold, the cross-sectional average flow velocity corresponding to the most recent valid image frame is estimated based on the shore water surface elevation corresponding to the current image frame and the shore water surface elevation corresponding to the most recent valid image frame. The floating debris influence time range is determined by the floating debris characteristic size and the estimated flow velocity. Each velocity line of each image frame within the floating debris influence time range is marked as invalid, so that the corresponding image frames are excluded from the cross-sectional average flow velocity calculation. For each image frame marked as valid, the pixel brightness sequence of each velocity measurement line is read from the current image frame. The current length of each velocity measurement line is obtained by using the water surface elevation of the shore corresponding to the current image frame, the camera installation position parameters, and the reference length of each velocity measurement line. The pixel brightness sequences of each velocity measurement line are superimposed along the time dimension to construct a spatiotemporal image. The main direction angle of the texture is extracted from the spatiotemporal image. The water surface velocity of each velocity measurement line is obtained by using the main direction angle of the texture, the current length of each velocity measurement line, and the time window length. The cross-sectional average velocity is obtained by using the water surface velocity of each valid velocity measurement line.

[0006] Furthermore, the method for obtaining the reference length of each speed measuring line through camera installation position parameters and camera parameters includes: In the images captured by the camera during the still water period, the two endpoints of each velocity measurement line along the direction of water flow are calibrated to obtain the lateral pixel coordinates of the first endpoint of each velocity measurement line. Vertical pixel coordinates of the first endpoint of each speed measuring line Horizontal pixel coordinates of the second endpoint of each speed measuring line Vertical pixel coordinates of the second endpoints of each speed measuring line The difference in horizontal pixel coordinates of each speed measuring line is obtained by comparing the horizontal pixel coordinates of its first and second endpoints; the difference in vertical pixel coordinates of each speed measuring line is obtained by comparing the vertical pixel coordinates of its first and second endpoints; and the pixel length of each speed measuring line is obtained by comparing the differences in horizontal and vertical pixel coordinates. for: ; The vertical distance from the camera to the water surface during calm water conditions is obtained by comparing the camera's installation elevation with the water surface elevation during calm water conditions. Horizontal distance between each speed measuring line The distance from the center point of each speed measurement line along the direction perpendicular to the water flow to the orthographic projection point of the camera on the water surface during still water is given. The reference oblique distance of each speed measurement line is obtained by combining the vertical distance from the camera to the water surface during still water and the horizontal distance of each speed measurement line. ; By camera focal length Pixel length of each speed measuring line Slant distance from the reference speed measuring line Obtain the reference length of each speed measuring line .

[0007] Furthermore, the method for obtaining the current length of each speed measuring line by using the shore water surface elevation corresponding to the current image frame, the camera installation position parameters, and the reference length of each speed measuring line includes: Record the frame number of the current image frame as The vertical distance from the camera to the water surface is obtained by comparing the camera's installation elevation with the corresponding elevation of the water surface on the shore in the current image frame. Data acquisition stops when the elevation of the water surface on the shore corresponding to the current image frame is not lower than the camera's installation elevation; the current oblique distance of each speed measuring line is obtained by combining the current vertical distance from the camera to the water surface with the horizontal distance of each speed measuring line, where the current oblique distance of each speed measuring line is... ; The length correction ratio of each speed measuring line is obtained by comparing the current oblique distance of each speed measuring line with the reference oblique distance of each speed measuring line. ; The current length of each speed measuring line is obtained by comparing its length correction ratio with the reference length of each speed measuring line. .

[0008] Furthermore, the method for obtaining the floating object determination threshold by repeatedly collecting echo intensity values ​​from a frequency-modulated continuous wave radar level gauge includes: Under conditions of still water and no floating debris, the echo intensity values ​​were continuously collected multiple times using a frequency-modulated continuous wave radar level gauge. The total number of echo intensity value collections was recorded as the echo intensity collection count. The baseline mean echo intensity is obtained by combining all echo intensity values ​​with the number of echo intensity acquisitions. for: ; in For the first The echo intensity values ​​were collected once; the standard deviation of the echo intensity was obtained by comparing all echo intensity values ​​with the mean of the echo intensity reference. for: ; The floating object detection threshold is obtained by using the mean echo intensity, a preset multiplier, and the standard deviation of the echo intensity. ;in The preset multiplier is greater than 1.

[0009] Furthermore, the method for performing a currentized estimation of the cross-sectional average flow velocity corresponding to the most recent preceding valid image frame, and determining the time range of the floating object's influence by using the floating object's feature size and the currentized estimated flow velocity, includes: Floating object characteristic dimensions This is a preset percentile for the size sequence of floating objects along the water flow direction obtained from camera images during historical flood periods; the sequence number of the nearest preceding valid frame in time to the current image frame is denoted as... The average flow velocity of the cross section corresponding to the most recent valid image frame in time is denoted as... The elevation of the water surface along the shore corresponding to the most recent valid image frame in time is recorded as the elevation of the water surface of the previous valid frame. ; Record the frame number of the current image frame as ; by the elevation of the water surface along the shore corresponding to the current image frame , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame ; The duration of the floating object's influence is obtained by estimating the flow velocity using the feature size of the floating object and the current image frame. ; The time range of the floating object's influence is obtained by combining the current image frame acquisition time with the duration of the floating object's influence. The time range of the floating object's influence is a time interval with the current image frame acquisition time as the midpoint and the duration of the floating object's influence as the interval width.

[0010] Furthermore, the elevation of the water surface along the shore corresponding to the current image frame... , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame In this method, if there is no preceding valid image frame in the current image frame, the average flow velocity in the hydrological station's historical records for the time period corresponding to the current flood season is recorded as the historical reference average flow velocity. The duration of the floating object's influence in the current image frame is obtained by comparing the floating object's feature size with the historical average flow velocity. .

[0011] Furthermore, after marking each velocity measurement line in each image frame within the time range affected by the floating object as invalid, the echo intensity value corresponding to the current image frame is compared with the floating object determination threshold for each subsequent image frame. If the echo intensity value of the current image frame is lower than the floating object detection threshold, the current image frame will be counted in the consecutive valid frame count. If the echo intensity value corresponding to the current image frame is not lower than the floating object detection threshold, the consecutive valid frame count is reset to zero and the counting restarts, and the floating object influence time range is redefined; the continuous recovery duration is obtained by combining the consecutive valid frame count with the camera frame rate, where the camera frame rate... The frame rate at which the camera continuously captures images at a fixed frame rate, and the duration of continuous recovery. ; When the duration of continuous recovery exceeds the time window length, each speed measurement line will be remarked as valid.

[0012] Furthermore, before remarking each velocity measurement line as valid, if there are no valid image frames before the current time, the output of the cross-sectional average flow velocity is stopped; otherwise, for the image frames marked as invalid, the cross-sectional average flow velocity corresponding to the most recently valid image frame is used as the output of the cross-sectional average flow velocity at the current time; the number of consecutively marked invalid image frames is recorded as the consecutive invalid frame count. The duration of consecutive invalid frames is obtained by counting consecutive invalid frames and the camera frame rate. ; When the duration of continuous invalidity exceeds the preset maximum duration of invalidity. When the average flow velocity of the cross section is stopped, the output will resume after each velocity measurement line is remarked as valid.

[0013] Furthermore, the elevation of the water surface along the shore corresponding to the current image frame... , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame The methods include: The preset speed measurement section reference elevation is denoted as... The water depth corresponding to the current image frame is obtained by comparing the elevation of the water surface along the shore with the preset velocity measurement section reference elevation. The water depth corresponding to the previous valid frame is obtained by comparing the water surface elevation of the previous valid frame with the preset velocity measurement section reference elevation. ; The water depth ratio is obtained by comparing the water depth corresponding to the current image frame with the water depth corresponding to the previous valid frame. The estimated flow velocity of the current image frame is obtained by using the water depth ratio, the average flow velocity of the cross section corresponding to the previous valid image frame, and the preset water depth-flow velocity relationship index. The current image frame estimates the flow velocity by reflecting the power-law relationship between the average flow velocity at the cross-section and the water depth. .

[0014] Furthermore, the method for corresponding each frame of the image captured by the camera to the elevation of the water surface on the shore and the echo intensity value at the corresponding sampling time in time includes: By inputting a shared time reference signal to both the camera and the frequency-modulated continuous wave (FMCH) radar level gauge, the acquisition times of each image frame from the camera and each sampling time from the FMCH radar level gauge are recorded based on a unified time reference. When the sampling interval of the FMCH radar level gauge is greater than the frame interval of the camera, it will not exceed the current image frame acquisition time. The most recent radar sampling time is recorded as the previous sampling time. The radar sampling time immediately following the previous sampling time is recorded as the next sampling time. The time interpolation coefficients are obtained by using the current image frame acquisition time, the previous sampling time, and the subsequent sampling time. for: ; Using the time interpolation coefficient and the elevation of the water surface at the previous sampling time... Elevation of the shore water surface corresponding to the subsequent sampling time The elevation of the water surface along the shore corresponding to the current image frame is obtained, where the elevation of the water surface along the shore corresponding to the current image frame is obtained. for: ; Using the time interpolation coefficients and the echo intensity value corresponding to the previous sampling time. echo intensity value corresponding to the next sampling time The echo intensity value corresponding to the current image frame is obtained, where the echo intensity value corresponding to the current image frame is... for: ; When the sampling interval of the frequency-modulated continuous wave radar level gauge is no greater than the camera frame interval, each frame of image is directly correlated with the shore water surface elevation and echo intensity value of the most recent radar sampling in time.

