JP2011257389A - Structure displacement measuring method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a structure displacement measuring method capable of a dynamic measurement as well as a completely contactless measurement of a minute displacement equal to or less than 1 mm from a distant position even for an image type measuring method and capable of estimating a measurement accuracy.SOLUTION: The structure displacement measuring method includes the steps: to perform a dynamic photographing; to compare a measured image with a reference image; to calculate an amount of pixel displacement; and to calculate a displacement by converting a measured displacement represented in the number of pixel into a displacement in a unit of a real scale based on a photographing distance and a photographing angle and to calculate a measurement error as well. Depending on a site condition causing a large measurement error, the method changes a photographing condition. In order to reduce the effect of an atmospheric distortion, the method adapts a limitation in the photographing distance and allows a measurement to be performed by a night photographing. Also in the presence of the atmospheric distortion in measurement data, the method can reduce the effect thereof by filter processing.

Description

  The present invention mainly relates to a structure displacement measuring method for measuring a displacement of a structure that is non-contact and simple and dynamic.

  The deflection of the bridge girder as a structure is an important index for examining the running performance of a train and the soundness of the structure. Conventional bridge deflection measurement can be roughly divided into the following six methods depending on the sensor used. For contact measurement methods, strain gauge measurement method and optical fiber measurement method are used, and for non-contact measurement methods, light wave (laser) measurement method, ultrasonic measurement method, laser Doppler measurement method, image method measurement The method etc. are used (for example, refer patent document 1).

JP 2004-325209 A

  The strain gauge measurement method is a measurement method that is often used because the measurement sensor is inexpensive. However, in river bridges and overpasses that are difficult to enter, displacement gauges are difficult to install, so ensuring safety for workers is an important issue, and dynamic measurement is difficult. In addition, the optical fiber type measurement method can be monitored for a long time, but there are problems that it is necessary to attach a sensor at the time of construction, and that observation after construction is expensive.

  On the other hand, there is a light wave measurement method as a method for measuring the deflection of non-contact type bridge girders, but there are problems such as installation of scaffolding etc., measurement not possible when the bridge is under roads or rivers, and ensuring safety of workers is an important issue. is there. In addition, the ultrasonic measurement method has problems such as low measurement accuracy compared to other methods, difficulty in dynamic measurement, and difficulty in specifying a measurement location. The laser Doppler measurement method is the most excellent measurement method in the non-contact method. However, when the measurement distance is long, there is a problem that a target reflector is necessary and the system price is expensive.

  Image type measurement methods include “a method using a target”, “a method for projecting pattern light”, and “a method using template matching”, but “a method using a target” (eg, Patent Publication 2008-58178) Because it is necessary to install, it is not completely non-contact. The “method of projecting pattern light” (for example, Japanese Patent Publication No. 2011-2378) requires a pattern projector, which is not a simple method and is difficult to project a pattern to a distant place. The present invention belongs to the “method using template matching”. Examples of this method include Japanese Patent Publication No. 2006-329628 and Japanese Patent Publication No. 2006-343160, which are based on the assumption that neither the camera nor the atmosphere fluctuates, and when photographing from a distance using a high magnification telephoto lens. Can not respond. In addition, when shooting on a large structure with an enlarged local area, it is difficult to provide a non-contact size reference within the shooting range, and the method using only the camera can convert pixels to actual size. I can't do it.

  The present invention has been developed to solve the above-described problems, and an object of the present invention is to provide a structure displacement amount measuring method capable of easily and accurately measuring the displacement amount of a structure without contact. It is said.

  In order to solve the above-described problem, the structural displacement measurement method according to the present invention compares the reference image and the measurement image, calculates the pixel displacement, and calculates the actual displacement from the information on the shooting distance and shooting angle. This is a completely non-contact structure displacement measuring method.

  Further, the structure displacement measuring method according to the present invention includes a first step of capturing a reference image and a measurement image by a photographing device, and obtaining information of a photographing distance and a photographing angle by the measuring device, and the reference photographed above. A second step of calculating the pixel displacement amount by comparing the image with the measurement image, a third step of converting the pixel displacement amount into an actual displacement amount, and a fourth step of converting the image sequence number into real time. Step, a fifth step for estimating a position where the amount of change is 0 from the data converted in the third step and the fourth step, and data converted in the third step and the fourth step A sixth step of detecting and removing an abnormal value from the seventh step, a seventh step of determining a displacement amount from the converted data corrected by the fifth step and the sixth step, the fifth step and the fifth step 6 steps Characterized by comprising an eighth step of calculating a measurement error from the corrected transformed data variation 0 interval.

