Monitoring method of retained vibration depth and retained vibration time of vibratory stone pile based on image recognition

Abstract

The present invention discloses a method for monitoring the vibration depth and vibration time of vibratory crushed stone piles based on image recognition. The method comprises installing a suitable target marker at the end of the vibrator guide rod and constructing an image training set; then building a monitoring system including a CCD industrial camera, a computer storage and processing device, a wireless communication device, etc. according to the on-site construction situation; real-time shooting of the vibrator construction process, and using image recognition technology to extract the coordinate position of the marker in real time and calculate the construction depth of the vibrator; combining the compaction current time-history curve of the vibrator construction process, automatically calculating the vibration time of the vibrator, and finally outputting the construction depth and vibration time of the vibrator to the operating platform in real time for the operator to read. The method of the present invention is contactless monitoring, inherently safe, and has high monitoring accuracy and quality inspection efficiency, and is suitable for quality inspection of multiple vibratory crushed stone pile construction in complex sites.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering foundation treatment construction monitoring, and specifically to a method for monitoring the retained vibration depth and retained vibration time of vibro-impacted gravel piles based on image recognition. Background Art

[0002] Since there are a large number of important water conservancy and hydropower infrastructure under construction in southwest my country and other places, and the mechanical properties of the natural foundation covering layer in these areas are often difficult to meet the bearing capacity and anti-liquefaction requirements of large structures, it is often necessary to treat the original natural foundation soil to improve its physical and mechanical properties.

[0003] The vibro-compacted stone column technology was proposed in the 1960s and has been widely used in the past few decades due to its reinforcement effect and rich anti-liquefaction connotation. At the same time, vibro-compacted stone columns have become one of the most economical foundation treatment methods because of their convenient construction and low technical requirements. Vibro-compacted stone columns mainly control the vibrator to fully retain vibration at each depth, so that the surrounding soil at each depth is vibrated and compacted, and a stone column body that meets the design requirements is formed. The key factor in controlling the quality of stone column formation is whether the vibrator can fully retain vibration at the designed depth, so that the compaction current and retention time during the retention stage meet the design requirements.

[0004] However, since the vibrator construction work is carried out below the surface, the depth of the vibrator foundation treatment has always been determined by the operator by observing the changes in the scale lines on the vibrator guide rod, and the vibration retention time at the corresponding depth is also determined by the operator in combination with the compaction current time history curve. This construction method has a strong human subjective factor, it is difficult to ensure the construction quality, and it requires high operating experience, energy and attention of the operator, which is not conducive to long-term construction. Summary of the invention

[0005] In order to solve the problems existing in the background technology, monitor the vibration depth of vibrating stone piles in real time, and determine the vibration time of the vibrator at different vibration depths, the purpose of the present invention is to provide a method for monitoring the vibration depth and vibration time of vibrating stone piles based on image recognition. By deploying a CCD industrial camera at the vibrating construction site, the position of the vibrator guide rod marker is captured in real time, and the vibration depth of the vibrator is located using an image recognition algorithm. Combined with the compaction current time history curve, the vibration depth and vibration time of the vibrator are further marked to assist operators in making judgments and construction.

[0006] The technical solution adopted by the present invention is as follows, comprising the following steps:

[0007] S1. First, select a marker and install it on the end of the vibrator guide rod; then construct an image training set by taking a certain number of construction site images;

[0008] S2. Setting up a CCD industrial camera and a computer storage and processing device at the construction site, using the CCD industrial camera to capture images of the vibrator during the entire vibratory construction process, and transmitting the images of the construction process to the computer storage and processing device in real time;

[0009] S3. Use computer storage and processing equipment to process the input image of the construction process, extract the real space coordinates of the marker in real time, obtain the real-time construction depth D of the vibrator, and draw a real-time vibratory depth-time curve.

[0010] S4. According to the real-time vibration impact depth time curve, the vibration retention time and the vibration retention depth are obtained by using the pre-determined compaction current design value and the vibration retention depth change rate threshold.

