Method for measuring center offset of bottom of prefabricated sinking well under water

Abstract

The application discloses a kind of underwater excavation prefabricated sinking well bottom center offset measurement methods, including steps:1, in the blade foot ring (11) of sinking well (1), fixed beam (2) is equipped, and fixed beam center is connected by suspending rope (3) float ball (4);2, float ball is floated on water surface (5) by suspending rope;3, binocular camera (6) is equipped above float ball, and the two cameras of binocular camera are symmetrically located on the two sides of float ball, establish measurement point coordinate model, and the coordinates of float ball center in camera coordinate system are obtained using binocular camera and measurement point coordinate model;4, calibrate camera coordinate system and convert into world coordinate system, obtain the coordinates of float ball center under world coordinate system;5, the coordinates of float ball center under world coordinate system are compared with design center position, and the center offset of sinking well bottom is obtained.The application can quickly and accurately measure the center position of sinking well bottom, to judge whether the center of sinking well bottom is offset and obtain offset.

Description

TECHNICAL FIELD

[0001] The present application relates to a method for measuring the center of the bottom of a sinking well, in particular to a method for measuring the center deviation of the prefabricated sinking well bottom in underwater excavation. BACKGROUND

[0002] The sinking well method is also known as the sinking well method. Before the soil is excavated, a section of the well wall is partially sunk into the soil by its own weight at the designed position of the well shaft. Then, under its cover, the well wall is built accordingly while excavating and sinking. In the process of excavation and sinking, in order to avoid the occurrence of construction safety accidents caused by the center deviation of the sinking well bottom, the center position of the sinking well bottom needs to be measured to ensure that the center of the sinking well bottom is within the allowable deviation range.

[0003] Because the measurement point is located at the bottom of the sinking well, the space is small and the light is poor, so it is impossible to accurately measure the center position of the well bottom. Especially in the construction of prefabricated sinking well in underwater excavation, the sinking well bottom is located underwater, and the construction personnel cannot effectively measure the center position of the sinking well bottom. Therefore, it is necessary to provide a method for measuring the center deviation of the prefabricated sinking well bottom in underwater excavation, which can quickly and accurately measure the center position of the sinking well bottom, and is used to judge whether the center of the sinking well bottom deviates and obtain the deviation. SUMMARY

[0004] The purpose of the present application is to provide a method for measuring the center deviation of the prefabricated sinking well bottom in underwater excavation, which can quickly and accurately measure the center position of the sinking well bottom, and is used to judge whether the center of the sinking well bottom deviates and obtain the deviation.

[0005] The present application is implemented as follows:

[0006] A method for measuring the center deviation of the prefabricated sinking well bottom in underwater excavation, comprising the following steps:

[0007] Step 1: A fixed beam is arranged in the blade foot ring at the bottom of the sinking well, and a floating ball is connected by a hanging rope at the center of the fixed beam;

[0008] Step 2: As the sinking well sinks, the floating ball floats on the water surface through the hanging rope;

[0009] Step 3: A binocular camera is arranged above the floating ball, a measurement point coordinate model is established, and the coordinates P(X, Y, Z) of the center of the floating ball in the camera coordinate system are obtained by using the binocular camera and the measurement point coordinate model;

[0010] Step 4: The camera coordinate system is calibrated, the camera coordinate system is converted into the world coordinate system, and the coordinates P(x, y, z) of the center of the floating ball in the world coordinate system are obtained. w w w w ​​​​

[0011] Step 5: According to the comparison of the coordinates of the floating ball center in the world coordinate system and the design center position of the caisson, the center offset of the caisson bottom is obtained.

[0012] The fixed beam is in a cross-shaped structure, and the center of the cross-shaped structure is located on the central axis of the blade foot ring 11.

[0013] The projection of the bottom edge of the binocular camera on the water surface is perpendicular to the projection of the fixed beam on the water surface, and the two cameras of the binocular camera are symmetrically located on the two sides of the floating ball.