[0015] (3) Compared with the prior art, the beneficial effects of the present invention are that, under the condition of multi-sensor fusion of shore-based fixed camera and frequency-modulated continuous wave radar water level gauge, the physical length of the velocity line is corrected in real time with the current water surface elevation, eliminating the scale error caused by water level changes; floating objects are identified by echo intensity value, making the identification independent of the image and unaffected by illumination; and the previous flow velocity is corrected by the change of the current water surface elevation relative to the previous effective frame, reducing the estimation deviation of the influence range of floating objects during the passage period. Attached Figure Description

[0016] The above and / or other aspects of this application will become more apparent from the description of certain embodiments with reference to the accompanying drawings, in which: Figure 1 This is a block diagram of a multi-sensor fusion hydrological data acquisition method according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the actual device structure according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Before providing examples, it is necessary to describe the application scenario of this invention. A hydrological department in a certain region has set up a flow measurement section on a mountain stream in the southwestern mountainous area. This river has a large gradient, and the floodwaters rise and fall rapidly during the flood season. The average annual precipitation exceeds 1400 mm, mainly concentrated from July to September. Historical observation data shows that the maximum flood rise at this section is approximately 6 meters. Based on the historical highest flood level of this section, the installation location is determined. A camera is fixedly installed on the side wall of a concrete bridge pier on one side of the flow measurement section, with an installation elevation higher than the historical highest flood level. The camera's field of view covers the entire water surface of the section. A frequency-modulated continuous wave radar level gauge is installed on the same side of the bridge pier, with the antenna also positioned higher than the historical highest flood level. The sampling interval is set to 5 seconds due to limitations in the on-site wireless transmission bandwidth, and the camera frame rate is 25 frames per second. The sampling times of both types of sensors are aligned to the same time reference using satellite timing signals. Figure 2As shown, a camera and radar level gauge are both installed on the shore. The radar beam is pointed towards the water surface, and the camera's line of sight is towards the velocity measurement line on the water surface. Floating objects are seen floating on the water. This station is an unattended automatic station; data is transmitted in real-time to the provincial flood control command platform via mobile communication network, making it impossible to rely on manual intervention to eliminate interference frames on-site. The upstream area of ​​this basin is covered with a large area of ​​bamboo forests and mixed coniferous and broad-leaved forests. During floods triggered by heavy rainfall, a large number of bamboo segments, branches, grass mats, and a small amount of agricultural waste flow into the river channel with slope runoff and appear at the velocity measurement section. Historical records show that floating objects appear as frequently as once every few minutes during floods, with single passages lasting from several seconds to about one minute, and sizes ranging from tens of centimeters to about 2 meters along the water flow direction. Heavy rainfall in this area mostly occurs at night, with more than half of the flood peaks occurring between midnight and 6:00 AM. Insufficient light during this period significantly reduces the reliability of identifying floating objects based solely on image content. During the rising water phase, the surface flow velocity increases rapidly with increasing water depth. Several minutes can pass between the detection of a floating object and the generation of the next valid frame, during which the water surface elevation can rise by several centimeters to tens of centimeters.

[0019] Example 1: As Figure 1 As shown in the figure, this embodiment provides a method for acquiring hydrological data through multi-sensor fusion, the method comprising: A camera and a frequency-modulated continuous wave radar water level gauge are fixedly installed on the bank foundation on one side of the river cross section. The frequency-modulated continuous wave radar water level gauge synchronously outputs the elevation of the water surface on the bank and the echo intensity value during each sampling, so that each frame of the image captured by the camera corresponds to the elevation of the water surface on the bank and the echo intensity value at the corresponding sampling time in time. During calm water conditions without floating debris, the floating debris determination threshold is obtained by repeatedly collecting echo intensity values ​​from the frequency-modulated continuous wave radar level gauge. The reference length of each velocity line is obtained through the camera installation position parameters and camera parameters. During flood season, the echo intensity value corresponding to the current image frame is compared with the floating debris determination threshold frame by frame. When the echo intensity value is lower than the floating debris determination threshold, each velocity line of the current image frame is marked as valid. When the echo intensity value is not lower than the floating debris determination threshold, the cross-sectional average flow velocity corresponding to the most recent valid image frame is estimated based on the shore water surface elevation corresponding to the current image frame and the shore water surface elevation corresponding to the most recent valid image frame. The floating debris influence time range is determined by the floating debris characteristic size and the estimated flow velocity. Each velocity line of each image frame within the floating debris influence time range is marked as invalid, so that the corresponding image frames are excluded from the cross-sectional average flow velocity calculation. For each image frame marked as valid, the pixel brightness sequence of each velocity measurement line is read from the current image frame. The current length of each velocity measurement line is obtained by using the water surface elevation of the shore corresponding to the current image frame, the camera installation position parameters, and the reference length of each velocity measurement line. The pixel brightness sequences of each velocity measurement line are superimposed along the time dimension to construct a spatiotemporal image. The main direction angle of the texture is extracted from the spatiotemporal image. The water surface velocity of each velocity measurement line is obtained by using the main direction angle of the texture, the current length of each velocity measurement line, and the time window length. The cross-sectional average velocity is obtained by using the water surface velocity of each valid velocity measurement line.

[0020] For example, a camera and a frequency-modulated continuous wave radar level gauge are fixedly installed on the bank foundation on one side of the river cross-section. The camera is mounted on the side wall of a bridge pier, the top of a measurement house, or a dedicated steel frame, and its field of view covers the water surface area of ​​the flow velocity cross-section to be measured. The frequency-modulated continuous wave radar level gauge is installed on a stable bracket near the camera, with its antenna pointing towards the water surface and its beam illumination area located at a representative water surface position. In this specific embodiment, the camera installation elevation... Meters, camera frame rate Frames per second, image resolution Pixels, number of speed camera lines Sampling interval of frequency modulated continuous wave radar level gauge seconds, time window length The parameters mentioned above generally remain unchanged during operation after installation and commissioning. The camera can be configured with a near-infrared supplementary light or a low-light industrial camera, which can still form sufficient brightness contrast on the water surface at night to support spatiotemporal image texture extraction; unlike the scheme that relies on visible light images to identify the type of floating objects, the floating object discrimination in this application only depends on the radar echo intensity and is not affected by the supplementary lighting method.

[0021] The frequency-modulated continuous wave radar level gauge synchronously outputs the elevation of the water surface at the shore during each sampling. Two signals, the echo intensity value A (represented in volts as a linear voltage value in this embodiment), are sent to the main control unit. The water surface elevation is obtained by the instrument's signal processing unit through echo frequency spectrum analysis: the instrument emits microwaves with linearly increasing frequency towards the water surface, and the frequency difference between the echo and the emitted wave is proportional to the distance from the instrument to the water surface; the strongest main peak in the frequency spectrum corresponds to continuous large-area reflection from the water surface, and its beat frequency is converted into distance to obtain the water surface elevation; the echo intensity value A is the peak amplitude of this main peak.

[0022] Regarding the detection of floating objects using echo intensity value A, it should be noted that: In calm water, the reflection of radar microwaves is close to specular reflection; most of the incident energy propagates along the deflection direction, with only a small amount returning to the instrument antenna. Therefore, the echo intensity value A remains low and fluctuates little during calm water. Solid floating objects (trees, grass mats, foam plastics, construction waste, etc.) have a much rougher surface than calm water, causing diffuse scattering of radar microwaves. The scattered energy is distributed in all directions, and the scattered power received by the antenna increases significantly with the size of the floating object, resulting in an increase in the echo intensity value A. When a floating solid is on the water surface, it is at a close distance to the surface, and additional scattered energy is superimposed at the main peak frequency, causing the overall amplitude A of the main peak to increase. The scattering peak of the part of the floating object protruding above the water surface is adjacent to the main peak due to distance offset (usually a few centimeters to tens of centimeters corresponding to a small frequency difference) and does not affect the main peak. The location of the main peak frequency is determined, but it significantly increases A. This difference (the back energy difference between the diffuse scattering intensity of the solid and the specular reflection of the calm water surface) is usually several times larger than the natural fluctuations during the still water period, and can be distinguished from normal fluctuations within the threshold statistical framework.

[0023] Regarding the water level during the passage of floating objects It should be noted that: when a floating object floats on the water surface, the distance from its top to the instrument antenna is approximately the distance corresponding to the water surface elevation minus the height of the floating object above the water surface (usually several centimeters to tens of centimeters). The secondary peaks caused by scattering from these floating objects correspond to different beat frequencies in the frequency spectrum than the main peak, and there is a distinguishable distance between them and the main peak on the frequency axis. The instrument's main peak extraction algorithm uses the strongest peak as the source of the water surface elevation. The energy of the secondary peaks generated by a single or a few floating objects is usually lower than that of the main peak and will not change the frequency position of the main peak. Therefore, the water surface elevation output by the instrument is not accurate. The location still corresponds to the actual water surface position during the passage of floating objects. On-site verification method: Inspect the period of floating object passage (i.e.,...) (during the period of significant increase) Whether the time series is continuous with the water level change trend of the preceding and following periods indicates that the water level measurement has not been disturbed.

[0024] By inputting a shared time reference signal into the camera and the frequency-modulated continuous wave radar level gauge, the camera's... Frame image acquisition time With frequency modulated continuous wave radar level gauge Second sampling time All data are recorded based on the same time coordinate. The sampling interval of the frequency-modulated continuous wave radar level gauge... At that time, for each frame of image Corresponding shore water surface elevation With echo intensity value It is obtained by linear interpolation of two adjacent sample values; when At that time, each image frame directly corresponds to the most recent radar sampling value in time. It should be noted that, to ensure the detected transit duration is no less than [a certain duration], [further details are needed]. For floating object events, the echo intensity sampling interval of the frequency-modulated continuous wave radar level gauge should meet the following requirements. In the application scenario of this application, the typical transit time of floating objects is several seconds to about 1 minute. It is recommended to set the echo intensity sampling interval to no more than 1 second. The upload interval of water surface elevation can be kept unchanged at 5 seconds to reduce the communication bandwidth usage.

