Displacement monitoring system and method based on machine vision

Abstract

The present invention relates to a displacement monitoring system based on machine vision, wherein the front end of a telephoto lens of the system is fixedly connected with a non-visible light filter, the rear end of the telephoto lens is fixedly connected with a machine vision sensor, a servo motor is fixedly connected below the machine vision sensor, the machine vision sensor and the servo motor are respectively connected with a microcomputer processor data line, the microcomputer processor is also connected with a data telemetry module data line, a plurality of non-visible light fixed objects are fixed on a building foundation monitored by the machine vision sensor, the data telemetry module is wirelessly connected with a remote data receiving module, and an alarm module is arranged on the remote data receiving module. The machine vision-based monitoring system for the displacement of the top of a building foundation pile solves the problems of low monitoring accuracy, low efficiency, inability to continuously monitor, and large errors existing in the prior art.

Description

Technical field:

[0001] The present invention relates to the technical field of building displacement monitoring, and in particular to a machine vision-based system and method for monitoring the displacement of a construction engineering support pile top, a concrete structure and a steel member. Background technology:

[0002] With the development of my country's economy and social construction, there are more and more construction projects, and the importance and difficulty of displacement monitoring systems and methods for construction projects are becoming increasingly prominent. As an important part of building displacement monitoring, it is particularly difficult to monitor the displacement of construction projects 24 hours a day. The existing monitoring method for the displacement of the top of the supporting piles of construction projects is to use a total station or a theodolite, and monitor manually with the naked eye after setting up the station. The defects of the existing monitoring method are mainly reflected in two aspects. On the one hand, the efficiency of monitoring by artificial vision plus a total station or theodolite is low, and the visual error of the monitoring personnel is large; on the other hand, it is difficult to monitor 24 hours a day due to the inconvenience of monitoring at night or in bad weather. That is, the monitoring accuracy and monitoring continuity of the existing monitoring methods cannot meet the requirements, and it also consumes human resources. Summary of the invention:

[0003] Purpose of the invention:

[0004] The present invention provides a machine vision-based construction engineering support pile top, concrete structure and steel member displacement monitoring system and method, which aims to solve the problem that the monitoring accuracy and monitoring continuity of the existing monitoring methods cannot meet the engineering requirements.

[0005] Technical solution:

[0006] A displacement monitoring system based on machine vision, wherein the front end of a telephoto lens of the system is fixedly connected with a non-visible light filter, the rear end of the telephoto lens is fixedly connected with a machine vision sensor, a servo motor is fixedly connected below the machine vision sensor, the machine vision sensor and the servo motor are respectively connected with a microcomputer processor data line, the microcomputer processor is also connected with a data telemetry module data line, a plurality of non-visible light fixed objects are fixed on a building foundation monitored by the machine vision sensor, the data telemetry module is wirelessly connected with a remote data receiving module, and an alarm module is arranged on the remote data receiving module.

[0007] The servo motor is fixedly arranged at the center of the support plate, the lower end surface of the support plate is fixedly connected to the bracket, the bottom end of the bracket is an inverted cone structure, a boss is arranged on the upper end surface of the support plate, a ball is arranged in the boss, the ball contacts the lower surface of the shell, the output shaft of the servo motor passes through the boss and is fixedly connected to the shell, and a protective cover is also fixedly arranged on the support plate, the protective cover includes a fixed plate, a passive sliding cover and an active sliding cover, the curvature of the fixed plate, the passive sliding cover and the active sliding cover are consistent, the fixed plate is an arc-shaped plate structure with a square tube-shaped cross section, the fixed plate is fixed on the upper end surface of the support plate, the passive sliding cover is an arc-shaped plate with a square ring-shaped cross section composed of an inner wall and an outer wall, the square ring-shaped cavity formed between the inner wall and the outer wall of the passive sliding cover is clamped with the fixed plate, the active sliding cover is an arc-shaped plate structure, and the inner cavity formed by the inner wall of the passive sliding cover is clamped with the active sliding cover.

[0008] A displacement monitoring system method based on machine vision comprises the following steps:

[0009] Step 1: Install a displacement monitoring system based on machine vision;

[0010] Step 2: calibrate the monitoring points and obtain the coordinates of monitoring point 0 according to the acquisition method;

[0011] Step 3, the first collection cycle on day a; obtain the first collection cycle data according to the collection method, automatically save the records to the hard disk and upload them to the cloud, a≥1;

[0012] Step 4, the second collection cycle on day a; obtain the data of the second collection cycle according to the collection method, automatically save the records to the hard disk and upload them to the cloud, a≥1;

[0013] Step 5, the i-th collection cycle on the a-th day; obtain the i-th collection cycle data according to the collection method, automatically save the record to the hard disk and upload it to the cloud, a≥1, 1≤i≤24; on the a+1-th day, obtain the i-th collection cycle data on the a+1-th day according to the collection method, automatically save the record to the hard disk and upload it to the cloud;

[0014] Step six, calculate the offset;

[0015] Step 7, determine whether to alarm;

[0016] Step 8. Repeat the above steps to complete automatic monitoring and automatic alarm.

[0017] The steps of the collection method in step 1 to step 5 are:

[0018] 1) Adjust the starting position angle of the machine vision sensor according to the position of the monitoring point 0 of the invisible light fixed object: control the output shaft of the servo motor through the microcomputer processor to drive the rigid body connected with the invisible light filter and the telephoto lens and the machine vision sensor to rotate to the position angle where the monitoring point 0 of the invisible light fixed object is fixed;

[0019] 2) Image acquisition: The microcomputer processor controls the machine vision sensor to acquire images, capturing black and white images of invisible objects;

[0020] 3) Image processing: Perform median filtering on the black and white image collected in step 2), and then perform threshold segmentation;

[0021] 4) Data processing: Using the area and center algorithm of the specified gray value image of the image processing algorithm, calculate the center coordinates X0_0 and Y0_0 and the diameter d of the invisible light fixed object in pixels;

[0022] 5) Calculate actual size: Calculate the actual coordinate data according to the actual size conversion formula. The computer processor and network switch automatically record the data X0_0 and Y0_0 of the point, automatically save the record and upload it to the cloud;

[0023] 6) Collect all data: Continue to execute 1) to 4) to obtain the second monitoring point data X0_1 and Y0_1, obtain the nth monitoring point X0_n and Y0_n within the range of 200 degrees, n ≥ 2, automatically save the records and upload them to the cloud;

[0024] 7) Stop acquisition: stop the servo motor and stop acquisition;

[0025] 8) Servo motor 5 resets: that is, servo motor 5 returns to the starting position of step 1) to wait for the next cycle acquisition.