[0014] The step 3 comprises the following sub-steps:

[0015] Step 3.1: An optimized measuring point coordinate model is established and trained by using a Yolov5 target detection algorithm;

[0016] Step 3.2: During the underwater caisson construction process, the photo of the floating ball is obtained by the binocular camera and input into the measuring point coordinate model;

[0017] Step 3.3: The measuring point coordinate model outputs the horizontal position information of the center of the floating ball in the camera coordinate system of the binocular camera, and calculates the coordinates P(X, Y, Z) of the center of the floating ball in the camera coordinate system.

[0018] The input of the measuring point coordinate model is the shooting image of the binocular camera, and the output of the measuring point coordinate model is the horizontal position information of the center of the floating ball in the camera coordinate system of the binocular camera, that is, P1(u1, v1) and P2(u2, v2).

[0019] Compared with the prior art, the present application has the following beneficial effects:

[0020] 1、The present application has a cross-shaped beam coaxially arranged in the blade foot ring at the bottom of the caisson, and the floating ball is floated on the water surface directly above the center of the cross-shaped beam by using the suspension rope, so as to extend the offset of the center of the caisson bottom to the water surface, realize the measurement of the center of the caisson bottom under water, and quickly determine whether the center of the caisson bottom has deviated by comparing the coordinate position of the floating ball with the design coordinate position of the caisson center, and quickly calculate the offset, which is convenient and efficient, and further ensures the vertical sinking construction of the caisson, improves the construction quality and safety.

[0021] 2、The present application adopts a Yolov5 CNN neural convolution network to establish a measuring point coordinate model, and adopts a binocular camera to obtain a floating ball position photo input into the measuring point coordinate model, and outputs the horizontal position of the center of the floating ball through the measuring point coordinate model, so as to obtain the coordinates of the floating ball in the camera coordinate system, and calibrate and convert it into the coordinates in the world coordinate system, so that the measurement process is simple, the measurement error of the measuring point coordinate model is less than 5mm after training and optimization, and the measurement accuracy of the center offset of the caisson bottom is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is the operation schematic diagram of the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0023] Figure 2 is the installation elevation view of the binocular camera in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0024] Figure 3 is the installation plan view of the binocular camera in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0025] Figure 4 is the error between the prediction frame and the calibration frame of the measurement point coordinate model (training set) in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0026] Figure 5 is the error between the prediction frame and the calibration frame of the measurement point coordinate model (validation set) in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0027] Figure 6 is the target detection error mean of the measurement point coordinate model (training set) in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0028] Figure 7 is the target detection error mean of the measurement point coordinate model (validation set) in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0029] Figure 8 is the floating ball positioning error evaluation model of the measurement point coordinate model in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application;

[0030] Figure 9 is the measurement principle diagram of the binocular camera in the underwater excavation prefabricated caisson bottom center offset measurement method of the present application.

[0031] In the figure, 1 is a caisson, 11 is a blade foot ring, 2 is a fixed beam, 3 is a suspension rope, 4 is a floating ball, 5 is a water surface, 6 is a binocular camera, 7 is a basketball, and 8 is a camera. DETAILED DESCRIPTION

[0032] The present application will be further described below in combination with the drawings and specific embodiments.

[0033] An underwater excavation prefabricated caisson bottom center offset measurement method, comprising the following steps:

[0034] Please refer to the drawings Figure 1, step 1: set the fixed beam 2 in the blade ring 11 at the bottom of the caisson 1, and connect the floating ball 4 in the center of the fixed beam 2 through the hanging rope 3.

[0035] Preferably, the fixed beam 2 is in a cross structure, and the center of the cross structure is located on the central axis of the blade ring 11.

[0036] Step 2: as the caisson 1 sinks, the floating ball 4 floats on the water surface 5 through the hanging rope 3.

[0037] The floating ball 4 floats on the water surface 5 at the upper part of the caisson 1, and the center position of the floating ball 4 is coaxial with the center position of the blade ring 11 at the bottom of the caisson 1. If the center of the floating ball 4 coincides with the design center of the caisson 1, the center of the bottom of the caisson 1 is not offset, otherwise, the center of the bottom of the caisson 1 is offset.