  The present invention can provide a structure displacement amount measuring method that can easily and accurately measure the displacement amount of a structure without contact.

It is a figure which shows the basic flowchart of the structure dynamic displacement measuring method which concerns on this invention. It is a figure which shows the flowchart of the pre-processing operation | work of the measurement imaging | photography of FIG. It is a block diagram which shows the structural example which concerns on the structure dynamic displacement measuring method of this invention. It is a figure which shows the standard deviation of a still range in case the influence of atmospheric fluctuation is large. It is a figure which shows the effect of a smoothing filter. It is a figure which shows the relationship between the standard deviation of a still range, and optimal filter size. It is a figure which shows the example of the exclusive scale used for the applicability determination process of a measurement. It is an outline figure of photography of a target bridge, and distance and angle measurement. It is a figure which shows the flowchart of displacement amount calculation of FIG. It is a figure which shows calibration of imaging distance. It is a figure which shows the focal distance correction | amendment of a camera.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The drawings relating to the structure are schematic and different from the actual ones. Moreover, the part from which the relationship and ratio of a mutual dimension differ also in between drawings is contained.

  First, an outline will be described. In the structure dynamic displacement measuring method in this embodiment, dynamic imaging is performed, a reference image and a measurement image are compared, and a pixel displacement is calculated. Then, the actual displacement amount is calculated from the information on the shooting distance and shooting angle. The actual displacement amount is calculated by converting the displacement amount expressed by the number of pixels into a unit of an actual scale. In addition, the actual measurement error is calculated. The actual measurement error is an actually measured error and is different from a measurement estimation error described later. If the actual measurement error is large due to local conditions, change the shooting conditions. In order to mitigate the effects of atmospheric fluctuations, a shooting distance constraint was introduced. In addition, when shooting must be performed at a distance, measurement by night shooting with small atmospheric fluctuations is possible. If the measurement data is affected by atmospheric fluctuations, it can be reduced by filtering.

  As an example of a method for easily realizing such a structural dynamic displacement measurement method, non-contact bridge deflection measurement using a consumer digital camera is proposed. In this method, the amount of movement between frames of a series of images obtained by continuously photographing a bridge at a high magnification is automatically calculated in sub-pixel order by image processing, and the actual displacement is calculated based on the distance measured by the laser and the focal length. Convert. A subpixel is a unit smaller than one pixel. By calibrating the combination of the camera and lens to be photographed in advance, the influence of the change in focal length due to focus adjustment is removed, and the actual displacement is calculated with high accuracy. Hereinafter, it explains in detail using a drawing.

  FIG. 1 shows the main processing steps of the structural dynamic displacement measuring method of the present invention. First, as shown in S1, preprocessing work for measurement imaging is performed. As shown in FIG. 2, the pre-processing work is roughly divided into a pre-measurement-photographing simulation work (S11 to S16, S19) and a pre-measurement-photographing site confirmation work (S17 to S18, S19).

  As shown in FIG. 3, the system for realizing such a structure dynamic displacement measuring method includes an imaging device 1, an arithmetic processing device 2 such as a portable personal computer or a server, and a measurement applicability determination unit 3. ing. The arithmetic processing device 2 includes an input unit 21, a registration unit 22, a dynamic displacement amount measurement unit 23, an output unit 24, a storage unit 25, a generated displacement amount prediction unit 26, an imaging setting condition calculation unit 27, a displacement calculation unit 28, an abnormality A value detection / removal unit 29 and a measurement accuracy estimation unit 30 are included.

  Hereinafter, the simulation work before measurement photographing will be described. In this work, first, target bridge registration (S11) is performed. Here, the name of the bridge for which the deflection is measured, the bridge structure conditions (length between spans, structure (single girder, continuous girder), material (metal / concrete)), passing train conditions (vehicle weight, vehicle length, axle length, number of vehicles) , Train speed, estimated occupancy rate) are input from the input unit 21, and the target bridge information is registered in the registration unit 22 and stored in the storage unit 25.

  Next, the generated displacement amount prediction (S12) is performed. The generated displacement amount prediction unit 26 predicts and calculates the maximum deflection amount, the deflection waveform, and the maximum deflection frequency from the bridge structure condition and the passing train condition by a simulation method.