[0025] Before operation during the formal flood season, two calibrations were completed under calm water conditions with no floating debris: First, the floating debris detection threshold was obtained by repeatedly collecting echo intensity values ​​from the frequency-modulated continuous wave radar level gauge. Secondly, the reference length of each speed measuring line is obtained through the camera installation location parameters and camera parameters. .

[0026] During the flood season, each frame of the image Take the corresponding echo intensity value Threshold for determining floating objects Comparison: When When the current image frame's velocity measurement lines are marked as valid, the spatiotemporal image velocity measurement calculation begins; when At that moment, it is determined that there are floating objects causing distortion of the spatiotemporal image texture, based on the elevation of the water surface on the shore corresponding to the current image frame. The elevation of the water surface along the shore corresponding to the most recent preceding valid image frame in time. The average flow velocity of the cross section corresponding to the most recent preceding valid image frame. The estimated flow rate is obtained by performing current-based estimation. By the characteristic size of the floating object and Determine the time range of the floating object's influence, and mark each velocity line in each image frame within that range as invalid; the methods for current estimation and determination of the time range of influence will be detailed later.

[0027] For each image frame marked as valid, perform spatiotemporal image velocity calculations according to the following steps: The first step is to read each speed measurement line from the current image frame. ( The pixel brightness sequence on the image. Each speed measurement line is a set of pixels arranged along the speed measurement line direction and uniformly sampled. The number of pixels is determined by the endpoint coordinates determined during calibration. The brightness value of this set of pixels is read in each frame to form a brightness array.

[0028] The second step is to use the elevation of the water surface along the shore corresponding to the current image frame. Camera installation location parameters and reference lengths of each speed measuring line Get the current length of each speed measuring line .

[0029] The third step is to continuously frame( In this specific implementation method frame, (Rounding operation) for each speed measuring line The pixel brightness sequences are superimposed along the time axis to construct the spatiotemporal image of each speed measurement line. The spatiotemporal image is a two-dimensional array: the horizontal axis represents the spatial direction (corresponding to the direction of the speed measurement line from the first endpoint to the second endpoint), and the width is... Pixel, physical range is Meters; the vertical axis is the time direction (earlier frames on top, later frames on the bottom), and the height is [missing information]. Frame, physical time span is Second.

[0030] The fourth step is to extract the principal orientation angle of the texture from the spatiotemporal image. Its meaning is to denote the physical extent of the horizontal (spatial) axis of the spatiotemporal image as... The physical range of the meter and vertical (time) axis is denoted as Seconds constitute a coordinate system with physical dimensions; in this coordinate system, water surface velocity tracers (brightness features such as water ripples and foam that move with the water flow) form slanted texture stripes in the spatiotemporal image. The angle of inclination of the fringe relative to the horizontal axis (spatial axis), with a range of values. Corresponding to positive water flow ( The texture tilt angle extracted from the spatiotemporal image in pixels is denoted as... (Relative to the horizontal axis, in pixel coordinates), it is the same as The conversion relationship is as follows: ; extract One possible method is the gradient structure tensor method: for each pixel of the spatiotemporal image... Calculate the horizontal gradient With vertical gradient ,exist Accumulated tensor components within the neighborhood window , , Take the tensor matrix The direction of the principal feature vector is the principal direction of the stripes, which can be converted into the angle of inclination relative to the horizontal axis. The gradient structure tensor method is an existing method in the field of image processing, and its derivation will not be repeated in this specific implementation.

[0031] Fifth step, through the main direction angle of the texture Current length of each speed measuring line With time window length Calculate the surface velocity of the water at each velocity measurement line. . and The correspondence is as follows: In the physical coordinate system (horizontal axis) meters, vertical axis (seconds), water surface velocity tracer at speed Exercise, in Move horizontally within seconds Meters. Correspondingly, in this coordinate system, the texture stripes originate from a certain initial spatial position and undergo a horizontal displacement of... meters, vertical displacement is Seconds. Normalize the horizontal axis to [0,1] (i.e., divide by). ), normalize the vertical axis to [0,1] (i.e., divide by ), Then, the horizontal displacement of the fringes in the normalized coordinate system is... The vertical displacement is 1. Let the angle of inclination of the fringes relative to the normalized horizontal axis satisfy: ; The water surface velocity is obtained after transformation. for: ; In short, the faster the flow rate, the flatter the stripes. The smaller, The smaller, The larger the flow rate, the steeper the stripes. The larger, (The smaller). (The previous...) Converted to Substituting the derivative into the above formula, we get... Equivalently, the above formula can also be rewritten directly using the angle in pixel coordinates. This specific implementation method adopts the former.

[0032] Step 6: Calculate the average flow velocity of the cross-section using the surface velocity of each effective velocity measuring line, and let... This is the set of valid speed measurement line indices in the current frame. Number of effective velocity lines; average flow velocity across the cross section. for: ; when When, output And update the status synchronously: value update assignment to , value update assignment to , value update assignment to ;when At this time, the current frame is not output and the state is not updated.

[0033] The calculation process of the average flow velocity in a cross-section considering the cross-sectional shape: Let... For the first The surface velocity correction factor for each velocity measurement line converts the surface velocity into the average velocity of the cross section along that vertical line. It is determined by comparison measurements using an on-site current meter, typically ranging from 0.8 to 0.95. When on-site comparison data is unavailable, a value can be used. =0.85 is used as the initial value, and will be updated after data is accumulated. For the first The width of the water surface represented by each velocity line is determined by a cross-sectional topographic map or on-site measurements. The average flow velocity of the cross section is calculated as follows: .

[0034] For time window length It should be noted that: The meaning is the time span of the spatiotemporal image (i.e., the length of time covered by the superimposed frames). Number of frames superimposed when constructing a spatiotemporal image. The calculation requires the use of The condition for resuming speed measurement after the subsequent floating object passes is that the continuous recovery time must exceed [a certain duration]. The recovery requirement is that the accumulated time exceeds [a certain duration]. The subsequent relabeling is to ensure that the first complete spatiotemporal image (time span) constructed thereafter is valid. All frames in the dataset are valid frames without floating objects, thus ensuring reliable texture extraction results. This specific implementation method takes... Seconds, corresponding frame.

[0035] Regarding pixel angles, it should be noted that in a pixel-unit spatiotemporal image (width) Pixels, High In the frame), the water surface velocity tracer is in Intra-frame horizontal movement Pixels, vertical movement Pixel, the stripe tilt angle relative to the horizontal axis in pixel coordinates is ,satisfy Transformed to .

[0036] Example 2: Based on Example 1, this example further provides the method for obtaining the reference length of each speed measuring line through camera installation position parameters and camera parameters, including: In the images captured by the camera during the still water period, the two endpoints of each velocity measurement line along the direction of water flow are calibrated to obtain the lateral pixel coordinates of the first endpoint of each velocity measurement line. Vertical pixel coordinates of the first endpoint of each speed measuring line Horizontal pixel coordinates of the second endpoint of each speed measuring line Vertical pixel coordinates of the second endpoints of each speed measuring line The difference in horizontal pixel coordinates of each speed measuring line is obtained by comparing the horizontal pixel coordinates of its first and second endpoints; the difference in vertical pixel coordinates of each speed measuring line is obtained by comparing the vertical pixel coordinates of its first and second endpoints; and the pixel length of each speed measuring line is obtained by comparing the differences in horizontal and vertical pixel coordinates. for: ; The vertical distance from the camera to the water surface during calm water conditions is obtained by comparing the camera's installation elevation with the water surface elevation during calm water conditions. Horizontal distance between each speed measuring line The distance from the center point of each speed measurement line along the direction perpendicular to the water flow to the orthographic projection point of the camera on the water surface during still water is given. The reference oblique distance of each speed measurement line is obtained by combining the vertical distance from the camera to the water surface during still water and the horizontal distance of each speed measurement line. ; By camera focal length Pixel length of each speed measuring line Slant distance from the reference speed measuring line Obtain the reference length of each speed measuring line .

[0037] For example, according to the aforementioned installation parameters ( rice, (frames / s), taking the water surface elevation during the still water period. m (determined by the average of 200 consecutive readings from a frequency-modulated continuous wave radar level gauge during the still water period), specifically explained by obtaining the reference length of each velocity measuring line through camera installation position parameters and camera parameters. This method is to be implemented during the still water period.

[0038] In images captured by cameras during the still water period, each speed measurement line is calibrated using manual operation of calibration software or photogrammetry software. ( The two endpoints along the direction of water flow (i.e., the pixel positions of the starting and ending points of the velocity measurement line in the image) are calibrated to obtain the lateral pixel coordinates of the first endpoint of each velocity measurement line. Vertical pixel coordinates of the first endpoint of each speed measuring line Horizontal pixel coordinates of the second endpoint of each speed measuring line Vertical pixel coordinates of the second endpoint of each speed measuring line (The origin of the coordinate system is taken from the top left corner of the image.) Axis to the right is positive. (Axial direction downwards is positive). Taking a speed camera as an example: Speed ​​camera , , Speed ​​camera , , Speed ​​camera , , Speed ​​camera , , Speed ​​camera , , Endpoint calibration can also be performed by pre-setting reference markers with known three-dimensional coordinates at the endpoints of the speed measurement line, and then obtaining pixel coordinates after identification by image measurement software.

[0039] The pixel length of each speed measuring line is calculated by using the differences between the horizontal pixel coordinates of the first and second endpoints of each speed measuring line, and the differences between the vertical pixel coordinates of the first and second endpoints of each speed measuring line. : ; It should be noted that, The meaning is the pixel distance between the two endpoints of the speed measurement line in the image. For example: Pixels. For example: Pixel.