[0026] Step 3) The median filter output in the image processing method is

[0027] g(x,y)=med{f(xk,yl),(k,l∈W)} (1)

[0028] Among them, f(xk,yl) is the original image, g(x,y) is the processed image, and W is a two-dimensional template, usually a 5*5 area.

[0029] The area calculation formula of the region of the specified gray value image using the image processing algorithm in step 4) is:

[0030] S={(r,c)∈R|Gmin<=Fr.c<=Gmax} (2)

[0031] Among them, S is the output point area, (r, c) is a certain point, R is the specified range of grayscale values, Gmin is 0, Fr.c is the grayscale value of a certain point, and Gmax is 255.

[0032] The center algorithm formula of the specified gray value image using the image processing algorithm in step 4) is:

[0033]

[0034]

[0035] Among them, X is the horizontal axis coordinate, Y is the vertical axis coordinate, W is the image pixel width, H is the image pixel height, and f(i, j) represents the grayscale value at the pixel point (i, j).

[0036] Step 5) The formula for converting the actual size is:

[0037] K=D / d (5)

[0038] Where K is a constant; D is the actual diameter in mm; d is the calculated diameter in pixels.

[0039] The interval between the i-th collection cycle and the previous collection cycle is 1 hour or 2 hours.

[0040] The method for calculating the pile top displacement in step six is ​​to calculate the mean value of each monitoring point in one day; Xn_M=1 / i*∑Xn,Yn_M=1 / i*∑Yn, and subtract the mean value Xn_Mp, Yn_Mp of each monitoring point in the previous day's collection cycle to calculate the offset.

[0041] Advantages and effects:

[0042] 1. The machine vision-based monitoring system for the displacement of the top of the pile of the building foundation solves the problems of low monitoring accuracy, low efficiency, inability to continuously monitor, and large errors in the prior art. The present invention realizes automatic monitoring of the displacement of the top of the pile of the building foundation, thereby reducing the various disadvantages brought about by manual visual total station or theodolite monitoring.

[0043] 2. The present invention makes full use of non-visible light imaging technology. The non-visible light with a wavelength of 940 nanometers to 1000 nanometers is not affected by other light and has the ability to bypass and penetrate. It can also achieve 24-hour monitoring in bad weather or at night.

[0044] 3. The present invention can automatically draw a displacement curve, can perform real-time online monitoring, automatically save data and upload it to the cloud, and remotely alarm.

[0045] 4. After the monitoring system is established for the first time, the present invention does not require manual intervention and automatically completes the monitoring, saving manpower.

[0046] 5. All monitoring data are automatically uploaded to cloud storage and saved in a dedicated folder to avoid human tampering and ensure the authenticity of the data.

[0047] 6. The absolute position encoder of the servo motor has a resolution of 262144 pulses / revolution, which is much higher than the resolution of the total station or theodolite of 108000 seconds / revolution. Description of the drawings:

[0048] Figure 1 It is a schematic diagram of the structure of the displacement monitoring system of the present invention;

[0049] Figure 2 It is the principle diagram of the present invention;

[0050] Figure 3 It is a side view of the machine vision sensor bracket structure;

[0051] Figure 4 This is the main view of the machine vision sensor bracket structure;

[0052] Figure 5 is a schematic diagram of the boss structure;

[0053] Figure 6 It is a structural schematic diagram of the connection between the fixed plate, the passive sliding cover and the active sliding cover;

[0054] Figure 7 It is a schematic diagram of the process of the present invention;

[0055] Labels in the figure: 1. microcomputer processor, 2. machine vision sensor, 3. telephoto lens, 4. invisible light filter, 5. servo motor, 5-1. output shaft, 6. invisible light fixing object, 7. data remote transmission module, 8. alarm module, 9. remote data receiving module, 10. network switch, 11. bracket, 12. support plate, 13. boss, 13-1. ball bearing, 14. protective cover, 14-1. fixing plate, 14-2. passive sliding cover, 14-2-1. inner wall, 14-2-2. outer wall, 14-3. active sliding cover, 15. shell. Specific implementation method:

[0056] The technical solution of the present invention will be described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0057] The building foundation refers to the load-bearing structure below the ground of the building, such as foundation pit, pedestal, frame columns, ground beams, etc.

[0058] like Figure 1-2 As shown, a displacement monitoring system based on machine vision includes a microcomputer processor 1, a machine vision sensor 2, a telephoto lens 3, a non-visible light filter 4, a servo motor 5, a non-visible light fixation object 6, a data remote transmission module 7, an alarm module 8 and a remote data receiving module 9;

[0059] The machine vision sensor 2, telephoto lens 3 and non-visible light filter 4 are rigidly connected to form a rigid body. The front end of the telephoto lens 3 of the system is fixedly connected with the non-visible light filter 4, the rear end of the telephoto lens 3 is fixedly connected with the machine vision sensor 2, and the bottom of the telephoto lens 3 is fixedly connected with the servo motor 5. The machine vision sensor 2 and the servo motor 5 are connected to the microcomputer processor 1 through the network switch 10 star data line or bus. The servo motor 5 is a bidirectional drive motor that can drive the output shaft 5-1 to rotate clockwise or counterclockwise, thereby driving the telephoto lens 3 to rotate and scan in a circle on a horizontal plane parallel to the ground. The horizontal rotation angle is 0 degrees to 200 degrees, which can just cover the horizontal plane of the building foundation. The microcomputer processor 1 is also connected to the data transmission module 7 data line through the network switch 10. Multiple non-visible light stator 6 are fixed on the building foundation monitored by the telephoto lens 3. The data transmission module 7 is wirelessly connected to the remote data receiving module 9, and the remote data receiving module 9 is provided with an alarm module 8. This system can monitor continuously. It avoids the problems of low efficiency and large visual error of monitoring by manual vision plus total station or theodolite; it avoids the situation that it is inconvenient to monitor at night or in bad weather, making it difficult to achieve 24-hour monitoring.