[0038] Please refer to the attached Figure 2 and attached Figure 3 , step 3: set the binocular camera 6 above the floating ball 4, establish a measurement point coordinate model, and use the binocular camera 6 and the measurement point coordinate model to obtain the coordinates P(X, Y, Z) of the center of the floating ball 4 in the camera coordinate system.

[0039] Please refer to the attached Figure 3 , preferably, the projection of the bottom edge of the binocular camera 6 on the water surface 5 is perpendicular to the projection of the fixed beam 2 on the water surface 5, and the two cameras of the binocular camera 6 are symmetrically located on both sides of the floating ball 4.

[0040] The step 3 includes the following sub-steps:

[0041] Step 3.1: use Yolov5 target detection algorithm to establish and train the optimized measurement point coordinate model.

[0042] The input of the measurement point coordinate model is the shooting image of the binocular camera 6, and the output of the measurement point coordinate model is the horizontal position information of the center of the floating ball 4 in the camera coordinate system of the binocular camera 6, that is, P1(u1, v1) and P2(u2, v2).

[0043] Yolo is a target detection algorithm. Yolo redefines object detection as a regression problem, which applies a single convolutional neural network (CNN) to the entire image, divides the image into grids, and predicts the class probability and bounding box of each grid. Yolov5 is an iterative version of Yolo, which adds adaptive anchor box, adaptive image scaling and loss function function compared with the previous version. Yolov5 is an open source framework based on GPL3.0, which has a strong advantage in the rapid deployment of models.

[0044] Due to the influence of light, stains, angles and other factors on the scene, simple image RGB-based recognition cannot guarantee the recognition rate of the center of the floating ball 4. Therefore, deep learning needs to be introduced to establish a measurement point coordinate model to obtain good recognition effect.

[0045] Use a camera with a resolution of 1280*720 to take 40 groups of floating ball 4 photos, and label each group of photos as a training set. The 40 groups of photos can also be divided into training and validation sets according to the proportion, and the measurement point coordinate model is trained and optimized.

[0046] The Backbone stage, i.e. the network stage of extracting features, is a convolutional neural network that aggregates and forms image features at different image granularities. The Backbone stage includes Focus and CSP. The beginning of Backbone is focus, which contains standard convolution. Focus uses the slice operation to integrate W and H information into the channel, and does not cause information loss in the downsampling process. Then use 3*3 convolution to extract features, so that the feature extraction is more sufficient. CSP splits the feature map into two parts, one part is convolved, and the other part is concatenated with the result of the previous convolution operation. In the classification problem, it can reduce the amount of calculation, and in the detection problem, it can improve the learning ability of CNN.

[0047] In the Neck stage, the Neck stage is located between the Backbone stage and the Head (output network) stage, and is a series of network layers that mix and combine image features. The Neck stage passes the image features to the prediction layer. The Neck stage includes SPP and PAN. SPP uses different size windows for pooling to improve feature extraction. PAN guides the shallow network with the deep network, and can get more advanced semantic information. It has a bottom-up process, which improves the positioning ability of the measurement point coordinate model for the center of the floating ball 4, and the positioning is more dependent on shallow information.

[0048] The Head stage is the final detection and result output, which outputs the center of the floating ball 4 in the camera coordinate system of the binocular camera 6, i.e. the horizontal position information P1(u1, v1) and P2(u2, v2).

[0049] Please refer to the attached Figure 4 and the attached Figure 5 , train(Training set) / box_loss and val(Validation set) / box_loss are the error between the predicted box and the labeled box. The smaller the box_loss, the more accurate the predicted box, i.e. the more accurate the horizontal position of the center of the floating ball 4. After 1000 times of training of 40 groups of floating ball 4 photos, the error is less than 1%, and this index will affect the measurement error.