[0011] The step S1 is specifically as follows:

[0012] S11, select two markers, and fix both markers on the vibrator guide rod; one of the markers is located at the top of the vibrator guide rod, and the other marker is located at a position just exposed above the ground;

[0013] S12. Use a camera to shoot the construction site from multiple angles to obtain a number of images of the construction site in which the vibrator is captured; the images of the construction site include images in which no markers are captured and images in which markers are captured; then, the markers in the images in which the markers are captured are labeled; after the labeling is completed, an image training set is constructed using all the construction site images.

[0014] The step S2 is specifically as follows:

[0015] S21: First, the model and installation position of the CCD industrial camera are selected so that the CCD industrial camera can capture the vibratory construction process within the full length range of the vibrator;

[0016] S22: Then, the CCD industrial camera is placed at the construction site according to the selected installation position, and then wireless communication equipment and computer storage and processing equipment are set up at the construction site, and then the CCD industrial camera and the computer storage and processing equipment are connected;

[0017] S23: Start the CCD industrial camera, use the CCD industrial camera to take images of the vibrator during the entire vibratory construction process, and transmit the images of the vibratory construction process to a computer storage and processing device in real time;

[0018] The step S3 is specifically as follows:

[0019] S31, constructing a recognition and positioning model for locating markers in a computer storage and processing device, inputting the image training set in step 1 into the constructed recognition and positioning model for training, and obtaining the trained recognition and positioning model after the training is completed; the input of the recognition and positioning model is the construction site image in the image training set, and the output is the pixel coordinates of the marker in the construction site image;

[0020] S32, after the image of the construction process is transmitted to the computer storage and processing device, the target detection range corresponding to the i-th frame of the construction image in the image recognition algorithm is determined according to the size of the vibrator in the i-th frame of the construction image; the target detection range corresponding to the i-th frame of the construction image includes the outer contour of the vibrator in the i-th frame of the construction image;

[0021] S33, inputting the i-th frame of the construction image into a well-trained recognition and positioning model with an image recognition algorithm, the image recognition algorithm performs image recognition and tracking on the markers in the construction image within the target detection range, extracting the pixel coordinates of the markers; and then converting the pixel coordinates of the markers into spatial coordinates;

[0022] S34, using the spatial coordinates of the marker in the i-th frame construction image, obtain the construction depth D of the vibrator at the corresponding time t, and the construction depth D is obtained by processing according to the following formula:

[0023]

[0024] Where H is the height of the vibrator at time t, h is the height difference between the centers of the two markers at time t, z1 is the height of the marker at the bottom at time t, and z2 is the height of the marker at the top at time t;

[0025] S35, storing the construction depth D and time data of the vibrator corresponding to the i-th frame of the construction image in a computer storage and processing device, and outputting the construction depth D and time data to an operating platform through a wireless communication device for synchronous display;

[0026] S36. Repeat steps S32 to S35 multiple times to obtain the spatial coordinates of the markers in each frame of the construction image and the construction depth D of the vibrator at different times; then, establish a relationship curve between the construction depth D and time t as a real-time vibration depth time curve of the vibrator.

[0027] The step S4 is specifically as follows:

[0028] S41, collect the compaction current E of the vibrator in real time through computer storage and processing equipment, and establish the relationship curve between the compaction current E and time t as the real-time monitoring curve of the compaction current of the vibrator; combine the real-time vibration depth time curve of the vibrator, and calculate the compaction current E greater than E min And the rate of change of the retained vibration depth f is less than fmax The time interval is taken as the vibration retention period;

[0029] Among them, E min is the given design value of compaction current; f max is a given threshold value of the change rate of the retained vibration depth; in the above formula, the change rate of the retained vibration depth f of the vibrator is obtained according to the following formula:

[0030]

[0031] Where D is the construction depth;

[0032] S42, taking the total duration of each vibration retention time period as the vibration retention time of the vibrator, and taking the average construction depth D in each vibration retention time period as the vibration retention depth.

[0033] The color of the marker is different from that of the vibrator and the surrounding environment; the marker is completely wrapped around the periphery of the vibrator guide rod.