[0040] Elevation measurement via camera installation Water surface elevation during still water period Obtain the vertical distance from the camera to the water surface during still water conditions It should be noted that, The meaning is the distance from the optical center of the camera to the surface of the water in still water in the vertical direction, i.e., the vertical height difference. In this specific embodiment... Meters. Horizontal distance between each speed measuring line. Defined as: the point on the still water surface from the optical center of the camera (denoted as point). ), along the direction perpendicular to the water flow (i.e., the cross-sectional direction) to the velocity measuring line The horizontal distance to the center point is determined by on-site total station measurement or UAV orthophoto measurement. In this specific implementation method... rice, rice, rice, rice, Meters, entered as parameters after installation. Vertical distance from the camera to the water surface during still water conditions. Horizontal distance from each speed measuring line The baseline oblique distance of each speed measuring line is obtained: ; It should be noted that, The meaning is the distance from the camera's optical center to the speed measuring line. center point The spatial linear distance (oblique distance). The optical center of the camera and... The vertical distance is , and The horizontal distance is The three elements form a right triangle. By the Pythagorean theorem, the oblique distance is... Where the right angle is place ( The optical center of the camera is in the vertical direction. arrive (The direction is horizontal, and the two directions are orthogonal). For example: Rice. With For example: rice.

[0041] By camera focal length Pixel length of each speed measuring line Slant distance from the reference speed measuring line Obtain the reference length of each speed measuring line The principle is based on the pinhole imaging (central perspective projection) model, where the effective projection distance from the optical center of the camera to the center point of the velocity measurement line is... The physical length on the water surface is The speed measuring line, together with the optical center of the camera, forms the base of an isosceles triangle, and the angle subtended by this triangle is... (Radians, approximated by small angles); this angle corresponds to the image sensor. (radians); the two equations are combined and transformed into The accuracy of this similar triangle relationship depends on how well the pinhole camera model approximates the actual camera. After using industrial-grade lenses and camera calibration (removing radial and tangential distortion), the approximate accuracy is acceptable within the range for engineering applications. (Camera focal length) Determined by images obtained from calibration plates in the laboratory or in the field; in this specific embodiment Pixels. For example: Rice. With For example: rice.

[0042] Example 3: Based on Example 2, this example further provides the method for obtaining the current length of each speed measuring line by using the shore water surface elevation corresponding to the current image frame, the camera installation position parameters, and the reference length of each speed measuring line, including: Record the frame number of the current image frame as The vertical distance from the camera to the water surface is obtained by comparing the camera's installation elevation with the corresponding elevation of the water surface on the shore in the current image frame. Data acquisition stops when the elevation of the water surface on the shore corresponding to the current image frame is not lower than the camera's installation elevation; the current oblique distance of each speed measuring line is obtained by combining the current vertical distance from the camera to the water surface with the horizontal distance of each speed measuring line, where the current oblique distance of each speed measuring line is... ; The length correction ratio of each speed measuring line is obtained by comparing the current oblique distance of each speed measuring line with the reference oblique distance of each speed measuring line. ; The current length of each speed measuring line is obtained by comparing its length correction ratio with the reference length of each speed measuring line. .

[0043] For example, according to the aforementioned parameters ( rice, , , (Specifically, this describes the process for each frame of the image during the flood season.) Calculate the current length of each speed measuring line. The method. The elevation of the water surface along the shore corresponding to the current image frame. The implementation details will be provided in the following embodiments.

[0044] Elevation measurement via camera installation Elevation of the shoreline water surface corresponding to the current image frame Obtain the vertical distance from the current camera to the water surface. ;in With the aforementioned The definition is the same (vertical distance from the camera's optical center to the water surface), the difference is that the current water surface elevation is used here. Alternative water surface elevation during still water period Updated with each frame. (Taking the flood season) For example, rice: Meters. Boundary checks are performed on each frame: if the elevation of the water surface along the shore corresponding to the current image frame is... Not lower than the camera installation elevation (Right now , equivalent to If the water level has risen above the camera's installation elevation, it indicates that the projection of the speed measuring line into the camera image is invalid. Immediately stop all acquisition and calculation for the current frame, issue a "water level exceeded" warning, and wait... Restore later. When selecting the installation location, ensure... It is higher than the highest historical flood level and has a margin of no less than 0.5m.

[0045] Vertical distance from the current camera to the water surface Horizontal distance from each speed measuring line (After installation, it remains fixed) to obtain the current diagonal distance of each speed measuring line. ;in and All distances are straight-line distances from the optical center of the camera to the center point of the speed measuring line; only the vertical distance varies depending on the still water period. Switch to the current frame , Unchanged. , For example, rice: Meters; and benchmark values Compared to meters, it is shortened by about This indicates that the distance between the speed measuring line and the camera significantly shortened after the water level rose. : Rice, compared to The length of the speed measuring line is shortened by approximately 7.0%. The closer the speed measuring line is to the camera... The smaller the coefficient, the greater the change in oblique distance caused by water level changes. This is the direct reason for the inconsistent correction range of different speed measurement lines in fixed camera installation scenarios. A single coefficient cannot be used to uniformly correct all speed measurement lines.

[0046] Based on the current oblique distance of each speed measuring line Slant distance from the reference speed measuring line Obtain the correction ratio for the length of each velocity measuring line. ;in The meaning is: compared with the still water period, the speed measurement line in the current frame... The scaling factor based on the oblique distance to the camera. Because... (Still water period) and (The current frame is directly derived from similar triangles) at the same pixel length and focal length Below, the ratio of the two is ,therefore It is also the scaling factor of the current physical length of the speed measuring line relative to the reference length. For example: ;by : When the water level is higher than the still water level When the water level is lower than the still water level Both situations can be handled correctly.

[0047] Correction ratio based on the length of each speed measuring line Relative length of each speed measuring line Get the current length of each speed measuring line ;by For example: Rice; with : Meters. Each speed camera line. Based on the current frame Real-time updates, substituting into the aforementioned speed formula This enables adaptive scale correction for each frame.

[0048] When the camera reads each speed measurement line at a fixed pixel position, changes in the water level cause a corresponding change in the horizontal distance between the pixel ray and the intersection point of the water surface. The current horizontal distance of each speed measurement line should be corrected accordingly. ; After substitution, the current oblique distance, length correction ratio, and current length of each speed measuring line are simplified as follows: , and .

[0049] When the camera is in a fixed physical position (each frame is based on the current water surface elevation) When reprojecting the endpoint coordinates and reading the speed measuring line, Keep the calibration value unchanged during the static water period and use calculate.

[0050] Example 4: Based on Example 1, this example further provides the method for obtaining the floating object determination threshold by repeatedly collecting echo intensity values ​​from a frequency-modulated continuous wave radar level gauge, including: Under conditions of still water and no floating debris, the echo intensity values ​​were continuously collected multiple times using a frequency-modulated continuous wave radar level gauge. The total number of echo intensity value collections was recorded as the echo intensity collection count. The baseline mean echo intensity is obtained by combining all echo intensity values ​​with the number of echo intensity acquisitions. for: ; in For the first The echo intensity values ​​were collected once; the standard deviation of the echo intensity was obtained by comparing all echo intensity values ​​with the mean of the echo intensity reference. for: ; The floating object detection threshold is obtained by using the mean echo intensity, a preset multiplier, and the standard deviation of the echo intensity. ;in The preset multiplier is greater than 1.

[0051] For example, based on the aforementioned installation conditions, the method for determining floating objects is specifically explained in detail during the still water period by obtaining the floating object determination threshold from the echo intensity values ​​collected multiple times by the frequency-modulated continuous wave radar level gauge. This method is implemented under still water conditions with no floating debris before the site is officially operational. Before calibration, visual inspection using camera images or on-site personnel confirms that there are no solid floating objects on the water surface, only a normal water surface condition. Under these conditions, the frequency-modulated continuous wave radar level gauge is driven to continuously collect data at the normal operating frequency. Secondary echo intensity value ( This specific implementation method takes... Sampling interval The total sampling time was approximately 16.7 minutes. Selection requirements: The number of samples must be large enough to minimize the random error (standard error) of the mean estimation. (relative to standard deviation) Small enough to guarantee the threshold The estimate is stable; take When, standard error To meet engineering precision requirements; minimum recommended The value should not be lower than this; otherwise, the estimated variance of the mean and standard deviation will be too large, making the threshold unstable.

[0052] Through all The baseline mean echo intensity is calculated by combining the echo intensity value with the number of echo intensity acquisitions. ; This represents the long-term average of the echo intensity values ​​from a frequency-modulated continuous wave radar level gauge under still water conditions with no floating debris. Its dimensions are... Same; in this specific implementation method volt.

[0053] Through all echo intensity values Compared with the baseline mean of echo intensity Calculate the standard deviation of echo intensity: ; yes The population standard deviation of a sample, in units of , The same applies, reflecting the natural fluctuation amplitude of echo intensity during the still water period; in this specific implementation method volt.

[0054] Based on the mean echo intensity Preset multiplier With echo intensity standard deviation Calculate the floating object determination threshold In this specific implementation, (Preset multiplier is greater than 1) Volts. It should be noted that... The meaning of the value is: Compared to the mean during the still water period higher This ensures that the probability of echo intensity exceeding the threshold is controlled under normal fluctuations during the still water period; If the echo intensity during the still water period follows a normal distribution, the normal fluctuation exceeds The probability of this is approximately 0.3%, meaning the probability of a normal water surface being mistakenly identified as having floating debris is approximately 0.3%; while the false negative rate for floating debris is relatively high due to the increase in echoes caused by typical floating debris. The size determines the difference between the two. Take a real number greater than 1 (including non-integer numbers, such as...). or The specific value is determined during site commissioning based on the following principles: If the echo intensity increment (relative to the mean during the still water period) caused by a typical floating object event is known to be... ,but Should meet To ensure that typical floating objects can be detected; at the same time Ensure that the threshold is higher than the normal fluctuation range.