[0060] like Figure 3-6As shown, the servo motor 5 is fixedly arranged at the center of the support plate 12. The support plate 12 is square and convenient for supporting the protective cover, so that the protective cover is large enough to cover and block the rigid body. The front end of the machine vision sensor 2 is fixed with a telephoto lens 3 through a connecting thread. The front end of the telephoto lens 3 is fixedly connected with a non-visible light filter 4 through a connecting thread to form a rigid body. The rigid body is fixedly arranged in the shell 15. The lower end surface of the support plate 12 is fixedly connected with the bracket 11. The bracket 11 is a tripod bracket. The bottom end of the bracket 11 is an inverted cone. The inverted cone facilitates the insertion of the bracket 11 into the soil, making the device more firmly fixed. The upper end surface of the support plate 12 is provided with a boss 13, and the boss 13 is higher than the plane of the support plate 12 to avoid the influence of the machine vision sensor 2 when there is a small amount of rain and snow on the plane of the support plate 12. A ball 13-1 is arranged in the boss 13, and the ball 13-1 contacts the lower surface of the shell 15. The ball 13-1 plays a role in supporting and reducing the friction of the rotation of the shell 15. The output shaft 5-1 of the servo motor 5 passes through the boss 13 and is fixedly connected to the shell 15. A protective cover 14 is also fixedly arranged on the support plate 12. The protective cover 14 is used to block rain and snow and prevent objects from hitting the rigid body. At the same time, the protective cover 14 is only a shed structure. There is no obstruction on both sides, which does not affect the field of view of the machine vision sensor 2 when rotating to take pictures, and is beneficial to the heat dissipation of the rigid body. The protective cover 14 includes a fixed plate 14-1, a passive sliding cover 14-2 and an active sliding cover 14-3. The curvatures of the fixed plate 14-1, the passive sliding cover 14-2 and the active sliding cover 14-3 are consistent. The fixed plate 14-1 is an arc-shaped plate structure with a square tube-shaped cross section. The fixed plate 14-1 is fixed to the upper end surface of the support plate 12 by bolts. The passive sliding cover 14-2 is an arc-shaped plate with a square ring-shaped cross section composed of an inner wall 14-2-1 and an outer wall 14-2-2. The square ring-shaped cavity formed between the inner wall 14-2-1 and the outer wall 14-2-2 of the passive sliding cover 14-2 is clamped with the fixed plate 14-1. The active sliding cover 14-3 is an arc-shaped plate structure. The inner cavity formed by the inner wall 14-2-1 of the passive sliding cover 14-2 is clamped with the active sliding cover 14-3. The square ring-shaped cavity formed between the inner wall and the outer wall of the passive sliding cover 14-2 is open at one end and closed at the other end. The open end is clamped with the fixed plate 14-1, and the closed end is clamped with the active sliding cover 14-3. That is, the thickness of the passive sliding cover 14-2 is greater than the thickness of the fixed plate 14-1 and the active sliding cover 14-3.

[0061] Pull the raised handle provided on the active slide cover 14-3, pull the handle, the active slide cover 14-3 drives the passive slide cover 14-2 to be pulled out from the fixed plate 14-1, at this time, the passive slide cover 14-2 is at the top of the arc, because the active slide cover 14-3 is sleeved in the passive slide cover 14-2, and the fixed plate 14-1 is also sleeved in the passive slide cover 14-2, so when rain and snow fall on the protective cover 14, they will flow down along the passive slide cover 14-2, and will not flow into the clamping parts between the passive slide cover 14-2 and the active slide cover 14-3, and between the passive slide cover 14-2 and the fixed plate 14-1, thereby avoiding the problem of easy rust; when it is necessary to adjust the machine vision sensor 2, the active slide cover 14-3 can be pushed into the passive slide cover 14-2, driving the passive slide cover 14-2 to be sleeved on the fixed plate 14-1, so as to facilitate the adjustment of the machine vision sensor 2.

[0062] The microcomputer processor 1 is an existing processor; the microcomputer processor 1 includes a CPU, memory, hard disk, display, keyboard, mouse and network switch. The CPU is Intel(R) Core(TM) i5-3320M CPU@2.6GHZ, the memory is 8GB, the hard disk is a 128G solid state drive, and it runs a windows7 64-bit operating system. The microcomputer processor uses Ethernet to connect the machine vision sensor and the servo motor driver as well as the data remote transmission module through the network switch 10.

[0063] The network switch 10 is an existing switch, which is a network device used for forwarding electrical (optical) signals. It can provide an exclusive electrical signal path for any two network nodes connected to the switch. The most common switch is an Ethernet switch, and the present invention uses 4 network ports.

[0064] The machine vision sensor 2 is a CCD photoelectric sensor industrial camera with global shutter, black and white channels, high resolution and high acquisition rate. The shape and size of the invisible light fixed object 6 are collected through the invisible light filter 4 by using the telephoto lens 3 and fed back to the microcomputer processor 1.

[0065] The microcomputer processor 1 is connected to the machine vision sensor 2 via USB or Ethernet, the servo motor 5 is connected to the microcomputer processor 1 via a bus, and the data transmission module 7 is connected to the microcomputer processor 1 via USB or Ethernet. The microcomputer processor can control the operation of the servo motor and continuously process the image of the invisible fixed object 6 obtained by the machine vision sensor, calculate the displacement data through the image algorithm, and send it to the remote data receiving and alarm module 8 through the data transmission module 7. At the same time, the data is uploaded to the cloud for permanent storage. The microcomputer processor draws the offset curve in real time, which is intuitively displayed.