[0050] Please refer to the attached Figure 6 and the attached Figure 7 train (training set) / Obj_loss and val (validation set) / Obj_loss are estimated as the average of the target detection error, the smaller the Obj_loss, the more accurate the target detection, that is, the more accurate the horizontal position of the center of the floating ball 4. After 1000 times of training of 40 groups of floating ball 4 photos, the error is less than 0.25%, and this index will affect the recognition rate.

[0051] The average measurement error of the measurement point coordinate model after 1000 times of training of 40 groups of floating ball 4 photos is less than 5mm. From the box_loss index, increasing the number of training and increasing the training samples can further improve the single measurement accuracy, which can be adjusted according to the actual engineering demand.

[0052] Please refer to the attached Figure 8 Place the camera 8 on the table and keep it horizontal, and use the laser level to ensure that the rectangular side of the table and the bottom side of the camera 8 are parallel, which is used to simulate the position relationship between the binocular camera 6 and the cross beam. The basketball 7 is in a deflated state, which is used to simulate the part of the floating ball 4 under water. The distance between the basketball 7 and the camera 8 is the distance between the binocular camera 6 and the floating ball 4.

[0053] Take 20 photos of the basketball 7 by the camera 8, and input the measurement point coordinate model to evaluate the positioning error of the floating ball 4, and get 20 groups of error data as shown in Table 1, wherein the actual value is the actual offset of the basketball 7, the measured value is the offset of the basketball 7 output by the measurement point coordinate model, and the error is the error of the measured value relative to the actual value.

[0054] Table 1 Floating ball positioning error evaluation table

[0055]

[0056]

[0057]

[0058]

[0059] From Table 1, the measurement variance of 20 groups is 0.082, and the average measurement error is 0.005.

[0060] Step 3.2: During the underwater sinking well construction process, the photos of the floating ball 4 are obtained by the binocular camera 6, and input into the measurement point coordinate model.

[0061] Step 3.3: The measurement point coordinate model outputs the horizontal position information of the center of the floating ball 4 in the camera coordinate system of the binocular camera 6, and calculates the coordinates P (X, Y, Z) of the center of the floating ball 4 in the camera coordinate system.

[0062] Please refer to the attached Figure 9 , the coordinate P(X, Y, Z) of the center of the floating ball 4 is calculated according to the following principle:

[0063] A is the position of the center of the floating ball 4 in the camera coordinate system, b is the distance between the optical centers (O1 and O2) of the left and right cameras of the binocular camera 6, and Z is the depth of the center of the floating ball 4 in the camera coordinate system. L and a R are two symmetric points of the left and right cameras in the camera coordinate system, f is the depth of the two symmetric points a L and a R in the camera coordinate system. Let d be the distance between a L and a R , then d = b - a L + a R . According to the properties of similar triangles, we have:

[0064]

[0065] Substitute d = b - aL + aR into equation (1) to obtain:

[0066]

[0067] According to the horizontal position information P1(u1, v1) and P2(u2, v2) of the floating ball 4, the coordinate value of the center of the floating ball 4 is obtained by triangulation:

[0068]

[0069] Step 4: Calibrate the camera coordinate system, convert the camera coordinate system to the world coordinate system, and obtain the coordinates P w (x w , y w , z w ) of the center of the floating ball 4 in the world coordinate system.

[0070] (1) Camera internal parameters: There are differences between factory and theory, which leads to image distortion, which needs to be corrected by calibration. The present application adopts Zhang's calibration method to calibrate the camera coordinate system, and does not discuss the calibration of camera internal parameters.

[0071] (2) Camera external parameters: Based on Zhang's calibration method, the conversion relationship from the camera coordinate system to the world coordinate system is realized.

[0072] The binocular camera 6 can be preferably a ZED-2i camera, which integrates an imu module. The imu module is the abbreviation of Inertial measurement unit, which is a device for measuring the three-axis attitude angle (or angular velocity) and acceleration of an object.

[0073] Set extrinsic matrix Wherein, the i-th row of R represents the coordinate of the unit vector of the i-th coordinate axis direction in the camera coordinate system in the world coordinate system, the i-th column of R represents the coordinate of the unit vector of the i-th coordinate axis direction in the world coordinate system in the camera coordinate system, and T represents the coordinate of the origin of the world coordinate system in the camera coordinate system; -R transpose *T represents the coordinate of the origin of the camera coordinate system in the world coordinate system.