  Next, the target measurement accuracy is determined (S13). Based on the result predicted in S12, the target accuracy of the deflection measurement of the target bridge is determined. Next, photographing condition setting (S14) is performed. In S14, the deflection measurement photographing conditions (digital camera, tripod (structure / weight), photographing distance, photographing mode (continuous shooting or moving image)) are set in consideration of the surrounding situation of the measurement target bridge. Thereafter, a measurement estimation error is calculated (S15).

It is the influence of atmospheric fluctuations that makes measurement estimation errors difficult to determine. The effects of atmospheric fluctuations are not only nonlinear, but also vary depending on weather conditions and locations. For this reason, an atmospheric fluctuation term should be provided in the measurement estimation error equation for evaluation. Error term (A × K × D) due to external force received by the photographing apparatus, error term due to photographed image (P × s / f × D), error term due to atmospheric fluctuation (B × D ^1.5 ) And other error terms (C), the simplest model is an equation in which terms such as the following equations are linearly combined.

  Here, A is a different coefficient depending on the environment, K is a coefficient specific to each equipment such as a tripod, D (m) is a shooting distance, P is a subpixel coefficient, s (μm) is a sensor size, f (mm) is a focal length, B is a coefficient related to the magnitude of atmospheric fluctuation during shooting, and C (mm) is the other error.

  First, in order to judge the effects of atmospheric fluctuations, the effects of large atmospheric fluctuations were investigated. FIG. 4 shows a standard error of measurement data having a large atmospheric fluctuation and a graph when the coefficient B is determined to express them using only the atmospheric fluctuation term. For reference, the estimation accuracy obtained by adding the first term and the second term is indicated by a dotted line.

  Next, a method for estimating the subpixel coefficient P will be described. Here, the local environmental conditions were kept constant, and the measurement was not affected by the first term, and the accuracy of the subpixel coefficient P was estimated. We also examined how the focal length, shooting mode, subject brightness, and magnitude of change affect the subpixel coefficient P. The steps 1 to 4 of the estimation method are shown below.

1. While keeping the environmental conditions constant, (1) Fluctuating objects and (2) Still objects are photographed, and (1)-(2) are calculated to obtain measurement data that captures only fluctuations.

2.1. Perform multiple measurements for the same variation in the same environment. The measurement data for a plurality of times is obtained by the same processing as.

3. The measurement accuracy is estimated based on the variation between multiple measurement data.

4). The value of P is obtained by dividing the measurement accuracy by the spatial resolution (= s / f × D).

By the way, when the overall accuracy when measuring the object i is σi, the accuracy corresponding to the first term is σ _i1 , and the accuracy of the second term is σ _i2 ,

It becomes. Measurement data that captures only fluctuations is canceled by σ _i1, so that only the effect of σ _i2 remains. If the accuracy of the measurement data that captures only the fluctuation is σ _m , the variable object is i = 1, and the stationary object is i = 2, the error propagation law

It becomes. If the stationary object and the changed object have the same accuracy (omitting i), σ ₂ = σ _m / √2.

Now, it is assumed that the difference between the first measurement and the second measurement is obtained, and σ _d is obtained as the mean square error. If there is no difference in accuracy between the first measurement and the second measurement, the error propagation law is applied in the same manner as Equation 3 to obtain σ _m = σ _d / √2. Therefore, the accuracy of the second term is obtained with σ ₂ = σ _d / 2. Therefore,

As a result, a sub-pixel coefficient is calculated. When measuring three or more times, the average value of σ _d is used. In addition, when a graph is obtained when shooting under good conditions, the mean square error with that graph is obtained, and the value of σ ₂ when shooting under good conditions is used.

Can be calculated as

  Next, a countermeasure when atmospheric fluctuation occurs will be described.

  The influence of atmospheric fluctuation can be alleviated to some extent by smoothing in the time direction and the spatial direction. However, it is not appropriate to largely smooth in the time direction in order to correctly grasp the deflection variation. On the other hand, if the matching area is made too large for smoothing in the spatial direction, there is a high possibility that the matching itself will fail. In addition, the processing time for matching also increases. Therefore, a method of smoothing a graph obtained by taking a plurality of small regions for matching can be considered.

  Next, a method for reducing atmospheric fluctuation using a filter will be described.