[0034] The height of the marker from the ground is not more than twice the given depth positioning accuracy ΔD, the height of the marker from the ground is not less than the preset vibro-compacted gravel pile construction depth, and the height difference between the two markers is a given height difference value h.

[0035] In order to realize automatic monitoring of the construction process of vibratory stone piles, the present invention proposes a real-time monitoring method for the vibration depth and vibration time of vibratory stone piles based on image recognition. The construction depth of the vibrator is monitored in real time by image recognition technology, and the vibration depth and vibration time of the vibrator are automatically calculated in combination with the construction depth time curve and the compaction current time curve. The vibration depth and vibration time are output to the operating platform in real time in the form of time interval marks for operators to read and judge, which ensures the construction quality while greatly reducing the workload of operators.

[0036] Beneficial effects of the present invention:

[0037] 1. The present invention realizes the real-time positioning of the vibration depth of the vibrator, which greatly improves the efficiency and accuracy and reduces the workload of the operators compared with the traditional method in which the operators observe the scale on the guide rod to judge the vibration depth.

[0038] 2. The present invention realizes the automatic calculation and marking of the vibration retention time of the vibro-compacting pile, which is more reasonable and accurate than the traditional method in which operators manually judge the vibration retention time. It greatly guarantees the construction quality of the vibro-compacted gravel pile and reduces the requirements for the operator's experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of on-site monitoring during the construction process of vibro-stone piles;

[0040] Figure 2 Flow chart of real-time positioning of vibration depth of vibro-stone piles;

[0041] Figure 3 Results of retained vibration depth and retained vibration time of vibro-compacted stone columns. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation cases. The following implementation cases are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the appended claims of the application.

[0043] The method of the present invention comprises the following steps:

[0044] S1. First, select a suitable marker and install it on the end of the vibrator guide rod; then construct an image training set by taking a certain number of construction site images;

[0045] S2. According to the position of the vibrator and the camera, as well as the height of the vibrator guide rod, etc., select a suitable CCD industrial camera and computer storage and processing equipment, set the CCD industrial camera and computer storage and processing equipment at the construction site, use the CCD industrial camera to capture images of the vibrator during the entire vibratory construction process, and transmit the images of the construction process to the computer storage and processing equipment in real time;

[0046] S3. Through image recognition technology and camera imaging principles, computer storage and processing equipment are used to process the input images of the construction process, extract the real space coordinates of the markers in real time, obtain the real-time construction depth D of the vibrator, and draw a real-time vibratory depth-time curve.

[0047] S4. According to the real-time vibration impact depth time curve, the vibration retention time and vibration retention depth are obtained through the predetermined compaction current design value and the vibration retention depth change rate threshold, and the calculation results are output to the operation platform for the operator to read in real time.

[0048] Wherein, step S1 is specifically as follows:

[0049] S11, select two markers, and fix both markers on the vibrator guide rod; one of the markers is located at the top of the vibrator guide rod, and the position of the other marker meets the following conditions: when the vibrator falls to the lowest point, the marker is just exposed above the ground;

[0050] S12. Use a camera to shoot the construction site from multiple angles to obtain a number of images of the construction site where the vibrator is captured; the construction site images include images where no markers are captured and images where markers are captured; then, the markers in the images where the markers are captured are labeled; after the labeling is completed, an image training set is constructed using all the construction site images.

[0051] The number of vibrator images is greater than 2000, and the shooting ratio, rotation angle and lighting conditions of each vibrator image should be as different as possible.

[0052] In the specific implementation, first determine the color of the vibrator guide rod on site, and select the corresponding contrasting color as the color of the marker according to the color of the vibrator guide rod. The marker can be an object that is easy to fix on the vibrator guide rod. The length of the marker should be sufficient to wrap the entire vibrator guide rod. Since the coordinates of the marker extracted by image recognition are the midpoint of the marker, the width of the marker should not exceed twice the allowable positioning accuracy ΔD of the vibrator. After determining the marker, the vibrator guide rod height and the pre-treated foundation depth should be combined to determine the part of the vibrator guide rod that is always exposed above the ground. The selected vibrator markers include two, the colors of the two should also be contrasting, and they are fixed at the end of the vibrator guide rod and the guide rod position just exposed above the ground. After the marker is fixed, the site should be photographed at different scales, rotation angles, and lighting, and multiple photos containing and excluding markers should be created. An image training set with no less than 2,000 photos is established, and the images containing markers should be marked with Label 1 and Label 2, respectively, such as Figure 1 shown.