[0066] The invisible light fixture 6 is a non-illumination electric light source with the main purpose of generating infrared radiation. It can be a thermal radiation infrared light source, a gas discharge infrared light source or a laser infrared light source. The present invention is preferably a laser infrared light source. The invisible light fixture 6 is a circular stainless steel structure with a diameter of 200 mm. It can emit a cylindrical waterproof, dustproof, and vibration-resistant infrared light source. The semiconductor laser is an existing product with the following characteristics: A Directionality: The light emitted by the laser is emitted in one direction, and the divergence angle of the light beam is small, close to the ideal parallel light; B Monochromaticity: The spectral width of the laser is very small, which is on the order of several nanometers. Therefore, it has a good monochromatic light source; C High brightness and high energy density. The specific structure of the invisible light fixture 6 is 100 mm in height, 200 mm in inner diameter, 210 mm in outer diameter, a stainless steel cylindrical sealed body, the top surface is sealed with tempered wear-resistant transparent glass, and the bottom surface is sealed with a stainless steel flange plate with fixed mounting holes. The sealed body is provided with a replaceable 18650 battery and five evenly distributed laser infrared lamp cores (model HG_IR3503).

[0067] Infrared radiation is electromagnetic radiation in a certain range with a wavelength greater than that of red light, with a wavelength of 0.78 to 1000 μm. It is divided into three bands: near infrared (codenamed IR-A, wavelength 0.78 to 1.4 μm), mid infrared (IR-B, 1.4 to 3 μm), and far infrared (IR-C, 3 to 1000 μm).

[0068] The corresponding infrared light sources are called near infrared, mid infrared and far infrared. Infrared light sources are commonly used for heating, physical therapy, night vision, communication, navigation, plant cultivation and livestock breeding.

[0069] There are 3 types:

[0070] 1. Thermal radiation infrared light source

[0071] It can be a black body, a powered silicon carbide rod, etc. A black body is an ideal thermal radiation infrared light source because at the same temperature, the radiation power density of a black body is the highest.

[0072] Incandescent bulbs can convert more than 75% of input electrical energy into infrared radiation and can be used as infrared light sources. Because the infrared radiation above 5000nm radiated by the incandescent body is absorbed by the glass shell, the incandescent lamp is a near-infrared and mid-infrared light source. In order to make full use of the infrared radiation of the incandescent bulb, a reflective glass shell can be used to concentrate the infrared radiation to the front through the aluminum reflective surface at the rear of the glass shell, further enhancing the infrared radiation effect. In addition, quartz tube-shaped infrared incandescent lamps can also be used as infrared light sources. It works on the principle of halogen tungsten cycle, has a small size, high mechanical strength, is easy to install and use, and has a lifespan of more than 5000 hours.

[0073] When the silicon carbide rod is electrically heated, it emits approximately black body radiation in the wavelength range of 2000 to 20000nm, and is a medium and far infrared light source.

[0074] By coating the surface of a heating object with oxides of titanium, zirconium, chromium, manganese, iron, nickel and silicon, or carbides of boron and silicon, a far-infrared light source can be made.

[0075] 2. Gas discharge infrared light source

[0076] Some gas discharge light sources generate infrared radiation when they discharge, and can be used as infrared light sources. The spectrum of xenon lamps is continuous and produces strong radiation in the near-infrared region. They are often used as solar simulation light sources and heat sources for melting special metals or materials. Their infrared radiation is easily modulated and can be used for infrared communications.

[0077] 3. Laser infrared light source

[0078] Some lasers can be used as infrared light sources. The radiation wavelength of yttrium aluminum garnet or neodymium glass solid laser is 1064nm, which is a near-infrared light source.

[0079] Semiconductor lasers made of different materials can be used as infrared light sources from near-infrared to far-infrared bands at different operating temperatures.

[0080] The third laser infrared light source is preferably used to fix the calibration object at the point to be monitored.

[0081] The telephoto lens 3 is an existing SLR camera lens with a long focal length. The front end of the telephoto lens 3 is connected to a non-visible light filter 4, which can filter out visible light with a wavelength below 940 nanometers, allowing non-visible light with a wavelength of 940 nanometers to 1000 nanometers to penetrate without being affected by other light in the body, and has the ability to bypass penetration, so that 24-hour monitoring can be achieved in bad weather or at night. The rear end of the telephoto lens 3 is connected to a machine vision sensor 2. The telephoto lens 3 is a SLR camera lens with a long focal length. The preferred model is EF 400mm f / 2.8L IS USM, which is a Canon telephoto lens with a non-visible light filter connected to the front end, and the filter diameter is 52mm (insertion type), which can filter out visible light with a wavelength below 940 nanometers to avoid interference from visible light. The rear end of the lens is connected to the C-type interface of the machine vision sensor.

[0082] Servo motor 5 is a high-precision motor with an absolute position encoder with a resolution of 262144 pulses / rev. It can accurately stop at a specified angle position with high repeatability. Its rotation angle is controlled by a microcomputer processor 1. Servo motor 5 is a high-precision Mitsubishi servo MR-J3 motor, which can accurately stop at a specified angle position with high repeatability. Its rotation angle is controlled by a microcomputer processor 1. The high-performance servo motor encoder of the MR-J3 series adopts an absolute position encoder with a resolution of 262144 pulses / rev. The speed loop frequency response is increased to 900HZ, with high speed / high torque characteristics.

[0083] The data transmission module 7 is a DTU data transmission module, which is a wireless mobile network and sends the calculated data of the microcomputer processor 1 to the remote data receiving module 9 and its alarm module 8. The DTU data transmission module is an existing product, and the available model is: WL-4010 / 4030 4G LTE DTU, which is fully network-accessible and can send the microcomputer processor calculated data to the remote data receiving module 9 and its alarm module 8 through a wireless network.

[0084] The remote data receiving and alarm module 8 is a mobile phone APP running monitoring data, or a computer running monitoring alarm software, which can receive data sent by the data remote transmission module 7. It is convenient to notify the staff in time.

[0085] The machine vision sensor 2 is a commonly used industrial camera, specifically a CDD machine vision sensor, which has the following characteristics:

[0086] Resolution: The number of pixels (Pixels) of each image captured by the camera. For the industrial camera selected this time, it directly corresponds to the number of pixels of the photoelectric sensor, using 640*480.

[0087] Pixel Depth: The number of bits per pixel data, which uses 8 bits.