  FIG. 5 is a graph obtained by superimposing a measurement graph (solid line) obtained by image measurement and a graph of actual measurement values (dotted line) obtained by the past measurement system. By increasing the value of the filter, the measurement graph (solid line) becomes smoother and closer to the graph (dotted line) of the previous measurement system. This is an example in which the effect of the filter appears well.

When using a filter in this way, it becomes a problem how much the filter size should be set. Therefore, an optimum filter size was investigated by applying filters of various sizes. As a result, as an overall rough tendency, as shown in FIG. 6, it was found that when the standard deviation is small, the filter size changes suddenly and gradually changes gradually. Examples of model equations showing such a tendency include a power function that takes an inverse number and a logarithm (log). Since the logarithm is the most applicable to the survey data this time, we decided to calculate the coefficient using the following model formula.

  Next, formulation of measurement errors during the season and night will be described.

  The first and second terms of Equation 1 (basic formula) are proportional to the shooting distance D, but the third term has a large non-linear effect depending on the shooting distance. When the atmospheric fluctuation is sufficiently small, it is hidden by the effects of the first and second terms, and no nonlinear effect appears until the shooting distance becomes considerably large. The effect appears.

  Here, since the influence of atmospheric fluctuations varies greatly depending on the shooting time, time zone, and atmospheric conditions, a restriction on the shooting distance that reduces the influence of atmospheric fluctuations is introduced. If a measurement estimation error is calculated using an estimation formula that eliminates terms due to atmospheric fluctuations, summer measurements and nighttime measurements can be performed.

  Next, on-site confirmation work before measurement photographing will be described. As field confirmation work before measurement photographing, field photographing (S17) is performed, and applicability determination (S18) is performed. Before taking a bridge deflection measurement using a photographic equipment, in order to check whether it is in a state suitable for deflection measurement photography such as the surrounding situation of the shooting location and weather conditions, Visually check the amount of fluctuation in the stationary state under the same conditions. Specifically, using a dedicated scale sheet attached to the liquid crystal monitor of the digital camera, it is determined in S18 whether the fluctuation of the stationary object is within the deflection measurement application range.

  The measurement accuracy of the imaging method varies greatly due to the influence of weather conditions such as on-site weather and wind, and environmental conditions such as vibrations caused by automobiles. For this reason, we have realized a mechanism to determine whether the image measurement method can be applied locally. In order to determine whether or not the image type measurement method is applicable, first, it is necessary to confirm whether a tripod, a camera, or a lens is bent, loose, or loose by applying force. Secondly, it is to confirm whether the string of the stone bag is stretched or grounded. Third, check the effects of wind, vibration, and atmospheric fluctuations. Fourthly, the state of exposure or the like of the measurement target portion is checked on the liquid crystal monitor to confirm whether or not the photographing is possible (shutter speed, aperture, ISO sensitivity).

  After confirming the above items, check with the dedicated scale sheet. Here, the dedicated scale will be described. In order to measure the displacement of a measurement object that fluctuates dynamically with a camera, 1) Consider the effects of vibrations on the camera itself, and 2) Consider the effects of the atmosphere between the camera and the measurement object. There is a need. Therefore, it is necessary to estimate a measurement error by monitoring a stationary object before measurement and to confirm that the measurement estimation error is within an allowable range. Since the effects of 1) and 2) are affected by the circumstances around the measurement position, it is desirable to be able to monitor immediately before the measurement at the site.

  As one of the methods, it is conceivable to evaluate the measurement estimation error by using a liquid crystal monitor on the back of the camera and observing at a maximum magnification. However, since the camera's LCD monitor does not have a function for measuring the amount of variation, an error amount can be estimated quantitatively by attaching a dedicated scale. Here, the liquid crystal monitor of the camera and the dedicated scale correspond to the measurement application determination unit 3.

  An allowable error range (hereinafter referred to as an allowable error range) is defined by a parameter K determined for each photographic equipment. By using a dedicated scale with a scale indicating the allowable error range when the parameter K is applied, it is possible to determine whether the environment being measured in the field is within the allowable error range. Simple and immediate judgment can be made without using an apparatus.

  Of the measurement estimation errors of Equation 1, D × A × K in the first term represents an error that is an environmental factor. A dedicated scale is used to determine the effect of this first term. Now, the magnitude δ of the error on the object caused by the camera shake of d pixels is given by the following equation.

δ = D / f × S × d
Therefore, the number of pixels d corresponding to an error of D × A × K can be calculated from the following equation.