[0053] The specific steps of step S2 are:

[0054] S21: First, the model and installation position of the CCD industrial camera are selected according to the height and position of the vibrator, so that the CCD industrial camera has a sufficient field of view to capture the vibratory construction process within the full length of the vibrator;

[0055] S22: Then, the CCD industrial camera is placed at the construction site according to the selected installation position, and then wireless communication equipment and computer storage and processing equipment are set up at the construction site, and then the CCD industrial camera and the computer storage and processing equipment are connected;

[0056] S23: Start the CCD industrial camera, use the CCD industrial camera to capture images of the entire vibratory construction process within the entire length range of the vibratory impactor, and transmit the images of the vibratory construction process to a computer storage and processing device in real time;

[0057] In the specific implementation, the lens type of the CCD camera is determined according to the distance between the construction position of the vibrator on site and the fixed point of the CCD camera to ensure that the camera has a sufficient field of view to capture the vibratory construction process within the full length of the vibrator. The computer storage and processing device is directly connected to the CCD camera. The optional systems of the computer storage and processing device include windows, mac, ios or android, etc. The system can run the image recognition algorithm to be adopted and the self-compiled algorithms such as the vibration depth calculation and vibration interval marking at high speed. The selected computer storage and processing device should have sufficient storage space to cache real-time image digital data. At the same time, the computer storage and processing device should be able to perform image recognition operations at a high rate, thereby realizing the rapid extraction of the coordinates of the marker. The system time of the CCD camera device is synchronized with the system time of the computer storage and processing device, and then the processed vibrator depth time history curve is transmitted to the vibrator operation platform in the form of a digital signal in real time using a wireless transmission device with high transmission, high noise resistance and low latency for real-time reading by the operator. The selected wireless communication equipment should have high transmission performance, high noise resistance and low latency, and be able to synchronously transmit the processed depth time history curve to the operating platform in real time to ensure time synchronization.

[0058] like Figure 2 As shown, step S3 is specifically as follows:

[0059] S31, constructing a recognition and positioning model for locating pixel coordinates of markers in a computer storage and processing device, inputting the image training set in step 1 into the constructed recognition and positioning model to train the image recognition algorithm, and obtaining the trained recognition and positioning model after the training is completed; the input of the recognition and positioning model is the construction site image in the image training set, and the output is the pixel coordinates of the markers in the construction site image. If the construction site image does not contain the markers, no pixel coordinates are output;

[0060] S32, after the image of the construction process (i.e., the construction image) is transmitted to the computer storage and processing device, the initial target detection range in the image recognition algorithm is determined according to the size of the vibrator in the i-th frame of the construction image, the image width range in the target detection range is w1-w2, and the image height range is h1-h2; the initial target detection range includes the outer contour of the vibrator in the i-th frame of the construction image;

[0061] In the specific implementation, an artificial range calibration is performed according to the size of the vibrator in the construction process image, and the image width (w1, w2) and height range (h1, h2) of the vibrator that may appear are defined as the target detection range in subsequent image recognition.

