High-precision measurement method, laser radar and system

Abstract

The present invention discloses a high-precision measurement method, laser radar, and system. The method comprises: obtaining three-dimensional point cloud data and two-dimensional pixel data of a region to be measured, wherein the three-dimensional point cloud data is obtained by a laser radar, and the two-dimensional pixel data is obtained by an RGB lens, wherein a calibration relationship is provided between the laser radar and the RGB lens; preprocessing the three-dimensional point cloud data and two-dimensional pixel data of the region to be measured to obtain target point cloud data and target pixel data of a target to be measured in the region to be measured; obtaining three-dimensional calculated data of the target to be measured in the target pixel data based on the calibration relationship and the target point cloud data; and calibrating the target point cloud data using the three-dimensional calculated data to obtain measurement data of the target to be measured. The present invention can improve the accuracy of laser radar measurement data, particularly for the measurement of indoor doors and windows, significantly improving the accuracy.

Description

Technical Field

[0001] The present invention relates to a high-precision actual measurement method, a laser radar and a system. Background Art

[0002] Actual measurement refers to a method that uses measurement tools to conduct on-site testing and measurement to truly reflect product quality data. According to relevant quality acceptance standards, the error in measurement control engineering quality data is within the range allowed by national housing construction standards.

[0003] The project development stages covered by actual measurements include the main structure, masonry, plastering, equipment installation, and finishing. The scope of measurement includes concrete structure, masonry, plastering, waterproofing, doors and windows, painting, and finishing.

[0004] Existing measurement tools have the disadvantage of low measurement accuracy, especially large measurement errors in the measurement of doors and windows. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of low measurement accuracy of the actual measurement tools in the existing technology, especially the large measurement errors in the measurement of doors and windows, and to provide a high-precision actual measurement method, laser radar and system that can improve the accuracy of laser radar measurement data, especially the measurement of indoor doors and windows, with greatly improved accuracy.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] A high-precision actual measurement method, comprising:

[0008] Acquire three-dimensional point cloud data and two-dimensional pixel data of a to-be-measured area, wherein the three-dimensional point cloud data is acquired by a laser radar, and the two-dimensional pixel data is acquired by an RGB lens, wherein a calibration relationship is established between the laser radar and the RGB lens;

[0009] Preprocessing the three-dimensional point cloud data and the two-dimensional pixel data of the area to be measured to obtain target point cloud data and target pixel data of the target to be measured in the area to be measured;

[0010] Acquire three-dimensional calculation data of the target to be measured in the target pixel data according to the calibration relationship and the target point cloud data;

[0011] The target point cloud data is calibrated using the three-dimensional calculation data to obtain measurement data of the target to be measured.

[0012] Preferably, the actual measurement method includes:

[0013] Performing semantic segmentation on the three-dimensional point cloud data of the area to be measured, dividing the three-dimensional point cloud data of the area to be measured into a plurality of point cloud plane models, wherein the preprocessing includes the semantic segmentation;

[0014] Obtain the point cloud plane model where the target to be measured is located;

[0015] Acquire a preprocessing target, wherein the preprocessing target is two-dimensional pixel data corresponding to the point cloud plane model where the target to be measured is located;

[0016] Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship;

[0017] The three-dimensional calculation data is obtained according to the pre-processing target projected on the point cloud plane model.

[0018] Preferably, the obtaining of the preprocessing target includes:

[0019] Selecting or identifying target pixel data of the target to be measured from the two-dimensional pixel data;

[0020] Obtaining the wall edge of the wall where the target pixel data is located in the two-dimensional pixel data;

[0021] Pixel data of the wall where the target pixel data is located is obtained according to the wall edge as the preprocessing target.

[0022] Preferably, the actual measurement method includes:

[0023] In the two-dimensional pixel data, the wall edge is identified in all directions based on the texture information and starting from the target pixel data.

[0024] Preferably, the target to be measured is a door frame or a window frame, and the obtaining of the three-dimensional calculation data according to the pre-processed target projected on the point cloud plane model includes:

[0025] Identify corner feature points of the target to be measured after projection and preprocessing the target;

[0026] The length between the corner feature points is obtained according to the correspondence between the point cloud plane model and the preprocessing target and the spatial coordinates of the point cloud plane model, and the three-dimensional calculation data includes the length between the corner feature points.