[0055] Example 5: Based on Example 1, this example further provides a method for determining the time range of floating object influence by performing a current-based estimation of the cross-sectional average flow velocity corresponding to the most recent preceding valid image frame, and by using the floating object feature size and the current-based estimated flow velocity. Floating object characteristic dimensions This is a preset percentile for the size sequence of floating objects along the water flow direction obtained from camera images during historical flood periods; the sequence number of the nearest preceding valid frame in time to the current image frame is denoted as... The average flow velocity of the cross section corresponding to the most recent valid image frame in time is denoted as... The elevation of the water surface along the shore corresponding to the most recent valid image frame in time is recorded as the elevation of the water surface of the previous valid frame. ; Record the frame number of the current image frame as ; by the elevation of the water surface along the shore corresponding to the current image frame , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame ; The duration of the floating object's influence is obtained by estimating the flow velocity using the feature size of the floating object and the current image frame. ; The time range of the floating object's influence is obtained by combining the current image frame acquisition time with the duration of the floating object's influence. The time range of the floating object's influence is a time interval with the current image frame acquisition time as the midpoint and the duration of the floating object's influence as the interval width.

[0056] For example, during flood season, the echo intensity value of the current image frame Not lower than the floating object determination threshold This method involves estimating the average flow velocity across the cross-section of preceding valid image frames and determining the time range of the floating object's influence. Specifically, it involves determining the characteristic size of the floating object. This is a preset percentile for the sequence of floating object dimensions along the water flow direction obtained from camera images during historical flood periods. The operation method is as follows: Identify floating object passage events in the image sequence of several historical flood events. Measure the projected dimension of the largest floating object appearing in each event along the water flow direction (i.e., along the extension direction of the velocity line). During measurement, convert the pixel count to the actual distance using a known reference (such as the ratio of the number of pixels corresponding to the water surface width to the actual water surface width, or by using the reference length of the velocity line). After accumulating no less than 30 samples, take the percentile. percentiles In this specific implementation method Meters. Taking the 85th percentile means that approximately 85% of historical floating debris does not exceed a certain size along the direction of water flow. This makes the estimation of the time range of impact more conservative (preferring to exclude a few more frames rather than introduce flow rate errors in contaminated frames).

[0057] When the site is newly built and has no historical flood image data. Initial setup method: Take Using meters as a general initial value, this value corresponds to the upper quantile size of most single floating events in various types of river channels in mountainous and plain areas, which is a conservative but practically acceptable magnitude; after data is accumulated as the station operates, it will be updated according to the above statistical method. With a larger The consequence of using initial values ​​is only the calculated duration of the effect. The output is slightly larger, which means that a small number of valid frames are excluded, and no erroneous flow rate is introduced into the output. This is a case where a certain safety margin is left.

[0058] During the algorithm's operation during the flood season, the following three state variables are continuously maintained (initially all are null values, indicating that there are no valid frame records yet): the sequence number of the most recent valid frame. (Whenever a frame is marked as valid and the cross-sectional average flow velocity calculation is completed, let) value update assignment to The average cross-sectional velocity corresponding to the preceding valid image frame); (and Synchronous assignment value update assignment to ); the elevation of the water surface along the shore corresponding to the preceding valid image frame. (and Synchronous assignment value update assignment to The three state variables are updated synchronously during valid frames and retain their last assigned values ​​(not cleared) during invalid frames, for use in current estimation during floating object events. Taking a flood event as an example: the frame number of the currently detected floating object... Status Record , meters per second, Meters; Current frame water surface elevation rice.

[0059] The elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame The specific calculation method is described in the subsequent implementation methods. This is only used as an intermediate quantity in calculating the duration of the floating object's influence in this section, and is not used as the final output cross-sectional average flow velocity.

[0060] By the characteristic size of the floating object Estimated flow rate with the current image frame Duration of the effect of floating objects ;in, The meaning is: the dimension along the direction of water flow is Floating objects, at speed The time required to pass through the velocity measurement section, i.e., the duration of the impact of the floating object on the texture extraction of the image frame of that section. For example, For example, meters per second: seconds, which translates to approximately [number of frames] frame.

[0061] Based on the current image frame acquisition time Duration of the impact of floating objects The time range of the floating object's influence is obtained, and this range is... For the midpoint, The time interval with the specified width is: ; by Second, For example, seconds: Within a range of approximately 17 frames, all velocity measurement lines in each image frame are marked as invalid.

[0062] It should be noted that, with The feasibility of taking a symmetrical interval as the midpoint lies in: the moment when the frequency-modulated continuous wave radar level gauge detects an increase in echo intensity. This does not equal the moment when the floating object reaches directly above the speed measurement line. The floating object's journey from entering the radar beam's illumination area upstream to leaving the speed measurement line... This could correspond to the early stage of a floating object entering the irradiation zone (when the object has not yet fully reached the speed measurement line), or it could correspond to when the object is already above the speed measurement line; the actual time the object reaches the speed measurement line relative to... The direction and magnitude of the offset, in The exact moment cannot be determined by a single echo intensity value. A symmetrical interval should be considered. Make the scope of influence Extending both front and back Under conditions where the location of a floating object is uncertain, the time when the floating object might actually reach the velocity measurement line is covered, reducing the probability of contaminated frames being included (or valid frames being incorrectly excluded) due to location uncertainty. The acquisition time... Each frame The speed measuring lines are marked as invalid.

[0063] Example 6: Based on Example 5, this example further provides the elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame In this method, if there is no preceding valid image frame in the current image frame, the average flow velocity in the hydrological station's historical records for the time period corresponding to the current flood season is recorded as the historical reference average flow velocity. The duration of the floating object's influence in the current image frame is obtained by comparing the floating object's feature size with the historical average flow velocity. .

[0064] For example, the handling method when there is no preceding valid image frame in the current image frame is as follows: each time a floating object is detected ( And it is necessary to calculate and estimate the flow velocity. First, check the state variables. Has it already been assigned a value? The algorithm initializes to a null value (represented by a unique identifier, distinct from any valid frame number) upon startup, and only after a frame has completed the calculation of the cross-sectional average flow velocity and been marked as valid. Only then is the frame number assigned. If during inspection... If the value is empty, it is determined that there is no preceding valid image frame in the current image frame, and the process described in this specific embodiment is transferred.

[0065] Under these conditions, the historical average flow velocity of the cross section corresponding to the current flood season in terms of season and magnitude in the historical records of hydrological stations is recorded as the historical reference average flow velocity. . The specific method for obtaining the data is as follows: Consult the multi-year average flow velocity statistics for flood events of similar magnitude and duration in the same month of this website's hydrological yearbook, and take the multi-year average as the data. If the historical records of this station are insufficient, the estimated parameter values ​​of the same hydrological zone in the hydrological manual or the records of neighboring stations during the same period can be referenced. In this specific implementation, if the current flood season is July, the average historical record for the same period is... Meters per second. Based on the size characteristics of the floating object. Compared with historical reference average flow velocity Obtain the duration of the floating object's influence corresponding to the current image frame. ; In this specific embodiment Second. With the aforementioned The units are the same, both representing the estimated time it takes for a floating object to pass through a velocity measurement section; the difference lies in... The flow rate of the most recent valid frame was corrected based on the current water level, resulting in higher accuracy. Historical reference values ​​are used for a guaranteed approximation. replace Substituting into the aforementioned formula for the time range of influence, using the current frame time... For the midpoint, Determine the time range of influence based on the interval width: ; Will Each image frame and each velocity measurement line is marked as invalid, and the process then proceeds to the aforementioned recovery determination process, which is completely consistent with the handling method for cases with preceding valid frames. The accuracy is lower than When the initial When the deviation from the actual flow rate exceeds 50%, the estimation error of the affected area can reach [missing information]. Within ±30%, it is a guaranteed approximation rather than an exact value; as the first valid frame appears and is updated... , , Subsequently, all floating object events were handled using a more precise method. This situation is only used briefly during the initial startup phase.

[0066] Example 7: Based on Example 5, this example further provides that after marking each velocity line of each image frame within the time range of the floating object's influence as invalid, the echo intensity value corresponding to the current image frame is compared with the floating object determination threshold frame by frame for each subsequent image frame. If the echo intensity value of the current image frame is lower than the floating object detection threshold, the current image frame will be counted in the consecutive valid frame count. If the echo intensity value corresponding to the current image frame is not lower than the floating object detection threshold, the consecutive valid frame count is reset to zero and the counting restarts, and the floating object influence time range is redefined; the continuous recovery duration is obtained by combining the consecutive valid frame count with the camera frame rate, where the camera frame rate... The frame rate at which the camera continuously captures images at a fixed frame rate, and the duration of continuous recovery. ; When the duration of continuous recovery exceeds the time window length, each speed measurement line will be remarked as valid.

[0067] For example, according to the aforementioned time range of the floating object's impact. (or After marking all velocity lines in each image frame within the affected time range as invalid, the method for determining the validity of subsequent image frames frame by frame is explained in detail. After marking all velocity lines in each image frame within the affected time range as invalid, for all subsequent image frames whose acquisition time is later than the right endpoint of the affected time range, the following comparison is performed frame by frame: The echo intensity value corresponding to the current image frame is taken. Threshold for determining floating objects A size comparison is performed to determine whether the water surface has been restored to cleanliness and whether floating objects have passed the velocity measurement section. This comparison is executed in each frame, and the comparison logic is the same as the initial discrimination form in the aforementioned flood season start stage, the difference being that it is used here for restoration judgment rather than initial floating object detection.