D × A × K = D / f × S × d
Therefore
d = A × K × f / S
The photographing equipment safety factor coefficient K and the pixel size S are fixed values for each camera, and if the focal length of the lens is maximized, an approximate value of the focal length f is also determined. That is, if the setting at the time of shooting is determined, the number of pixels d corresponding to the measurement error becomes a constant value regardless of the shooting distance D.

  By the way, the resolution of the camera sensor and the resolution of the liquid crystal monitor do not match. In fact, the number of LCD monitor elements is much smaller than the number of camera sensor elements. Therefore, it has an enlargement function for displaying the details of the subject. There are roughly two types of methods, a method of changing the display magnification of a monitor adopted in many single-lens reflex cameras, and a method of using a digital zoom adopted in many compact cameras.

  For example, if the number of image pixels at the time of shooting is W × H and the number of dots on the liquid crystal monitor is w × h, it is only necessary to display at a magnification of H / h or higher because the target is vertical displacement. If the maximum display magnification is m, m × h / H is the number of display dots per pixel. Further, when the vertical width (mm) of one dot of the liquid crystal monitor is s, the fluctuation width δ (mm) on the liquid crystal monitor corresponding to the allowable error range can be calculated by the following formula.

δ = A × K × f / S × m × h / H × s
An example of a dedicated scale is shown in FIG.

  We changed the camera type and the equipment safety factor coefficient (tripod type), attached a special scale sheet to the liquid crystal panel, set the liquid crystal panel conditions, and observed a stationary object. If it is within the measurement application range, the process proceeds to S2. If it is not within the measurement application range, the shooting conditions are changed in consideration of environmental conditions (vibration, wind, atmospheric fluctuation, etc.), and the process proceeds to S14.

  Here, in addition to the information registered in S11, track name, shooting position sketch (GPS observation value), measurement location, measurement shooting date, weather condition (weather, humidity, wind), shooting distance, shooting operator name, analysis work Information such as a person's name is registered in the registration unit 22 using the input unit 21 or the like, and photographing related data is captured.

  When the process proceeds to S2 within the measurement application range, the target bridge measurement photographing is performed. That is, based on the determined shooting conditions, the deflection measurement shooting of the target bridge is performed at the site. Measurement shooting is performed continuously (dynamic shooting) from before the train passes until the train passes. As shown in FIG. 8, continuous shooting or moving image shooting is performed from the side of the center of the bridge 50 with the digital camera 11. Shooting is performed including several seconds before passing and several seconds after passing. The photographing equipment, frame rate (F), and focal length (f) are determined in consideration of the expected amount of deflection, the required measurement accuracy, the length of the digits, and the train speed. Further, the distance (D) from the camera to the imaging region of the bridge and the inclination angle (Φ) from the imaging position are measured with a handheld laser distance meter 12 or the like.

  Next, the dynamically captured image data is stored in the storage unit 25. A processed image is selected from the stored data (S3), and the captured image is determined (S4). In S4, it is determined whether the image taken at the time of deflection measurement is continuous shooting (still image) or moving image. In the case of a still image, the process proceeds to S6, and in the case of a moving image, conversion from a moving image to a still image (S5), that is, conversion into a still image for each frame is performed. These processes are performed by the dynamic displacement measuring unit 23.

  In S6, the maximum displacement is finally calculated, and the detailed procedure is shown in FIG. First, a reference image (for example, an image at no load) is compared with a measurement image (for example, an image at the time of passing a train), and further, a predetermined process is performed using the reference image as a template and the measurement image as a search plate.

  First, one of the images photographed before passing is set as a reference image R, and time-series image data of a portion that can be regarded as moving uniformly due to deflection is defined as Cn (n: number of captured images). A time-series movement amount Δypix in the vertical direction between the reference image R and another series of images Cn is automatically measured in units of pixels. A least square matching algorithm is applied as a method for automatically and stably calculating the measurement in sub-pixels. Luminance correction is performed so that the luminance difference is minimized by least square matching (S61), change amount measurement is performed in units of pixels (S62), and graphing is performed (S63). From the graphed data, the number of pixels on the vertical axis is converted into a deflection amount (mm), and the number of shots on the horizontal axis is converted into elapsed time (seconds) (S64).

(1) Relationship between image coordinates (pixels) and photo coordinates (mm) When the photo image coordinates are (u, v), the photo coordinates (x , y) is expressed as follows when the size of one element of the sensor is s (mm / pixel) and the number of pixels is w × h.