[0088] Maximum frame rate (Frame Rate) / line rate (Line Rate): The rate at which the camera collects and transmits images. For area array cameras, it is generally the number of frames collected per second (Frames / Sec), and 25 frames / second is used.

[0089] Exposure mode (Exposure) and shutter speed (Shutter): Use area array camera frame exposure with high shutter speed.

[0090] Pixel Size: Pixel size and pixel number (resolution) together determine the size of the camera target surface. The camera pixel size is 3μm.

[0091] Spectral Range: refers to the sensitivity of the pixel sensor to different light waves. The response range is 350nm-1000nm. The camera adds a non-visible light filter 4 in front of the target to filter out visible light and avoid visible light interference.

[0092] like Figure 7 As shown, a displacement monitoring system method based on machine vision includes the following steps:

[0093] Step 1: Install a displacement monitoring system based on machine vision;

[0094] The visual monitoring component is fixed at the intersection of the vertical and horizontal directions from the building foundation. The center of gravity of the visual monitoring component is 2-3 meters from the ground. Multiple invisible light stators 6 are evenly fixed at the displacement monitoring points of the pile tops around the building foundation, and the monitoring points are 8 meters to 15 meters apart. The various structures of the displacement monitoring system based on machine vision are connected and debugged for standby use. The aperture and focal length of the telephoto lens are automatically adjusted to make the circular invisible light stator 6 at the monitoring point clearly visible. The invisible light stator 6 is fixed using the existing method, and the flange set at its bottom is fixed to the building foundation by bolts.

[0095] Step 2: calibrate the monitoring points and obtain the coordinates of monitoring point 0 on day a according to the acquisition method;

[0096] The zeroth collection cycle on the ath day

[0097] (1) Adjusting the position angle of the machine vision sensor 2 according to the monitoring point 0 of the invisible light stator 6: The output shaft 5-1 of the servo motor 5 is controlled by the microcomputer processor 1 to drive the rigid body connected with the invisible light filter 4, the telephoto lens 3 and the machine vision sensor 2 to rotate to the position angle of the monitoring point 0 where the invisible light stator 6 is fixed.

[0098] (2) Image acquisition: The microcomputer processor 1 controls the machine vision sensor 2 to acquire an image, and acquires a black-and-white image of the invisible light fixed object 6.

[0099] (3) Image processing: Smoothing filter processing is performed on the black and white image collected in step 2, and then threshold segmentation processing is performed. Median filtering is a nonlinear signal processing technology based on sorting statistics theory that can effectively suppress noise. The basic principle of median filtering is to replace the value of a point in a digital image or digital sequence with the median value of each point in a neighborhood of the point, so that the surrounding pixel values ​​are close to the true value, thereby eliminating isolated noise points. The method is to use a two-dimensional sliding template with a certain structure to sort the pixels in the board according to the size of the pixel value, and generate a monotonically rising (or falling) two-dimensional data sequence.

[0100] (4) Data processing: Using the area and center algorithm of the specified gray value image of the image processing algorithm, calculate the center coordinates X0_0 and Y0_0 and the diameter d of the invisible light marker 6 in pixels.

[0101] (5) Calculate the actual size: Divide the actual diameter D of the invisible light measuring object by d to calculate the actual size represented by each pixel in the image. The microcomputer processor 1 automatically records the data X0_0 and Y0_0 of the point, automatically saves the record and uploads it to the cloud.

[0102] (6) Collect all data: Continue to execute 1 to 4 to obtain the second monitoring point data X0_1 and Y0_1. Obtain the nth monitoring point X0_n and Y0_n within the range of 200 degrees, automatically save the records to the hard disk and upload them to the cloud.

[0103] (7) Stop acquisition: Stop the servo motor and stop acquisition.

[0104] (8) Servo motor 5 resets: that is, servo motor 5 returns to the starting position of step 1 and waits for the next cycle of acquisition.

[0105] Step 3: The first collection cycle on day a

[0106] Repeat steps 1 to 8 of the zeroth acquisition cycle acquisition method to obtain the data of the first acquisition cycle

[0107] X1_0, Y1_0; X1_1, Y1_1; X1_n Y1_n

[0108] Automatically save records to hard drive and upload to the cloud.

[0109] Step 4: Second collection cycle on day a

[0110] Repeat steps 1 to 8 of the zeroth acquisition cycle acquisition method to obtain the data of the second acquisition cycle.

[0111] X2_0, Y2_0; X2_1, Y2_1; X2_n, Y2_n

[0112] Automatically save records to hard drive and upload to the cloud.

[0113] Step 5: The i-th collection cycle on the a-th day

[0114] Repeat steps 1 to 8 of the zeroth collection cycle collection method to obtain the data of the i-th collection cycle on the a-th day, automatically save the data to the hard disk and upload it to the cloud. 1≤i≤24; on the a+1th day, obtain the i-th collection cycle data of the a+1th day according to the collection method, automatically save the data to the hard disk and upload it to the cloud. The monitoring cycle is in days until the day when the monitoring ends.

[0115] The interval between the i-th collection cycle and the previous collection cycle is 1 hour or 2 hours. When the interval between the collection cycles is 1 hour, a total of 24 collections are made in one day, that is, i≤24; when the interval between the collection cycles is 2 hours, a total of 12 collections are made in one day, that is, i≤12. If the collection cycle is less than 1 hour, the monitoring point data will be unstable, causing data interference; if the collection cycle is greater than 2 hours, the interval between the collection cycles is longer, which is not conducive to timely monitoring of displacement and timely alarm.

[0116] Step 6: Calculate the displacement of the pile top

[0117] Calculate the mean of each monitoring point on day a+1; Xn_M=1 / i*∑Xn,Yn_M=1 / i*∑Yn, and subtract the mean Xn_Mp, Yn_Mp of each monitoring point in the collection period at the same time on day a to calculate the offset. The mean Xn_M, Yn_M of each monitoring point on day a is Xn_Mp, Yn_Mp_n in the calculation of the pile top displacement on the second day, and so on.

[0118] Since the data at each monitoring point may fluctuate greatly due to weather changes or changes in construction progress, average value calculation will avoid false alarms and reduce unnecessary calculations.