[0068] If the echo intensity value corresponding to the current image frame Below the floating object detection threshold (Right now The current image frame is counted in the consecutive valid frame count. (each satisfied) The frame makes Add 1, Initially 0). If the echo intensity value corresponding to the current image frame Not lower than the floating object determination threshold (Right now (including equal to), will Clear to 0 and restart counting from the next frame; simultaneously, use the current frame time... Based on this, the time range of the floating object's influence in this new floating object event is redefined using the aforementioned method. Each velocity measurement line in each image frame within the newly determined time range is marked as invalid (if at this time...). If a value has already been assigned, then use it. Estimate, if If it is still empty, then use the aforementioned method. (Method). The time range of the impact of multiple floating object events is handled by taking the union of the results: image frames within each determined time range of impact are marked as invalid, and frames in overlapping parts retain the invalid mark, so as not to cause logical conflicts due to multiple overlays. The purpose of clearing to zero is to ensure that consecutive low echo intensity frames must be re-accumulated from the next frame to guarantee... The duration represented is "the duration of consecutive periods without exceeding the limit since the most recent event of echo intensity exceeding the limit," and counts interrupted in between are not continued. Counting is done through consecutive valid frames. With camera frame rate Calculate the duration of continuous recovery ; in, In this specific embodiment, the frame rate at which the camera continuously captures images at a fixed frame rate is... Frames per second. For example, a frame: The recovery conditions have not yet been met. For example, a frame: seconds, more than Seconds, trigger recovery.

[0069] When the duration of continuous recovery Exceeding the time window length (Right now When the time is right, each speed measurement line will be remarked as valid, and the aforementioned spatiotemporal image speed measurement calculation path will be restored. The meaning is that the time span of the spatiotemporal image is Only when continuous accumulation exceeds Only after a certain number of floating-object-free frames can the next complete spatiotemporal image constructed based on these frames guarantee that all frames are free of floating-object interference, and only then can the texture principal orientation angle extraction results be reliable; if in If the process is not completed in time, several frames adjacent to the area affected by the floating object may still remain in the next spatiotemporal image, causing residual interference to the texture extraction results. In this specific embodiment, Second correspondence Resumption is triggered at frame intervals. After being re-marked as valid, Clear to zero, then return to executing on subsequent frames. and A comparison.

[0070] Example 8: Based on Example 7, this example further provides that before remarking each velocity measurement line as valid, if there are no valid image frames before the current time, the output of the cross-sectional average flow velocity is stopped; otherwise, for the image frames marked as invalid, the cross-sectional average flow velocity corresponding to the most recently marked valid image frame is used as the output of the cross-sectional average flow velocity at the current time; the number of consecutively marked invalid image frames is recorded as the consecutive invalid frame count. The duration of consecutive invalid frames is obtained by counting consecutive invalid frames and the camera frame rate. ; When the duration of continuous invalidity exceeds the preset maximum duration of invalidity. When the average flow velocity of the cross section is stopped, the output will resume after each velocity measurement line is remarked as valid.

[0071] For example, the cross-sectional average flow velocity output strategy and timeout stop output method for each image frame marked as invalid are specifically described before each speed measurement line is remarked as valid (i.e., during the entire period when it is still in an invalid state). Specifically, before each speed measurement line is remarked as valid, the following output decision is performed for each image frame marked as invalid: First, it is checked whether there are any image frames that have been marked as valid before the current time, and the determination is based on the state variable. Has a valid assignment (non-empty value) been obtained? If the data is in a null state (i.e., no valid frame has appeared before the current time), the output of the average cross-sectional velocity will stop, a "data missing" flag will be sent to the external data receiver, and no values ​​will be output externally until the first valid frame appears, at which point output can resume. If a valid frame has already been assigned a value (i.e., a valid frame has appeared before the current time), then for the currently marked invalid image frame, the average flow velocity of the cross section corresponding to the most recently marked valid image frame is used. The current cross-sectional average flow velocity is output externally. A "keep output" quality label is attached to the output to distinguish it from the real-time calculation results of normal valid frames, so that downstream data users know that the value comes from historical valid frames rather than direct measurement of the current frame. This refers to the flow rate of the most recent valid frame recorded by the aforementioned state variables, which remains unchanged during invalid frames.

[0072] The number of consecutively marked invalid image frames is recorded as the consecutive invalid frame count. (Each invalid frame makes) Increment by 1; when the algorithm transitions from an invalid state back to a valid state, i.e., as mentioned above. After triggering recovery and generating the next valid output frame, (Reset). Counting consecutive invalid frames. With camera frame rate Calculate the duration of continuous invalidity ;by frame, For example, frames per second: The time has not exceeded the preset maximum invalid duration. Seconds, continue output .when (Right now When a frame is reached, the output of the average flow velocity of the cross section is stopped, a "data missing" flag is sent to the outside, and the process waits for the aforementioned... After triggering the remarking of each speed camera as valid, the normal output path is restored. Reset to zero.

[0073] and The differences between them are: Tracking "Since the most recent The number of consecutive low echo intensity frames since the event is used to determine whether the recovery conditions are met. Tracking the "cumulative number of frames the current invalid state continues" is used to determine if the duration of maintaining output has exceeded the limit. The reset conditions for the two are different: In any Reset immediately. It is only cleared to zero after returning to a valid state and generating a valid output frame. Within the same invalid period, It may be reset and re-accumulated multiple times (each new floating object event triggers a reset), and It continues to increase monotonically during this period, only returning to zero at the end of the period. Therefore, it always has... (within the same invalid time period).

[0074] It should be noted that, The value should be greater than the typical maximum duration of historical floating object events at the site (to ensure that output does not stop prematurely during normal floating object passage), while not being too large (to prevent old values ​​with decreased output accuracy over a long period of time when floating objects cover the site at high density from misleading flood control decisions). This specific implementation method uses... seconds, suggested value range The duration of each floating object event is determined by combining historical data on the duration of floating object events during site debugging.

[0075] Example 9: Based on Example 5, this example further provides the elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame The methods include: The preset speed measurement section reference elevation is denoted as... The water depth corresponding to the current image frame is obtained by comparing the elevation of the water surface along the shore with the preset velocity measurement section reference elevation. The water depth corresponding to the previous valid frame is obtained by comparing the water surface elevation of the previous valid frame with the preset velocity measurement section reference elevation. ; The water depth ratio is obtained by comparing the water depth corresponding to the current image frame with the water depth corresponding to the previous valid frame. The estimated flow velocity of the current image frame is obtained by using the water depth ratio, the average flow velocity of the cross section corresponding to the previous valid image frame, and the preset water depth-flow velocity relationship index. The current image frame estimates the flow velocity by reflecting the power-law relationship between the average flow velocity at the cross-section and the water depth. .

[0076] For example, this embodiment specifically describes the process of determining the elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Calculate the estimated flow rate for the current image frame The method involves setting the reference elevation of the preset speed measurement section as... , The elevation of the lowest point on the riverbed of the velocity measurement section (or the lowest elevation point on a fixed reference cross section) is determined through on-site leveling measurements and compared with... The elevation datum used is consistent, and this specific implementation method takes... Rice. Special note: It is the elevation of the bottom of the cross-section (measured from bottom to top), while the aforementioned It is the vertical distance from the camera to the water surface (measured from top to bottom), and the two have different meanings.

[0077] The elevation of the water surface along the shore corresponding to the current image frame. Reference elevation of the preset speed measurement section Get the water depth corresponding to the current image frame ;in This indicates the current water depth relative to the reference bottom of the velocity measurement section; when hour (Under normal operating conditions), when hour This is an abnormal situation (the water level drops below the riverbed, which does not exist in reality). When this situation is triggered, the data should be checked for anomalies, and the estimation should be suspended. This specific implementation method takes... Rice, then Meters. Water surface elevation via preceding valid frames. Reference elevation of the preset speed measurement section Obtain the water depth corresponding to the previous valid frame. ; This indicates the water depth at the time of the preceding valid frame. In this specific implementation, we take... Rice, then rice. Selection requirements: It must be lower than the lowest possible water level under all operating conditions (including extremely low water levels during the dry season) to ensure and Hengcheng; it is recommended to take a point at least 0.3 meters below the historical lowest water level of the cross-section as the reference point. To leave room for maneuver.

[0078] Based on the water depth corresponding to the current image frame Water depth corresponding to the previous valid frame Obtain water depth ratio ; For a dimensionless quantity, when The water level rises relative to the previous valid frame (when the water level rises). ), expected flow rate increase; when When the water level falls, the expected flow velocity decreases; when (Right now When the estimated flow velocity is equal to In this specific implementation method By water depth ratio Average flow velocity of the cross section corresponding to the preceding valid image frame Relationship index with preset water depth and flow velocity Estimate the flow rate from the current image frame Preset water depth-velocity relationship index Preset parameters greater than zero This specific implementation method takes .

[0079] It should be noted that, The meaning is: In uniform flow in an open channel, Manning's formula gives the average velocity of the cross-section. ,in For roughness coefficient, For hydraulic radius, For the water surface slope. For a wide and shallow rectangular open channel (width... Much greater than the water depth ), hydraulic radius ; assuming in a short period of time and If approximately invariant, then That is, the relationship between flow velocity and water depth The velocity is directly proportional to the power of the number of times. Therefore, the velocity ratio between the two moments is: ,correspond For non-wide and shallow rectangular cross-sections (trapezoidal, compound, and irregular cross-sections), the relationship between the hydraulic radius and water depth is more complex, and the exponential deviation of the power relationship is greater. The actual situation needs to be determined through on-site calibration. value.

[0080] Specifically, The calibration method involves selecting several pairs of data from the station's historical synchronous water depth-velocity data (such as measured cross-sectional flow and concurrent water level records during historical flood events, or velocity measurement results from valid frames of historical spatiotemporal image velocity measurements and concurrent water level data). ( ,suggestion Then, perform the following steps.