[0027] Preferably, the actual measurement method includes:

[0028] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0029] Align the 3D point cloud data and 2D initial data of the area to be measured;

[0030] The three-dimensional point cloud data of the area to be measured is matched with the two-dimensional initial data of all the areas to be measured, and the optimal two-dimensional initial data is selected as the two-dimensional pixel data by minimizing the weighted energy error.

[0031] Preferably, the actual measurement method includes:

[0032] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0033] Obtain target point cloud data of the target to be measured;

[0034] Obtaining the moment when the laser radar is aimed at the target to be measured;

[0035] Two-dimensional initial data matching the moment is acquired as the two-dimensional pixel data.

[0036] Preferably, the actual measurement method includes:

[0037] identifying edges of preprocessed objects in two-dimensional pixel data;

[0038] Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship;

[0039] Obtaining whether the lengths of the two edge points projected on the point cloud plane model match the lengths of the corresponding points on the point cloud plane model; if not, ignoring the current edge pixel and performing the step of identifying the edge of the pre-processed target in the two-dimensional pixel data again; if so, stretching the pre-processed target and covering it on the point cloud plane model;

[0040] Acquiring the three-dimensional calculation data from the pre-processed target after stretching;

[0041] Determine whether the 3D calculation data matches the target point cloud data. If not, use the 3D calculation data as the measurement data of the target to be measured.

[0042] The present invention also provides a high-precision laser radar system, which includes a laser radar, an RGB lens, a processing module and a server. The laser radar system is used to implement the actual measurement method described above.

[0043] The present invention also provides a laser radar, which is used in the laser radar system as described above.

[0044] Based on the common sense in this field, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0045] The positive progress effect of the present invention is:

[0046] The present invention can improve the accuracy of laser radar measurement data, especially the measurement of indoor doors and windows, greatly improving the accuracy.

[0047] The present invention can improve the accuracy of data measured solely by lidar. The accuracy of existing lidar is affected by many factors, such as radar accuracy, image resolution, and line fitting variance. Tests have shown that under ideal conditions, the error of the measurement results of the present invention can be reduced from 5mm to 1mm. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the actual measurement method of Example 1 of the present invention. DETAILED DESCRIPTION

[0049] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0050] Example 1

[0051] This embodiment provides a high-precision laser radar system, which includes a laser radar, an RGB lens, a processing module, and a server.

[0052] In this embodiment, the laser radar includes the RGB lens, which is arranged on the outer shell of the laser radar, fixed to the laser radar and calibrated with each other.

[0053] In other implementations, the RGB lens can be fixed on the laser radar as an external device, that is, fixed separately, and then calibrated with the laser radar through a calibration method.

[0054] The laser radar is used to obtain three-dimensional point cloud data of a test area, and the RGB lens is used to obtain two-dimensional pixel data of the test area. A calibration relationship is included between the laser radar and the RGB lens.

[0055] The processing module is used to pre-process the three-dimensional point cloud data and the two-dimensional pixel data of the area to be measured to obtain the target point cloud data and the target pixel data of the target to be measured in the area to be measured;

[0056] The area to be tested is a room and a wall.

[0057] The 3D point cloud data includes the 3D point cloud data of the walls of the room. The 3D point cloud data of a wall will have some interference data after scanning, such as other walls connected to the wall, ceiling, ground, etc.

[0058] The two-dimensional pixel data is a two-dimensional image, which is photographed at preset time intervals during the scanning process to obtain a plurality of two-dimensional pixel data.

[0059] In this embodiment, the target to be measured is a door or a window, the target point cloud data is the three-dimensional data of the door, and the target pixel data is the two-dimensional data of the door.

[0060] The processing module is used to obtain three-dimensional calculation data of the target to be measured in the target pixel data according to the calibration relationship and the target point cloud data;

[0061] The processing module is used to calibrate the target point cloud data using the three-dimensional calculation data to obtain measurement data of the target to be measured.

[0062] The server is used to receive the measurement data for use by other processing modules and processing terminals.