[0119] The offset of monitoring point 0 in the X direction is (X0_M–X0_Mp)*K=BX_0

[0120] The offset of monitoring point 0 in the Y direction is (Y0_M–Y0_Mp)*K=BY_0

[0121] The offset of the first monitoring point in the X direction is (X1_M–X1_Mp)*K=BX_1

[0122] The offset of the first monitoring point in the Y direction is (Y1_M–Y1_Mp)*K=BY_1

[0123] The offset of the second monitoring point in the X direction is (X2_M–X2_Mp)*K=BX_2

[0124] The offset of the second monitoring point in the Y direction is (Y2_M–Y2_Mp)*K=BY_2

[0125] The offset of the nth monitoring point in the X direction is (Xn_M–Xn_Mp)*K=BX_n

[0126] The offset of the nth monitoring point in the Y direction is (Yn_M–Yn_Mp)*K=BY_n

[0127] Step 7: Determine whether to alarm

[0128] The microcomputer processor 1 draws the offset curve of the data curve of each monitoring point on the a+1 day and the data of the monitoring points on the a+1 day and the a day in real time, which is intuitively displayed. The offset is sent to the remote data receiving 9 and its alarm module 8 through the wireless mobile network via the data remote transmission module 7. If it exceeds 5mm, an alarm is triggered, and if it does not exceed 5mm, no alarm is triggered.

[0129] Step 8. Repeat the above steps to complete automatic monitoring and automatic alarm.

[0130] The steps of the collection method in step 1 to step 5 are:

[0131] 1) Adjust the starting position angle of the machine vision sensor 2 according to the position of the monitoring point 0 of the invisible light stator 6: Control the output shaft 5-1 of the servo motor 5 through the microcomputer processor 1 to drive the rigid body connected with the invisible light filter 4 and the telephoto lens 3 and its machine vision sensor 2 to rotate to the position angle where the monitoring point 0 of the invisible light stator 6 is fixed.

[0132] 2) Image acquisition: The microcomputer processor 1 controls the machine vision sensor 2 to acquire images, and acquires a black and white image of the invisible light fixed object 6.

[0133] 3) Image processing: The black and white image collected in step 2) is subjected to median filtering and then threshold segmentation.

[0134] The median filter in the image processing method outputs the two-dimensional median filter as follows:

[0135] g(x,y)=med{f(xk,yl),(k,l∈W)} (1)

[0136] Among them, f(xk,yl) is the original image, g(x,y) is the processed image, and W is a two-dimensional template, usually a 5*5 area.

[0137] 4) Data processing: Using the area and center algorithm of the specified gray value image of the image processing algorithm, calculate the center coordinates X0_0 and Y0_0 and the diameter d of the invisible light marker 6 in pixels.

[0138] The area calculation formula of the gray value image is: the diameter d is calculated based on the area value

[0139] S={(r,c)∈R|Gmin<=Fr.c<=Gmax} (2)

[0140] Among them, S is the output point area, (r, c) is a certain point, R is the specified range of grayscale values, Gmin is 0, Fr.c is the grayscale value of a certain point, and Gmax is 255.

[0141] The center algorithm formula of the gray value image is:

[0142]

[0143]

[0144] Among them, X is the horizontal axis coordinate, Y is the vertical axis coordinate, W is the image pixel width, H is the image pixel height, and f(i, j) represents the grayscale value at the pixel point (i, j).

[0145] 5) Calculate actual size: Calculate the actual coordinate data according to the actual size conversion formula, and the computer processor and network switch (1) automatically record the data X0_0 and Y0_0 of the point, automatically save the record and upload it to the cloud.

[0146] The actual size formula is:

[0147] K=D / d (5)

[0148] Where K is a constant; D is the actual diameter in mm; d is the calculated diameter in pixels.

[0149] 6) Collect all data: Continue to execute 1) to 4) to obtain the second monitoring point data X0_1 and Y0_1, and obtain the nth monitoring point X0_n and Y0_n within the range of 200 degrees, n ≥ 2, automatically save the records and upload them to the cloud.

[0150] 7) Stop acquisition: Stop the servo motor and stop acquisition.

[0151] 8) Servo motor 5 resets: that is, servo motor 5 returns to the starting position of step 1) to wait for the next cycle acquisition.

[0152] The interval between the i-th collection cycle and the previous collection cycle is 1 hour or 2 hours. When the interval between the collection cycles is 1 hour, there are 24 collections in a day, that is, i≤24. When the interval between the collection cycles is 2 hours, there are 12 collections in a day, that is, i≤12.

[0153] Example 1

[0154] A displacement monitoring system based on machine vision, the system comprises a microcomputer processor 1, a machine vision sensor 2, a telephoto lens 3, a non-visible light filter 4, a servo motor 5, a non-visible light fixation object 6, a data remote transmission module 7, an alarm module 8 and a remote data receiving module 9;

[0155] The machine vision sensor 2, telephoto lens 3 and non-visible light filter 4 are rigidly connected to form a rigid body. The front end of the telephoto lens 3 of the system is fixedly connected with the non-visible light filter 4, the rear end of the telephoto lens 3 is fixedly connected with the machine vision sensor 2, and the servo motor 5 is fixedly connected below the machine vision sensor 2. The machine vision sensor 2 and the servo motor 5 are connected to the microcomputer processor 1 through a network switch 10 star data line or bus. The servo motor 5 is a bidirectional drive motor that can drive the output shaft 5-1 to rotate clockwise or counterclockwise, thereby driving the telephoto lens 3 to rotate and scan in a circle on a horizontal plane parallel to the ground. The horizontal rotation angle is 0 degrees to 200 degrees, which can just cover the horizontal plane of the building foundation. The microcomputer processor 1 is also connected to the data transmission module 7 through the network switch 10. A plurality of non-visible light stator objects 6 are fixed on the building foundation monitored by the machine vision sensor 2. The data transmission module 7 is wirelessly connected to the remote data receiving module 9, and the remote data receiving module 9 is provided with an alarm module 8.