[0081] The first step is to calculate the water depth at each data point: ( ).

[0082] The second step, in Select a reference point from the data points (it is recommended to choose a data point with a moderate water depth), and record its water depth as . Flow velocity is ; Calculate the water depth ratio for each of the remaining data points. Ratio to flow rate (common (Regarding the data).

[0083] The third step is to establish a linear model in the logarithmic field: (No intercept, because when) hour , (This holds true for all cases). Solve using the least squares method. : ; The fourth step is to As a preset water depth-velocity relationship index Save. Typical calibration results: Wide and shallow rectangular open channel Trapezoidal or compound cross-section Irregular mountain river channels It can reach 0.80 to 0.90. If no historical data is available for calibration, then... As an initial value, it is updated as data accumulates. In natural river channels The reasonable range is usually 0.4 to 1.0, with a preset water depth-velocity relationship index. The default parameter is greater than zero. Determined by regression calibration of historical measured water depth and average flow velocity data of the velocity measurement section, reflecting the power-law relationship between the average flow velocity and water depth: for wide and shallow rectangular open channels Pick For trapezoidal or complex cross-sections whose width significantly expands with water depth Less than For deep and narrow cross sections Greater than .

[0084] Substitute the parameters of this specific implementation method into the following: meters per second, , ,calculate Take the natural logarithm multiplied by We get 0.04447, take the exponent. Therefore Meters per second. (This is in contrast to directly using...) Compared to meters per second, the current-level correction increases the estimated flow velocity by approximately 4.5%, corresponding to the duration of the floating object's impact. seconds (compared to using) Directly calculated (The time is shortened by about 0.031 seconds), which in this case means about one less frame is excluded, reducing the loss of valid data.

[0085] Example 10: Based on Example 1, this example further provides a method for making each frame of the image captured by the camera correspond to the elevation of the water surface on the shore and the echo intensity value at the corresponding sampling time in time, including: By inputting a shared time reference signal to both the camera and the frequency-modulated continuous wave (FMCH) radar level gauge, the acquisition times of each image frame from the camera and each sampling time from the FMCH radar level gauge are recorded based on a unified time reference. When the sampling interval of the FMCH radar level gauge is greater than the frame interval of the camera, it will not exceed the current image frame acquisition time. The most recent radar sampling time is recorded as the previous sampling time. The radar sampling time immediately following the previous sampling time is recorded as the next sampling time. The time interpolation coefficients are obtained by using the current image frame acquisition time, the previous sampling time, and the subsequent sampling time. for: ; Using the time interpolation coefficient and the elevation of the water surface at the previous sampling time... Elevation of the shore water surface corresponding to the subsequent sampling time The elevation of the water surface along the shore corresponding to the current image frame is obtained, where the elevation of the water surface along the shore corresponding to the current image frame is obtained. for: ; Using the time interpolation coefficients and the echo intensity value corresponding to the previous sampling time. echo intensity value corresponding to the next sampling time The echo intensity value corresponding to the current image frame is obtained, where the echo intensity value corresponding to the current image frame is... for: ; When the sampling interval of the frequency-modulated continuous wave radar level gauge is no greater than the camera frame interval, each frame of image is directly correlated with the shore water surface elevation and echo intensity value of the most recent radar sampling in time.

[0086] For example, this embodiment specifically describes a method for achieving frame-by-frame correspondence of data collected by two types of sensors. Specifically, by inputting a common time reference signal to the camera and the frequency-modulated continuous wave radar level gauge, the acquisition time of each image frame of the camera and the sampling time of each sampling of the frequency-modulated continuous wave radar level gauge are recorded based on a unified time reference, ensuring that any frame time... With arbitrary radar sampling time Difference on the same time axis The following are possible methods for implementing time reference synchronization: Method 1: GPS timing, where a GPS receiver broadcasts a 1PPS (1 second pulse signal) and a UTC time string to the camera control unit and the frequency-modulated continuous wave radar level gauge data acquisition module, respectively. Both ends record the acquisition time with a GPS timestamp, theoretically achieving an alignment accuracy better than 1ms. Method 2: Network Time Protocol (NTP) or Precision Time Protocol (PTP, conforming to the IEEE 1588 standard), connecting the control units of both types of sensors to the same local area network and using a unified clock server for timing synchronization. NTP typically achieves alignment accuracy in the millisecond range, while PTP can reach sub-millisecond to microsecond levels, both meeting the hydrological measurement requirements for time alignment accuracy (better than 100ms). Method 3: Hardware triggering, sending a frame acquisition trigger pulse to the camera and simultaneously sending a sampling trigger pulse to the frequency-modulated continuous wave radar level gauge sampling module. Both share the same trigger time signal, eliminating time alignment errors, suitable for configurations with unified sampling timing scheduling. Regardless of the method used, the result is the same: the camera's first frame acquisition trigger pulse is sent to the camera, and the second frame acquisition trigger pulse is sent to the frequency-modulated continuous wave radar level gauge sampling module. Both share the same trigger time signal, eliminating time alignment errors, suitable for configurations with unified sampling timing scheduling. Frame acquisition time With frequency modulated continuous wave radar level gauge Second sampling time Differences can be calculated on the same time axis. .

[0087] When the sampling interval of the frequency modulated continuous wave radar water level gauge (In this specific implementation method) (seconds) greater than the camera frame interval At time s, perform time interpolation: the time will not exceed the acquisition time of the current image frame. The most recent radar sampling time is recorded as the previous sampling time. ,Right now ;Will The next radar sampling time that follows is recorded as the subsequent sampling time. ,Right now ;but Based on the current image frame acquisition time Previous sampling time With the next sampling time Calculate the interpolation coefficients at different times: ; It is a dimensionless quantity, reflecting exist The relative position within the interval. Second, Second, seconds (corresponding) Frames per second, After frame 51, For example, (seconds): . This indicates that the current frame time is located approximately 40.8% of the time between two radar sampling times, meaning it is biased towards the previous sampling.

[0088] Interpolation coefficients at time The elevation of the water surface at the shore corresponding to the previous sampling time. Elevation of the shore water surface corresponding to the subsequent sampling time Obtain the elevation of the water surface along the shore corresponding to the current image frame: ; and The frequency-modulated continuous wave radar level gauge is respectively in and The measured water surface elevation reading at any given time is directly output and stored by the radar acquisition module at the sampling time; For the corresponding image frame time The estimated water surface elevation. rice, rice, For example: Meters. Linear interpolation assumes that the water level changes linearly with time between two radar sampling points (i.e., the water level rises or falls uniformly). This assumption is valid during the radar sampling interval. The second-hour rule usually holds true (the rate of change of water level is approximately constant over a short period of time). When the linear approximation error exceeds 10 seconds, a higher-order interpolation method (such as cubic spline interpolation based on multiple radar sample values ​​before and after) can be used. The calculation method refers to the existing technology of numerical analysis, and will not be described in detail in this specific implementation.

[0089] Interpolation coefficients at time Echo intensity value corresponding to the previous sampling time echo intensity value corresponding to the next sampling time Obtain the echo intensity value corresponding to the current image frame: ; and The frequency-modulated continuous wave radar level gauge is respectively in and The measured echo intensity value (in volts) at time t, and , Simultaneous output from radar at the same sampling time; For the corresponding image frame time The estimated echo intensity is used for the aforementioned frame-by-frame floating object discrimination (and...). (Comparison). volt, volt, For example: volt; Volts below Volt, this frame is determined to be free of floating objects, and all velocity measurement lines are marked as valid. For The necessity of interpolation: If interpolation is not performed and data is used directly... Volt correspondence All If the same echo intensity value is used for the frame, the floating object will appear in At any given time, the frame number accuracy for detecting floating objects is only 5 seconds (i.e., ) magnitude, affecting the central moment of the time range The determination error can reach tens of frames, but after interpolation, the error is reduced to the frame level. Second).

[0090] When the sampling interval of the frequency modulated continuous wave radar water level gauge No greater than the camera frame interval (Right now seconds, such as when the radar sampling frequency is not lower than When the frequency is (times / second), there is no need to perform time interpolation; each image frame is directly correlated with the shore water surface elevation and echo intensity values ​​of the most recent radar sampling in time: for image frames (Collection time) ), to find the one that makes it possible to detect the radar sampling time in all radar sampling moments. Minimum sampling Directly ordered , The time alignment error does not exceed Seconds are generally negligible in most hydrological measurement scenarios.

[0091] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 method for acquiring hydrological data through multi-sensor fusion, characterized in that, The method includes: A camera and a frequency-modulated continuous wave radar water level gauge are fixedly installed on the bank foundation on one side of the river cross section. The frequency-modulated continuous wave radar water level gauge synchronously outputs the elevation of the water surface on the bank and the echo intensity value during each sampling, so that each frame of the image captured by the camera corresponds to the elevation of the water surface on the bank and the echo intensity value at the corresponding sampling time in time. During calm water conditions without floating debris, the floating debris determination threshold is obtained by repeatedly collecting echo intensity values ​​from the frequency-modulated continuous wave radar level gauge. The reference length of each velocity line is obtained through the camera installation position parameters and camera parameters. During flood season, the echo intensity value corresponding to the current image frame is compared with the floating debris determination threshold frame by frame. When the echo intensity value is lower than the floating debris determination threshold, each velocity line of the current image frame is marked as valid. When the echo intensity value is not lower than the floating debris determination threshold, the cross-sectional average flow velocity corresponding to the most recent valid image frame is estimated based on the shore water surface elevation corresponding to the current image frame and the shore water surface elevation corresponding to the most recent valid image frame. The floating debris influence time range is determined by the floating debris characteristic size and the estimated flow velocity. Each velocity line of each image frame within the floating debris influence time range is marked as invalid, so that the corresponding image frames are excluded from the cross-sectional average flow velocity calculation. For each image frame marked as valid, the pixel brightness sequence of each velocity measurement line is read from the current image frame. The current length of each velocity measurement line is obtained by using the water surface elevation of the shore corresponding to the current image frame, the camera installation position parameters, and the reference length of each velocity measurement line. The pixel brightness sequences of each velocity measurement line are superimposed along the time dimension to construct a spatiotemporal image. The main direction angle of the texture is extracted from the spatiotemporal image. The water surface velocity of each velocity measurement line is obtained by using the main direction angle of the texture, the current length of each velocity measurement line, and the time window length. The cross-sectional average velocity is obtained by using the water surface velocity of each valid velocity measurement line.