[0063] The processing module can be a processing terminal such as a PC or a notebook, or it can be a computing chip integrated in the laser radar.

[0064] Furthermore, the processing module is used to:

[0065] Performing semantic segmentation on the three-dimensional point cloud data of the area to be measured, dividing the three-dimensional point cloud data of the area to be measured into a plurality of point cloud plane models, wherein the preprocessing includes the semantic segmentation;

[0066] Obtain the point cloud plane model where the target to be measured is located;

[0067] Acquire a preprocessing target, wherein the preprocessing target is two-dimensional pixel data corresponding to the point cloud plane model where the target to be measured is located;

[0068] This embodiment first obtains a three-dimensional model (point cloud plane model) and a two-dimensional image (pre-processed target) of the wall where the target to be measured is located.

[0069] The acquisition of the pre-processing target can be recognition or manual selection. Recognition can be recognition of the edge of the wall.

[0070] Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship;

[0071] By projecting the preprocessed target onto the point cloud plane model, the three-dimensional data or coordinates of the pixels on the preprocessed target can be obtained, thereby obtaining the spatial distance between each pixel point on the preprocessed target.

[0072] The target pixel data is the two-dimensional pixel data of the target to be measured, which can record the two-dimensional coordinates of each pixel point. The three-dimensional calculation data is the three-dimensional data converted from the two-dimensional image obtained by projecting the preprocessed target on the point cloud plane model.

[0073] The three-dimensional calculation data is obtained according to the pre-processing target projected on the point cloud plane model.

[0074] The three-dimensional calculation data is three-dimensional data obtained using a preprocessing target.

[0075] Furthermore, the processing module is used to:

[0076] Selecting or identifying target pixel data of the target to be measured from the two-dimensional pixel data;

[0077] Obtaining the wall edge of the wall where the target pixel data is located in the two-dimensional pixel data;

[0078] Pixel data of the wall where the target pixel data is located is obtained according to the wall edge as the preprocessing target.

[0079] The processing module is used for:

[0080] In the two-dimensional pixel data, the wall edge is identified in all directions based on the texture information and starting from the target pixel data.

[0081] The target to be measured is a door frame or a window frame, and the processing module is used to:

[0082] Identify corner feature points of the target to be measured after projection and preprocessing the target;

[0083] The length between the corner feature points is obtained according to the correspondence between the point cloud plane model and the preprocessing target and the spatial coordinates of the point cloud plane model, and the three-dimensional calculation data includes the length between the corner feature points.

[0084] The processing module is used for:

[0085] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0086] Align the 3D point cloud data and 2D initial data of the area to be measured;

[0087] The three-dimensional point cloud data of the area to be measured is matched with the two-dimensional initial data of all the areas to be measured, and the optimal two-dimensional initial data is selected as the two-dimensional pixel data by minimizing the weighted energy error.

[0088] The processing module is used for:

[0089] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0090] Obtain target point cloud data of the target to be measured;

[0091] Obtaining the moment when the laser radar is aimed at the target to be measured;

[0092] Two-dimensional initial data matching the moment is acquired as the two-dimensional pixel data.

[0093] The processing module is used for:

[0094] identifying edges of preprocessed objects in two-dimensional pixel data;

[0095] Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship;

[0096] Obtaining whether the lengths of the two edge points projected on the point cloud plane model match the lengths of the corresponding points on the point cloud plane model; if not, ignoring the current edge pixel and performing the step of identifying the edge of the pre-processed target in the two-dimensional pixel data again; if so, stretching the pre-processed target and covering it on the point cloud plane model;

[0097] Acquiring the three-dimensional calculation data from the pre-processed target after stretching;

[0098] For example, the point cloud plane model is a rectangular model, and the preprocessing target is usually a trapezoid. The left bottom side of the trapezoid is aligned with the left boundary of the rectangular model, and the right bottom side of the trapezoid is aligned with the right boundary of the rectangular model for stretching.

[0099] The pre-processing target is the two-dimensional pixel data corresponding to the point cloud plane model where the target to be measured is located;

[0100] Determine whether the 3D calculation data matches the target point cloud data. If not, use the 3D calculation data as the measurement data of the target to be measured.