(2) Relationship between camera coordinate system and target coordinate system When the focal length of the camera is f, θ is the horizontal rotation amount of the optical axis direction of the camera with respect to the train traveling direction (Z) of the bridge girder, and the camera tilt angle When φ is φ, the relationship between the camera coordinate system (x, y, -f) and the target coordinate system (X, Y, Z) that represents the measurement point of the bridge girder with the same coordinate origin as the camera coordinate system is as follows: It is expressed as follows.

  Here, ΔX and ΔY are a roll amount and a deflection amount at the measurement point (X, Y, Z), respectively. Δx and Δy are the amount of change on the photograph due to deflection and roll. λ is a scale variable. It was assumed that there was no change in the Z direction. The rotation amount around the optical axis of the camera is κ, and it is assumed that the bridge girder is not tilted.

Now, it is sufficient to use the value of the inclinometer mounted on the distance meter as it is, but θ and κ are obtained from the image. First, κ is searched for an edge that is directed in the vertical direction near the center of the image. Since it is a telephoto shooting, assuming that the scale factor λ does not change with small changes near the center, using the edge vector (dx, dy) on the image and the vector (0, dY, 0) on the object ,

It can be expressed as From this formula:
κ = arctan (-dx / dy)
Κ can be obtained as Next, using the vector (dx ', dy') on the image of the edge in the horizontal direction and the vector (0, 0, dZ) on the object,

Θ is solved from the following relational expression.

θ = arctan (sinφ (dx'cosκ + dy'sinκ) / (-dx'sinκ + dy'cosκ))
In the following, for the sake of simplicity, the discussion proceeds with κ = 0.

(3) Simplified calculation of deflection amount First, for the sake of simplicity, consider the case of shooting from the front, that is, θ = π / 2. At this time, can be modified as follows.

  Now, when the center (x, y) = (0, 0) of the photograph is taken as a measurement point, Z = 0, Ycosφ−Xsinφ = 0, that is, Y = Xtanφ holds when there is no deflection or roll.

At this time, if it is assumed that only deflection (= ΔY) has occurred, Δx = 0
Δy ≒ fΔYcosφ / X (cosφ + tanφsinφ)
Because it becomes a simple relationship like
ΔY = ΔyX (cosφ + tanφsinφ) / fcosφ
The amount of deflection can be calculated as If φ is small, tanφsinφ≈0, so ΔY = ΔyX / f may be set. Incidentally, if the approximate range is ± 1%, that is, 0.99 <(cosφ + tanφsinφ) / cosφ <1.01, −5.7 ° <φ <5.7 °.

  By the way, at the time of measurement, the horizontal distance D to the object near the center of the photograph is measured. If this horizontal distance D is used, the amount of deflection can be easily calculated without having to calculate θ.

  The relationship between the target coordinates X and Z in the vicinity of (x, y) = (0, 0) and the horizontal distance D is expressed as follows.

X = D sinθ, Z = -D cosθ
by the way,

Therefore, the above equation can be rewritten as follows using D in the vicinity of the center of the photo.

From this, when Y is calculated when there is no fluctuation, Y = D tanφ. Therefore, if it is assumed that there is no roll ΔX, ΔY = ΔyD / f in the range of -5.7 ° <φ <5.7 °
However, it is possible to calculate an almost accurate deflection amount. If the tilt angle exceeds this range, ΔY = ΔyD (cosφ + tanφsinφ) / fcosφ
It is sufficient to calculate using It should be noted that this formula can be applied only when the object appears near the center of the photograph and the horizontal distance is approximately D.

(4) Exact calculation If you want to calculate more strictly, you can solve the equation shown in (2) above in reverse. That is, the following equation is solved.

First, it is necessary to find the target coordinates (X, Y, Z) without fluctuation, but the number of expressions is 3, whereas the number of variables is 4 when λ is included. So you can't solve this situation. Therefore, it is assumed that the surface of the bridge girder to be measured is a substantially vertical plane. In this case, X is constant. Using the horizontal distance D to the approximate center of the photo, as already mentioned, X = D sinθ. Substituting this into the above equation as a variation amount of 0 solves Y and Z as follows.

Substituting X, Y, and Z obtained here into the formula, the following formula is finally obtained.