[0062] S33, inputting the i-th frame of the construction image into a well-trained recognition and positioning model with an image recognition algorithm, the image recognition algorithm in the recognition and positioning model performs image recognition and tracking on the markers in the construction image within the target detection range, and extracts the pixel coordinates of the markers; then converting the pixel coordinates of the markers into spatial coordinates;

[0063] In the specific implementation, i = 1, 2, 3...N; i represents the ordinal number of the construction image frame, and N is the total number of frames of the processed image. The image recognition algorithm is used to perform image recognition and tracking on the marker at the end of the vibrator, and the pixel coordinates of the marker are extracted. Further, according to the internal and external parameters of the camera, the camera imaging principle and the triangulation principle are used to convert the pixel coordinates of the two markers from bottom to top into spatial coordinates (x1, y1, z1) and (x2, y2, z2) respectively, to complete the processing of one frame of image. Among them, the image recognition algorithm selects a single-stage target detection algorithm, such as the Yolo algorithm, the RetinaNet algorithm or the SSD algorithm, etc., to ensure the accuracy while increasing the algorithm's computing speed as much as possible to ensure the real-time performance of the vibration depth positioning.

[0064] S34, using the spatial coordinates of the marker in the i-th frame construction image, obtain the construction depth D of the vibrator at the corresponding time t, and the construction depth D is obtained by processing according to the following formula:

[0065]

[0066] Where H is the height of the vibrator at time t, h is the height difference between the centers of the two markers at time t, z1 is the height of the marker at the bottom at time t, and z2 is the height of the marker at the top at time t;

[0067] S35, storing the construction depth D and time data (i.e., time t) of the vibrator corresponding to the i-th frame of the construction image in a computer storage and processing device, and outputting the construction depth D and time data to the operating platform through a wireless communication device for synchronous display;

[0068] S36. Calculate the vertical coordinate change Δz of the current marker according to the following formula:

[0069] Δz=D i -D i-1

[0070] Among them, D i is the construction depth corresponding to the i-th frame image; D0 is 0;

[0071] S37, updating the current target detection range according to the current vertical coordinate change Δz: after the update, the image width range in the target detection range remains unchanged, and the image height range is updated to h1 to h2-Δz;

[0072] That is, the left boundary h1 of the image height range remains unchanged, and the right boundary h2=h2-Δz. h2 is updated to the value of h2-Δz. After the update, the image height range is still displayed as h1~h2. The update is only a numerical update, and the sign remains unchanged.

[0073] In the specific implementation, before processing the next frame of image, the vertical coordinate change Δz of the marker at the end of the vibrator guide rod is first calculated, and based on this value, the width and height range of the original target detection is adjusted to ([w1, w2], [h1, h2-Δz]) through a self-compiled algorithm.

[0074] S38, returning to step S33 and repeatedly executing steps S32 to S37, using the updated target detection range to extract the spatial coordinates and construction depth of the next frame of construction image;

[0075] Repeat step S38 multiple times to obtain the spatial coordinates of the markers in each frame of the construction image and the construction depth D of the vibrator at different times; then, establish a relationship curve between the construction depth D and time t as a real-time vibration depth time curve of the vibrator.

[0076] In the specific implementation, step S4 is specifically as follows:

[0077] S41, collect the compaction current E of the vibrator in real time through computer storage and processing equipment, and establish the relationship curve between the compaction current E and time t as the real-time monitoring curve of the compaction current of the vibrator; combine the real-time vibration depth time curve of the vibrator, and calculate the compaction current E greater than E min And the rate of change of the retained vibration depth f is less than f max The time interval is taken as the vibration retention period;

[0078] Among them, E min is the given design value of compaction current; f max is a given threshold value of the change rate of the retained vibration depth; in the above formula, the change rate of the retained vibration depth f of the vibrator is obtained according to the following formula:

[0079]

[0080] Where D is the construction depth;

[0081] S42, taking the duration of each vibration retention time period as the vibration retention time of the vibrator in each vibration retention time period, and taking the average construction depth D in each vibration retention time period as the vibration retention depth in each vibration retention time period.

[0082] In the specific implementation, first determine the vibrator vibration depth change rate threshold f max and the compaction current design value E min The design value of compaction current is E minis the minimum threshold value determined by the gravel pile test, and the threshold value of the change rate of the retained vibration depth f max It is an indicator of the depth change rate of the vibrator operation, and its value is the maximum change rate artificially set according to the depth change rate of the actual vibrator during the vibration retention stage. Combining the real-time monitoring curve of the vibrator compaction current and the real-time vibration depth time curve determined in step S3, using a self-compiled algorithm, the compaction current E is greater than E min And the depth change rate f on the real-time vibration depth time curve is less than f max The time interval is marked on the real-time vibration depth time curve, and the time size within the marked segment is displayed.