[0156] A monitoring method for a building foundation pile top displacement monitoring system based on machine vision comprises the following steps:

[0157] Step 1: Install a displacement monitoring system based on machine vision;

[0158] When the displacement points of the top of the building foundation pile, concrete structures or steel components need to be monitored, any number of non-visible light stator objects 6 can be fixed. In this embodiment, five non-visible light stator objects 6 are used for illustration, marked as monitoring point 0, monitoring point 1, monitoring point 2, monitoring point 3 and monitoring point 4. At the corner points of the building foundation corresponding to these six points, the monitoring point spacing is 8-12 meters, which is selected according to the actual situation. In this embodiment, the monitoring point spacing is 10 meters. Fix the servo motor 5, and thread a rigid body (composed of non-visible light filter / telephoto lens / machine vision sensor) on the output shaft 5-1 of the servo motor 5. The distance between the rigid body and the building foundation is set according to the actual situation to ensure that the machine vision sensor can cover all monitoring points. Adjust the origin position of the servo motor so that it is in the starting direction of the monitoring point.

[0159] Use a gigabit network cable to connect the machine vision sensor, microcomputer processor and data transmission module to form a local area network, run the remote data receiving terminal, and establish a wireless communication connection with the data transmission module.

[0160] Confirm the servo motor origin angle position data and the angle position of monitoring point 0, as well as the angle positions of monitoring points 1 to 4. Adjust the telephoto lens aperture and focal length to make the circular invisible cursor 6 of the monitoring point clearly visible.

[0161] The servo motor 5 is run to return to the origin and is ready to start collecting the invisible cursor 6 at the monitoring point.

[0162] Step 2: calibrate the monitoring points and obtain the coordinates of monitoring point 0 on the first day according to the acquisition method;

[0163] Adjust the position angle of the machine vision sensor 2 according to the monitoring point 0 of the invisible light stator 6: control the output shaft 5-1 of the servo motor 5 through the microcomputer processor 1 to drive the rigid body connected with the invisible light filter 4 and the telephoto lens 3 and its machine vision sensor 2 to rotate to the position angle of the monitoring point 0 where the invisible light stator 6 is fixed.

[0164] Start the servo motor 5 and rotate it counterclockwise at 10 revolutions per second. Capture the invisible light calibrating object 6 at monitoring point 0. The image is processed by smoothing and filtering. The invisible light calibrating object 6 at monitoring point 0 is segmented by threshold. The area and center algorithm of the specified gray value image of the image processing algorithm is used to calculate the center coordinate of the calibration object in pixels as 257 / 315, and the diameter is 100. 200mm / 100=2 mm, that is, one pixel represents 2 mm. Automatically save the record and upload it to the cloud.

[0165] Collect all data: Continue to perform collection steps 1) to 4) to obtain the data X0_1 and Y0_1 of the second monitoring point on the first day. Get the nth monitoring point X0_n and Y0_n within the 180-degree range, automatically save the records to the hard disk and upload them to the cloud.

[0166] At monitoring point 1, the invisible fixed object is captured, and after the same operation, the center coordinates 123 / 321 in pixels are calculated. After capturing monitoring point 1, the servo motor moves to the next monitoring point.

[0167] At monitoring point 2, the invisible object is captured, and after the same operation, the center coordinate 40 / 324 in pixels is calculated. After capturing monitoring point 2, the servo motor moves to the next monitoring point.

[0168] At monitoring point 3, the invisible object is captured, and after the same operation, the center coordinate 43 / 480 in pixels is calculated. After the servo motor captures monitoring point 3, it moves to the next monitoring point.

[0169] At monitoring point 4, the invisible fixed object is captured, and after the same operation, the center coordinates in pixels are calculated as 123 / 700. After the servo motor captures monitoring point 4, it runs the origin return instruction and waits for the next cycle (second cycle) to capture.

[0170] Stop acquisition: Stop the servo motor and stop acquisition.

[0171] Servo motor 5 resets: that is, servo motor 5 returns to the starting position of step 1) to wait for the next cycle acquisition.

[0172] Step 3, the first collection cycle; after the zeroth collection cycle ends, delay for 1 hour, obtain the first collection cycle data according to the collection method, automatically save and record to the hard disk and upload to the cloud; every 1 hour delay, collect a collection cycle, and automatically save and record the collected data to the hard disk and upload to the cloud;

[0173] Step 4: Calculate the offset.

[0174] Calculate the mean of each monitoring point after the i-th collection cycle; Xn_M = 1 / i*∑Xn, Yn_M = 1 / i*∑Yn, subtract the mean of each monitoring point in the previous day's collection cycle to calculate the offset.

[0175] The coordinates of monitoring point 0 have not changed, the coordinates of monitoring point 1 have not changed, the coordinates of monitoring point 2 have not changed, the coordinates of monitoring point 3 have not changed, and the coordinates of monitoring point 4 have not changed.

[0176] Step 5: Subtract the mean data of each monitoring point after the i-th collection cycle from the mean data of each monitoring point on the previous day, and obtain that the offset is 0 mm. If it is less than 5 mm, no alarm will be issued.

[0177] Step 6: All coordinate data are automatically saved, uploaded to the cloud, and sent to the remote data receiving terminal mobile phone of relevant personnel through the DTU data remote module to complete automatic monitoring.

[0178] Step 7: Repeat the above steps, and all coordinate data will be automatically saved. Uploaded to the cloud, and sent to the remote data receiving terminal mobile phone of relevant personnel through the DTU data remote module. Automatic monitoring and alarm are completed.