2. The hydrological data acquisition method based on multi-sensor fusion according to claim 1, characterized in that, The method for obtaining the reference length of each speed measuring line through camera installation position parameters and camera parameters includes: In the images captured by the camera during the still water period, the two endpoints of each velocity measurement line along the direction of water flow are calibrated to obtain the lateral pixel coordinates of the first endpoint of each velocity measurement line. Vertical pixel coordinates of the first endpoint of each speed measuring line Horizontal pixel coordinates of the second endpoint of each speed measuring line Vertical pixel coordinates of the second endpoints of each speed measuring line The difference in horizontal pixel coordinates of each speed measuring line is obtained by comparing the horizontal pixel coordinates of its first and second endpoints; the difference in vertical pixel coordinates of each speed measuring line is obtained by comparing the vertical pixel coordinates of its first and second endpoints; and the pixel length of each speed measuring line is obtained by comparing the differences in horizontal and vertical pixel coordinates. for: ； The vertical distance from the camera to the water surface during calm water conditions is obtained by comparing the camera's installation elevation with the water surface elevation during calm water conditions. Horizontal distance between each speed measuring line The distance from the center point of each speed measurement line along the direction perpendicular to the water flow to the orthographic projection point of the camera on the water surface during still water is given. The reference oblique distance of each speed measurement line is obtained by combining the vertical distance from the camera to the water surface during still water and the horizontal distance of each speed measurement line. ; By camera focal length Pixel length of each speed measuring line Slant distance from the reference speed measuring line Obtain the reference length of each speed measuring line .

3. The hydrological data acquisition method based on multi-sensor fusion according to claim 2, characterized in that, The method for obtaining the current length of each speed measuring line by using the shore water surface elevation corresponding to the current image frame, the camera installation position parameters, and the reference length of each speed measuring line includes: Record the frame number of the current image frame as The vertical distance from the camera to the water surface is obtained by comparing the camera's installation elevation with the corresponding elevation of the water surface on the shore in the current image frame. Data acquisition stops when the elevation of the water surface on the shore corresponding to the current image frame is not lower than the camera's installation elevation; the current oblique distance of each speed measuring line is obtained by combining the current vertical distance from the camera to the water surface with the horizontal distance of each speed measuring line, where the current oblique distance of each speed measuring line is... ; The length correction ratio of each speed measuring line is obtained by comparing the current oblique distance of each speed measuring line with the reference oblique distance of each speed measuring line. ; The current length of each speed measuring line is obtained by comparing its length correction ratio with the reference length of each speed measuring line. .

4. The hydrological data acquisition method based on multi-sensor fusion according to claim 1, characterized in that, The method for obtaining the floating object determination threshold by repeatedly collecting echo intensity values ​​from a frequency-modulated continuous wave radar level gauge includes: Under conditions of still water and no floating debris, the echo intensity values ​​were continuously collected multiple times using a frequency-modulated continuous wave radar level gauge. The total number of echo intensity value collections was recorded as the echo intensity collection count. The baseline mean echo intensity is obtained by combining all echo intensity values ​​with the number of echo intensity acquisitions. for: ； in For the first The echo intensity values ​​were collected once; the standard deviation of the echo intensity was obtained by comparing all echo intensity values ​​with the mean of the echo intensity reference. for: ； The floating object detection threshold is obtained by using the mean echo intensity, a preset multiplier, and the standard deviation of the echo intensity. ;in The preset multiplier is greater than 1.

5. The hydrological data acquisition method based on multi-sensor fusion according to claim 1, characterized in that, The method for performing a current-based estimation of the cross-sectional average flow velocity corresponding to the most recent preceding valid image frame, and determining the time range of the floating object's influence by using the floating object's feature size and the current-based estimated flow velocity, includes: Floating object characteristic dimensions This is a preset percentile for the size sequence of floating objects along the water flow direction obtained from camera images during historical flood periods; the sequence number of the nearest preceding valid frame in time to the current image frame is denoted as... The average flow velocity of the cross section corresponding to the most recent valid image frame in time is denoted as... The elevation of the water surface along the shore corresponding to the most recent valid image frame in time is recorded as the elevation of the water surface of the previous valid frame. ; Record the frame number of the current image frame as ; by the elevation of the water surface along the shore corresponding to the current image frame , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame ; The duration of the floating object's influence is obtained by estimating the flow velocity using the feature size of the floating object and the current image frame. ; The time range of the floating object's influence is obtained by combining the current image frame acquisition time with the duration of the floating object's influence. The time range of the floating object's influence is a time interval with the current image frame acquisition time as the midpoint and the duration of the floating object's influence as the interval width.

6. The hydrological data acquisition method based on multi-sensor fusion according to claim 5, characterized in that, The elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame In this method, if there is no preceding valid image frame in the current image frame, the average flow velocity in the hydrological station's historical records for the time period corresponding to the current flood season is recorded as the historical reference average flow velocity. The duration of the floating object's influence in the current image frame is obtained by comparing the floating object's feature size with the historical average flow velocity. .

7. The hydrological data acquisition method based on multi-sensor fusion according to claim 5, characterized in that, After marking each velocity line in each image frame within the time range affected by the floating object as invalid, the echo intensity value corresponding to the current image frame is compared with the floating object determination threshold for each subsequent image frame. If the echo intensity value of the current image frame is lower than the floating object detection threshold, the current image frame will be counted in the consecutive valid frame count. If the echo intensity value corresponding to the current image frame is not lower than the floating object detection threshold, the consecutive valid frame count is reset to zero and the counting restarts, and the floating object influence time range is redefined; the continuous recovery duration is obtained by combining the consecutive valid frame count with the camera frame rate, where the camera frame rate... The frame rate at which the camera continuously captures images at a fixed frame rate, and the duration of continuous recovery. ; When the duration of continuous recovery exceeds the time window length, each speed measurement line will be remarked as valid.

8. The hydrological data acquisition method based on multi-sensor fusion according to claim 7, characterized in that, Before remarking each velocity measurement line as valid, if there are no valid image frames before the current time, the output of the cross-sectional average flow velocity is stopped; otherwise, for image frames marked as invalid, the cross-sectional average flow velocity corresponding to the most recently valid image frame is used as the current cross-sectional average flow velocity output; the number of consecutively marked invalid image frames is recorded as the consecutive invalid frame count. The duration of consecutive invalid frames is obtained by counting consecutive invalid frames and the camera frame rate. ; When the duration of continuous invalidity exceeds the preset maximum duration of invalidity. When the average flow velocity of the cross section is stopped, the output will resume after each velocity measurement line is remarked as valid.

9. The hydrological data acquisition method based on multi-sensor fusion according to claim 5, characterized in that, The elevation of the water surface along the shore corresponding to the current image frame. , Preceding valid frame water surface elevation Average flow velocity of cross section corresponding to the preceding valid image frame Estimate the flow rate from the current image frame The methods include: The preset speed measurement section reference elevation is denoted as... The water depth corresponding to the current image frame is obtained by comparing the elevation of the water surface along the shore with the preset velocity measurement section reference elevation. The water depth corresponding to the previous valid frame is obtained by comparing the water surface elevation of the previous valid frame with the preset velocity measurement section reference elevation. ; The water depth ratio is obtained by comparing the water depth corresponding to the current image frame with the water depth corresponding to the previous valid frame. The estimated flow velocity of the current image frame is obtained by using the water depth ratio, the average flow velocity of the cross section corresponding to the previous valid image frame, and the preset water depth-flow velocity relationship index. The current image frame estimates the flow velocity by reflecting the power-law relationship between the average flow velocity at the cross-section and the water depth. .

10. The hydrological data acquisition method based on multi-sensor fusion according to claim 1, characterized in that, The method for corresponding each frame of the image captured by the camera to the shore water surface elevation and echo intensity value at the corresponding sampling time in time includes: By inputting a shared time reference signal to both the camera and the frequency-modulated continuous wave (FMCH) radar level gauge, the acquisition times of each image frame from the camera and each sampling time from the FMCH radar level gauge are recorded based on a unified time reference. When the sampling interval of the FMCH radar level gauge is greater than the frame interval of the camera, it will not exceed the current image frame acquisition time. The most recent radar sampling time is recorded as the previous sampling time. The radar sampling time immediately following the previous sampling time is recorded as the next sampling time. The time interpolation coefficients are obtained by using the current image frame acquisition time, the previous sampling time, and the subsequent sampling time. for: ； Using the time interpolation coefficient and the elevation of the water surface at the previous sampling time... Elevation of the shore water surface corresponding to the subsequent sampling time The elevation of the water surface along the shore corresponding to the current image frame is obtained, where the elevation of the water surface along the shore corresponding to the current image frame is obtained. for: ； Using the time interpolation coefficients and the echo intensity value corresponding to the previous sampling time. echo intensity value corresponding to the next sampling time The echo intensity value corresponding to the current image frame is obtained, where the echo intensity value corresponding to the current image frame is... for: ； When the sampling interval of the frequency-modulated continuous wave radar level gauge is no greater than the camera frame interval, each frame of image is directly correlated with the shore water surface elevation and echo intensity value of the most recent radar sampling in time.

Images