[0074] Convert the camera coordinate system into the world coordinate system:

[0075]

[0076] Since the floating ball 4 floats on the horizontal surface, i.e. the water surface 5, the water surface 5 is taken as the world coordinate system x-o-y, the vertical direction is z, the rotation matrix obtained by the imu module is R', and R = R' -1 .

[0077] P c The point is taken as the origin of the world coordinate system, A is the measuring point, i.e. the floating ball 4, the bottom edges of the two cameras and the side of the cross beam are perpendicular, T = -P w .

[0078] Step 5: According to the comparison of the coordinates of the center of the floating ball 4 in the world coordinate system and the design center position of the caisson 1, the center offset of the bottom of the caisson 1 is obtained.

[0079] Since the floating ball 4 floats on the water surface 5 through the suspension rope 3, under the condition that the water surface 5 in the caisson 1 is stable, the x-axis and y-axis coordinates of the center of the floating ball 4 in the world coordinate system are the same as the x-axis and y-axis coordinates of the center of the bottom of the caisson 1, and the coordinates of the center of the bottom of the caisson 1 in the world coordinate system can be obtained according to the depth of the caisson 1.

[0080] The coordinates of the center of the bottom of the caisson 1 are compared with the design center of the caisson 1, so that whether the center of the bottom of the caisson 1 is offset and the offset amount of the center of the bottom of the caisson 1 can be determined.

[0081] The above is only a preferred embodiment of the present application, and is not used to limit the protection scope of the application, therefore, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for measuring the center offset of the bottom of a prefabricated caisson during underwater excavation, characterized in that: The method comprises the following steps: Step 1: a fixed beam (2) is arranged in the blade foot ring (11) at the bottom of the caisson (1), and a floating ball (4) is connected through a hanging rope (3) at the center of the fixed beam (2); Step 2: as the caisson (1) sinks, the floating ball (4) floats on the water surface (5) through the hanging rope (3); Step 3: a binocular camera (6) is arranged above the floating ball (4), a measurement point coordinate model is established, and the coordinates P(X, Y, Z) of the center of the floating ball (4) in the camera coordinate system are obtained by using the binocular camera (6) and the measurement point coordinate model; The step 3 comprises the following sub-steps: Step 3.1: an optimized measurement point coordinate model is established and trained by using a Yolov5 target detection algorithm; Step 3.2: during the underwater caisson construction process, the photo of the floating ball (4) is obtained by the binocular camera (6) and input into the measurement point coordinate model; Step 3.3: the measurement point coordinate model outputs the horizontal position information of the center of the floating ball (4) in the camera coordinate system of the binocular camera (6), and the coordinates P(X, Y, Z) of the center of the floating ball (4) in the camera coordinate system are calculated; Step 4: calibrate the camera coordinate system, convert the camera coordinate system to the world coordinate system, and obtain the coordinates P of the center of the floating ball (4) in the world coordinate system w (x w ,y w ,z w ) Step 5: according to the comparison of the coordinates of the center of the floating ball (4) in the world coordinate system and the designed center position of the caisson (1), the center offset of the bottom of the caisson (1) is obtained. The fixed beam (2) is in a cross-shaped structure, and the center of the cross-shaped structure is located on the central axis of the blade foot ring (11).

2. The method of claim 1, wherein the method further comprises: The projection of the bottom edge of the binocular camera (6) on the water surface (5) is perpendicular to the projection of the fixed beam (2) on the water surface (5), and the two cameras of the binocular camera (6) are symmetrically located on both sides of the floating ball (4).

3. The method of claim 1, wherein the method further comprises: The input of the measurement point coordinate model is the shooting image of the binocular camera (6), and the output of the measurement point coordinate model is the horizontal position information of the center of the floating ball (4) in the camera coordinate system of the binocular camera (6), that is, P1(u1, v1) and P2(u2, v2).

Images