[0083] The color of the marker has a relatively obvious color difference compared with the vibrator and the surrounding environment, such as the color of the marker is red, yellow, and blue; the marker can be completely wrapped around the outer circumference of the vibrator guide rod to ensure that the monitoring equipment can detect the marker in all directions.

[0084] The height of the marker from the ground shall not exceed twice the given depth positioning accuracy ΔD, the height of the marker from the ground shall not be less than the preset vibro-stone pile construction depth, and the height difference between the two markers shall be the given height difference value h.

[0085] Figure 1 This is the construction process of vibratory gravel piles at a certain engineering site. The vibrator guide rod is gray and has a total length of 25m. The designed gravel pile processing depth is 20m. The designed compaction current of the vibrator is 225A. The vibration depth varies within 30s and does not exceed 30cm. The designed positioning accuracy does not exceed 50cm.

[0086] In the first step, according to the characteristics of the vibrator, yellow and blue ribbons were selected as markers. The width of the ribbons did not exceed 1m and they were fixed at the end of the vibrator guide rod and 20m above the vibrator. After fixing, 5,000 photos of the site were randomly taken, of which 2,000 did not contain markers and the other 3,000 contained markers and the markers were marked as Label 1 and Label 2.

[0087] The second step is to start construction after the monitoring equipment is installed on site and conduct synchronous monitoring. The CCD camera transmits the real-time image digital signal to the computer storage and processing device, which performs the following steps: Figure 2 The process shown is used to complete the real-time calculation of the vibration depth.

[0088] Monitoring equipment includes CCD industrial cameras, computer storage and processing equipment, and wireless communication equipment.

[0089] The third step is to compare the vibration depth change rate f of the vibrator with the vibration depth change rate threshold of 1cm / s, and compare the compaction current time history data with the compaction current design value of 225A, automatically divide the time period that satisfies both the depth change rate f<1cm / s and the compaction current E>225A, give the time series that meets the requirements, and mark it with a blue background. At the same time, the data is synchronously transmitted to the crawler crane operating platform for real-time reading by the operator. The final data display result is shown in the figure. Figure 3 shown.