  By using this equation, not only the vertical deflection amount ΔY but also the roll amount ΔX can be calculated. In addition, the amount of deflection can be obtained by measuring a place that is off the center of the photograph.

  Based on the above theory, the displacement amount in pixel units is converted into the vertical displacement amount (deflection amount) in mm units, and the photograph sequence number (frame) is converted into the elapsed shooting time (in seconds).

Deflection formula ΔY = Δy × D × (cosφ + tanφsinφ) / fcosφ
Elapsed time conversion formula t = n / F
Here, before conversion by the above conversion formula, it is necessary to perform shooting distance calibration processing and focal length correction processing as shown in S69. As shown in FIG. 10, a distance meter is attached to the digital camera. ΔD is the difference between the measured value D 'measured by the distance meter and the shooting distance D. The focal length f is fixed, and multiple shots are taken at different shooting distances within the focus range. To do.

  When performing deflection measurement shooting at a close distance with AF (autofocus) enabled, the subject is shot in advance and the nominal focal length is calculated using an approximate expression calculated from the relationship between the shooting distance and the focal length. Perform the correction. An example is shown in FIG. The vertical axis represents the difference between the nominal focal length and the actual focal length, and the horizontal axis represents the shooting distance. When the shooting distance becomes smaller than a certain value, the difference in focal length increases rapidly. Using these data, focal length correction processing for the shooting distance is performed.

  Next, the position of the deflection amount 0 is estimated (S65). A section in which no deflection occurs (no-load section, section before passing train) is calculated, the average value of the section is calculated, the average value is subtracted from the entire graph, and the position of the deflection amount 0 is estimated. As mentioned above, each process of S61-S65, S69 is performed by the displacement calculation part 28. FIG.

  In S66, it is determined whether there is an abnormal value. For example, an abnormal value is determined by a median filter. Specifically, data that can be clearly determined to be abnormal is data that is either very large or small compared to adjacent data. Therefore, we decided to cut out abnormal values using a median filter.

  The median filter is a filter that takes an intermediate value. For example, when there is a series such as {1 15 2}, the intermediate value is a value located in the middle when sorting in ascending order like {1 2 15}. In this case, 2 is the intermediate value It corresponds to.

  Specifically, the movement amount ΔY at a certain time is the intermediate value of the movement amount of the front and rear three sheets (a value located in the middle when arranged from the maximum to the minimum, the median value) Δymed, and the threshold for abnormal value determination is Th , (ΔYmed−ΔY) ≧ Th, it is determined as an abnormal value (S66). If it is determined as an abnormal value, the abnormal value is corrected and removed using the following equation (S67).

If (ΔYmed-ΔY) ≥ Th ΔY = ΔYmed
If (ΔYmed-ΔY) <Th ΔY = ΔY
Then, the displacement amount (maximum deflection amount) is output (S68). As described above, the abnormal value detection / removal unit 29 executes S66 to S68.

  Next, the measurement accuracy estimation unit 30 calculates the standard deviation of the displacement in the section when the deflection amount 0 is estimated, and estimates the measurement accuracy. That is, the standard deviation of the displacement amount in the stationary state (deflection 0) before passing through the train is calculated, and the measurement accuracy of the measured deflection amount when passing through the train is estimated from this value (S7). Finally, information such as the maximum deflection amount, deflection waveform, estimated measurement accuracy, measurement location, measurement object, etc. of the deflection measurement result of the target bridge is output by the output unit 24 (S8).

  As described above, according to the structure displacement amount measuring method of the present invention, the displacement amount of the structure can be measured easily and accurately with complete non-contact. That is, even with the image-type measurement method, dynamic measurement and measurement of a minute deflection amount can be performed, and measurement accuracy can be estimated. Furthermore, high-precision measurement can be performed over a wide range of shooting from a short distance to a long distance, and the measurement accuracy can be made comparable even if the local conditions are different.

  Here, the case where the train passes through the bridge has been described as an example, but the present invention is not limited to this. That is, the displacement amount can be measured by a similar method even for structures other than bridges.

In order to solve the above-described problem, a structure displacement measuring method according to the present invention is a method of shooting a moving image or a continuous shot image after fixing a shooting device on a movable tripod, and out of the shot measurement images. A non-contact structure displacement amount that calculates the actual displacement amount from the information of the photographing distance and the photographing angle after comparing the reference image with the measurement image and calculating the pixel displacement amount. What measurement methods der, to calculate the pre-estimation error based on the performance of the environmental factors and photographic equipment, including air fluctuation, after the initial set the shooting conditions on the basis of the pre-estimation error, the actual measured in the field The imaging condition is reset based on the post-measurement error .