Claims

1. A displacement monitoring system based on machine vision, characterized in that: The front end of the telephoto lens (3) of the system is fixedly connected to a non-visible light filter (4), the rear end of the telephoto lens (3) is fixedly connected to a machine vision sensor (2), a servo motor (5) is fixedly connected below the machine vision sensor (2), the machine vision sensor (2) and the servo motor (5) are respectively connected to a microcomputer processor (1) data line, the microcomputer processor (1) is also connected to a data remote transmission module (7) data line, a plurality of non-visible light fixed objects (6) are fixed on the building foundation monitored by the machine vision sensor (2), the data remote transmission module (7) is wirelessly connected to a remote data receiving module (9), and an alarm module (8) is provided on the remote data receiving module (9); The servo motor (5) is fixedly arranged at the center of the support plate (12); the lower end surface of the support plate (12) is fixedly connected to the bracket (11); the bottom end of the bracket (11) is an inverted cone structure; the upper end surface of the support plate (12) is provided with a boss (13); a ball (13-1) is arranged in the boss (13); the ball (13-1) contacts the lower surface of the housing (15); the output shaft (5-1) of the servo motor (5) passes through the boss (13) and is fixedly connected to the housing (15); a protective cover (14) is also fixedly arranged on the support plate (12); the protective cover (14) comprises a fixed plate (14-1), a passive sliding cover (14-2) and an active sliding cover (14-3); the fixed plate (14-1), the passive sliding cover (14-2) and the active sliding cover (14-3) are The cover (14-2) and the active sliding cover (14-3) have the same curvature; the fixed plate (14-1) is an arc-shaped plate structure with a square tube-shaped cross section; the fixed plate (14-1) is fixed to the upper end surface of the support plate (12); the passive sliding cover (14-2) is an arc-shaped plate with a square ring-shaped cross section composed of an inner wall (14-2-1) and an outer wall (14-2-2); a square ring-shaped cavity formed between the inner wall (14-2-1) and the outer wall (14-2-2) of the passive sliding cover (14-2) is clamped with the fixed plate (14-1); the active sliding cover (14-3) is an arc-shaped plate structure; an inner cavity formed by the inner wall (14-2-1) of the passive sliding cover (14-2) is clamped with the active sliding cover (14-3).

2. A monitoring method using the machine vision-based displacement monitoring system according to claim 1, characterized in that: The steps include: Step 1: Install a displacement monitoring system based on machine vision; Step 2: calibrate the monitoring points and obtain the coordinates of monitoring point 0 according to the acquisition method; Step 3, the first collection cycle on day a; obtain the first collection cycle data according to the collection method, automatically save the records to the hard disk and upload them to the cloud, a≥1; Step 4, the second collection cycle on day a; obtain the data of the second collection cycle according to the collection method, automatically save the records to the hard disk and upload them to the cloud, a≥1; Step 5, the i-th collection cycle on the a-th day; obtain the i-th collection cycle data according to the collection method, automatically save the record to the hard disk and upload it to the cloud, a≥1, 1≤i≤24; on the a+1-th day, obtain the i-th collection cycle data on the a+1-th day according to the collection method, automatically save the record to the hard disk and upload it to the cloud; Step six, calculate the offset; Step 7, determine whether to alarm; Step 8. Repeat the above steps to complete automatic monitoring and automatic alarm.

3. The monitoring method of the displacement monitoring system based on machine vision according to claim 2, characterized in that: The steps of the collection method in step 1 to step 5 are: 1) adjusting the starting position angle of the machine vision sensor (2) according to the position of the monitoring point 0 of the invisible light stator (6): controlling the output shaft (5-1) of the servo motor (5) through the microcomputer processor (1) to drive the rigid body connected with the invisible light filter (4) and the telephoto lens (3) and the machine vision sensor (2) to rotate to the position angle where the monitoring point 0 of the invisible light stator (6) is fixed; 2) Image acquisition: The microcomputer processor (1) controls the machine vision sensor (2) to acquire images, and acquires a black and white image of the invisible light fixed object (6); 3) Image processing: Perform median filtering on the black and white image collected in step 2), and then perform threshold segmentation; 4) Data processing: using the area and center algorithm of the specified gray value image of the image processing algorithm, calculate the center coordinates X0_0 and Y0_0 and the diameter d of the invisible light fixed object (6) in pixels; 5) Calculate the actual size: Calculate the actual coordinate data according to the actual size conversion formula, and the microcomputer processor (1) automatically records the data X0_0 and Y0_0 of the point, automatically saves the record and uploads it to the cloud; 6) Collect all data: Continue to execute 1) to 4) to obtain the second monitoring point data X0_1 and Y0_1, obtain the nth monitoring point X0_n and Y0_n within the range of 200 degrees, n ≥ 2, automatically save the records and upload them to the cloud; 7) Stop acquisition: stop the servo motor and stop acquisition; 8) Servo motor 5 resets: that is, servo motor 5 returns to the starting position of step 1) to wait for the next cycle acquisition.

4. The monitoring method of the displacement monitoring system based on machine vision according to claim 3, characterized in that: Step 3) The median filter output in the image processing method is g(x,y)=med{f(xk,yl),(k,l∈W)} (1) Among them, f(xk,yl) is the original image, g(x,y) is the processed image, and W is a two-dimensional template, usually a 5*5 area.

5. The monitoring method of the displacement monitoring system based on machine vision according to claim 3, characterized in that: The area calculation formula of the region of the specified gray value image using the image processing algorithm in step 4) is: S ={(r,c)∈R|Gmin<=Fr.c<=Gmax} (2) Among them, S is the output point area, (r, c) is a certain point, R is the specified range of grayscale values, Gmin is 0, Fr.c is the grayscale value of a certain point, and Gmax is 255.

6. The monitoring method of the displacement monitoring system based on machine vision according to claim 3, characterized in that: The center algorithm formula of the specified gray value image using the image processing algorithm in step 4) is: Wherein, X is the horizontal axis coordinate, Y is the vertical axis coordinate, W is the image pixel width, H is the image pixel height, and f(i, j) represents the grayscale value at the pixel point (i, j).

7. The monitoring method of the displacement monitoring system based on machine vision according to claim 3, characterized in that: Step 5) The formula for converting the actual size is: K=D / d (5) Where K is a constant; D is the actual diameter in mm; d is the calculated diameter in pixels.

8. The monitoring method of the displacement monitoring system based on machine vision according to claim 2, characterized in that: The interval between the i-th collection cycle and the previous collection cycle is 1 hour or 2 hours.

9. The monitoring method of the displacement monitoring system based on machine vision according to claim 2, characterized in that: The method for calculating the displacement of the pile top in step 6 is to calculate the average value of each monitoring point within one day; Xn_M=1 / i*∑Xn,Yn_M=1 / i*∑Yn, and subtract the mean values ​​Xn_Mp, Yn_Mp of each monitoring point in the previous day's collection period to calculate the offset.

Images