Claims

1. A method for monitoring the vibration depth and vibration time of vibro-stone piles based on image recognition, characterized in that: The following steps are involved: S1. First, select a marker and install it on the end of the vibrator guide rod; then construct an image training set by taking a certain number of construction site images; S2. Setting up a CCD industrial camera and a computer storage and processing device at the construction site, using the CCD industrial camera to capture images of the vibrator during the entire vibratory construction process, and transmitting the images of the construction process to the computer storage and processing device in real time; S3, using computer storage and processing equipment to process the input image of the construction process, extracting the real space coordinates of the marker in real time, and obtaining the real-time construction depth D of the vibrator, and drawing a real-time vibratory depth time curve; S4, according to the real-time vibration impact depth time curve, the vibration retention time and vibration retention depth are obtained by using the predetermined compaction current design value and the vibration retention depth change rate threshold; The step S1 is specifically as follows: S11, select two markers, and fix both markers on the vibrator guide rod; one of the markers is located at the top of the vibrator guide rod, and the position of the other marker meets the following conditions: when the vibrator falls to the lowest point, the marker is just exposed above the ground; S12, using a camera to shoot the construction site from multiple angles to obtain a number of construction site images in which the vibrator is captured; the construction site images include images in which the markers are not captured and images in which the markers are captured; then marking the markers in the images in which the markers are captured; after the marking is completed, using all the construction site images to construct an image training set; The step S3 is specifically as follows: S31, constructing a recognition and positioning model for locating markers in a computer storage and processing device, inputting the image training set in step 1 into the constructed recognition and positioning model for training, and obtaining the trained recognition and positioning model after the training is completed; the input of the recognition and positioning model is the construction site image in the image training set, and the output is the pixel coordinates of the marker in the construction site image; S32, after the image of the construction process is transmitted to the computer storage and processing device, the target detection range corresponding to the i-th frame of the construction image in the image recognition algorithm is determined according to the size of the vibrator in the i-th frame of the construction image; the target detection range corresponding to the i-th frame of the construction image includes the outer contour of the vibrator in the i-th frame of the construction image; S33, inputting the i-th frame of the construction image into a well-trained recognition and positioning model with an image recognition algorithm, the image recognition algorithm performs image recognition and tracking on the markers in the construction image within the target detection range, extracting the pixel coordinates of the markers; and then converting the pixel coordinates of the markers into spatial coordinates; S34, using the spatial coordinates of the marker in the i-th frame construction image, obtain the construction depth D of the vibrator at the corresponding time t, and the construction depth D is obtained by processing according to the following formula: Where H is the height of the vibrator at time t, h is the height difference between the centers of the two markers at time t, z1 is the height of the marker at the bottom at time t, and z2 is the height of the marker at the top at time t; S35, storing the construction depth D and time data of the vibrator corresponding to the i-th frame of the construction image in a computer storage and processing device, and outputting the construction depth D and time data to an operating platform through a wireless communication device for synchronous display; S36. Calculate the vertical coordinate change Δz of the current marker according to the following formula: Δz=D i -D i-1 Among them, D i is the construction depth corresponding to the i-th frame image; D0 is 0; S37, updating the current target detection range according to the current vertical coordinate change Δz: after the update, the image width range in the target detection range remains unchanged, and the image height range is updated to h1 to h2-Δz; S38, repeating steps S32 to S37, and using the updated target detection range to extract the spatial coordinates and construction depth of the next frame of construction image; Repeat step S38 multiple times to obtain the spatial coordinates of the markers in each frame of the construction image and the construction depth D of the vibrator at different times; then, establish a relationship curve between the construction depth D and the time t as a real-time vibration depth time curve of the vibrator; The step S4 is specifically as follows: S41, collect the compaction current E of the vibrator in real time through computer storage and processing equipment, and establish the relationship curve between the compaction current E and time t as the real-time monitoring curve of the compaction current of the vibrator; combine the real-time vibration depth time curve of the vibrator, and calculate the compaction current E greater than E min And the rate of change of the retained vibration depth f is less than f max The time interval is taken as the vibration retention time period; Among them, E min is the given design value of compaction current; f max is a given threshold value of the rate of change of the retained vibration depth; In the above formula, the change rate f of the vibration depth of the vibrator is obtained according to the following formula: Where D is the construction depth; S42, taking the total duration of each vibration retention time period as the vibration retention time of the vibrator, and taking the average construction depth D in each vibration retention time period as the vibration retention depth.

2. The method for monitoring the vibration depth and vibration time of vibro-stone piles based on image recognition according to claim 1, characterized in that: The step S2 is specifically as follows: S21: First, the model and installation position of the CCD industrial camera are selected so that the CCD industrial camera can capture the vibratory construction process within the full length range of the vibrator; S22: Then, the CCD industrial camera is placed at the construction site according to the selected installation position, and then wireless communication equipment and computer storage and processing equipment are set up at the construction site, and then the CCD industrial camera and the computer storage and processing equipment are connected; S23: Start the CCD industrial camera, use the CCD industrial camera to capture images of the vibrator during the entire vibratory construction process, and transmit the images of the vibratory construction process to a computer storage and processing device in real time.

3. The method for monitoring the vibration depth and vibration time of vibro-stone piles based on image recognition according to claim 1, characterized in that: The color of the marker is different from that of the vibrator and the surrounding environment; the marker is completely wrapped around the periphery of the vibrator guide rod.

4. The method for monitoring the vibration depth and vibration time of vibro-stone piles based on image recognition according to claim 1, characterized in that: The height of the marker from the ground is not more than twice the given depth positioning accuracy ΔD, the height of the marker from the ground is not less than the preset vibro-compacted gravel pile construction depth, and the height difference between the two markers is a given height difference value h.

Images