First, an outline will be described. In the structure dynamic displacement measuring method in this embodiment, dynamic imaging is performed, a reference image and a measurement image are compared, and a pixel displacement is calculated. Then, the actual displacement amount is calculated from the information on the shooting distance and shooting angle. The actual displacement amount is calculated by converting the displacement amount expressed by the number of pixels into a unit of an actual scale. In addition, the actual measurement error (post-measurement error) is also calculated. The actual measurement error is an actually measured error and is different from a measurement estimation error (preliminary measurement error) described later. If the actual measurement error is large due to local conditions, change the shooting conditions. In order to mitigate the effects of atmospheric fluctuations, a shooting distance constraint was introduced. In addition, when shooting must be performed at a distance, measurement by night shooting with small atmospheric fluctuations is possible. If the measurement data is affected by atmospheric fluctuations, it can be reduced by filtering.

Claims (12)

  A completely non-contact structure displacement measurement method that compares a reference image and a measurement image, calculates a pixel displacement, and calculates an actual displacement from information on an imaging distance and an imaging angle. A first step of capturing a reference image and a measurement image with an imaging device, and obtaining information of an imaging distance and an imaging angle with the measurement device;
A second step of calculating a pixel displacement amount by comparing the reference image captured in the above and the measurement image;
A third step of converting the pixel displacement amount into an actual displacement amount;
A fourth step of converting the image sequence number into real time;
A fifth step of estimating a position where the amount of change is 0 from the data converted in the third step and the fourth step;
A sixth step of detecting and removing abnormal values from the data converted in the third step and the fourth step;
A seventh step of determining a displacement amount from the conversion data corrected in the fifth step and the sixth step;
A structure displacement amount measuring method comprising: an eighth step of calculating a measurement error from a section 0 of change amount of conversion data corrected in the fifth step and the sixth step. In the eighth step,
3. The structural displacement amount measuring method according to claim 2, wherein an actual measurement error is calculated according to local conditions. Before shooting in the first step,
3. The structural displacement measuring method according to claim 2, wherein the influence of local conditions is measured using a dedicated scale attached to the photographing apparatus, and when the influence is large, the photographing conditions are changed. Error term (A × K × D) due to external force received by the photographing apparatus, error term due to photographed image (P × s / f × D), error term due to atmospheric fluctuation (B × D ^1.5 3) The structure displacement amount measuring method according to claim 2, wherein when the other error term (C) is given, the measurement estimation error is calculated by the following equation.
Measurement estimation error σ = A × K × D + P × s / f × D + B × D ^1.5 + C
A coefficient that varies depending on the environment, K is a coefficient specific to the equipment, D is a shooting distance, P is a sub-pixel coefficient, s is a sensor size, f is a focal length, and B is a coefficient related to the magnitude of atmospheric fluctuation during shooting. .   6. The structure displacement amount measuring method according to claim 5, wherein an imaging condition is set according to the measurement estimation error. In the calibration of the imaging device,
The structure displacement amount measuring method according to claim 6, wherein the sub-pixel coefficient P is calculated by photographing an object that fluctuates simultaneously and an object that is stationary.   Simultaneously measure a stationary object close to the measurement object, remove the high-frequency component of atmospheric fluctuation, and reduce the influence of the atmospheric fluctuation by subtracting the measurement result of the stationary object from the measurement result of the measurement object. The structure displacement measuring method according to claim 6, wherein:   7. The structure displacement measuring method according to claim 6, wherein the influence of the atmospheric fluctuation is reduced by applying a filter according to the magnitude of the atmospheric fluctuation.   7. The measurement estimation error is calculated by an estimation formula in which a term due to the atmospheric fluctuation of the measurement estimation error is removed by introducing a restriction of an imaging distance that reduces the influence of the atmospheric fluctuation. Method for measuring the displacement of structures.   7. The structural displacement measuring method according to claim 6, wherein measurement at night is possible in order to reduce the influence of atmospheric fluctuation.   The structure displacement amount measuring method according to claim 1, wherein the reference image and the measurement image are obtained by using a photographing apparatus in which a focal length correction based on a photographing distance and a deviation correction between a measurement value obtained by a distance meter and a photographing distance are performed.