[0101] Using the above-mentioned laser radar system and laser radar, this embodiment further provides a method for actual measurement, including:

[0102] Step 100: Acquire 3D point cloud data and 2D pixel data of a to-be-measured area, wherein the 3D point cloud data is acquired by a laser radar, and the 2D pixel data is acquired by an RGB lens, wherein a calibration relationship is established between the laser radar and the RGB lens;

[0103] Step 101: Preprocess the three-dimensional point cloud data and the two-dimensional pixel data of the area to be measured to obtain target point cloud data and target pixel data of the target to be measured in the area to be measured;

[0104] Step 102: obtaining three-dimensional calculation data of the target to be measured in the target pixel data according to the calibration relationship and the target point cloud data;

[0105] Step 103: calibrate the target point cloud data using the three-dimensional calculation data to obtain measurement data of the target to be measured.

[0106] Step 101 specifically includes:

[0107] Step 1011: performing semantic segmentation on the three-dimensional point cloud data of the area to be measured, dividing the three-dimensional point cloud data of the area to be measured into a plurality of point cloud plane models, wherein the preprocessing includes the semantic segmentation;

[0108] Step 1012: Obtain a point cloud plane model of the target to be measured;

[0109] Step 1013: Acquire a pre-processed target, where the pre-processed target is two-dimensional pixel data corresponding to the point cloud plane model where the target to be measured is located;

[0110] Step 102 specifically includes:

[0111] Step 1021: projecting the pre-processed target onto the point cloud plane model according to the calibration relationship;

[0112] Step 1022: Acquire the three-dimensional calculation data according to the pre-processing target projected on the point cloud plane model.

[0113] Furthermore, step 1013 includes:

[0114] Selecting or identifying target pixel data of the target to be measured from the two-dimensional pixel data;

[0115] Obtaining the wall edge of the wall where the target pixel data is located in the two-dimensional pixel data;

[0116] Pixel data of the wall where the target pixel data is located is obtained according to the wall edge as the preprocessing target.

[0117] Specifically, the actual measurement method includes:

[0118] In the two-dimensional pixel data, the wall edge is identified in all directions based on the texture information and starting from the target pixel data.

[0119] The target to be measured is a door frame or a window frame, and step 1022 includes:

[0120] Identify corner feature points of the target to be measured after projection and preprocessing the target;

[0121] The length between the corner feature points is obtained according to the correspondence between the point cloud plane model and the preprocessing target and the spatial coordinates of the point cloud plane model, and the three-dimensional calculation data includes the length between the corner feature points.

[0122] Specifically, the actual measurement method includes:

[0123] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0124] Align the 3D point cloud data and 2D initial data of the area to be measured;

[0125] The three-dimensional point cloud data of the area to be measured is matched with the two-dimensional initial data of all the areas to be measured, and an optimal two-dimensional initial data is selected as the two-dimensional pixel data by minimizing the weighted energy error.

[0126] Specifically, the actual measurement method includes:

[0127] Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data;

[0128] Obtain target point cloud data of the target to be measured;

[0129] Obtaining the moment when the laser radar is aimed at the target to be measured;

[0130] Two-dimensional initial data matching the moment is acquired as the two-dimensional pixel data.

[0131] Preferably, the actual measurement method includes:

[0132] identifying edges of preprocessed objects in two-dimensional pixel data;

[0133] Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship;

[0134] Obtaining whether the lengths of the two edge points projected on the point cloud plane model match the lengths of the corresponding points on the point cloud plane model; if not, ignoring the current edge pixel and performing the step of identifying the edge of the pre-processed target in the two-dimensional pixel data again; if so, stretching the pre-processed target and covering it on the point cloud plane model;

[0135] Acquiring the three-dimensional calculation data from the pre-processed target after stretching;

[0136] Determine whether the 3D calculation data matches the target point cloud data. If not, use the 3D calculation data as the measurement data of the target to be measured.

[0137] Although specific embodiments of the present invention have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, and such changes and modifications are intended to fall within the scope of the present invention.

Claims

1. A high-precision measurement method, characterized in that: The actual measurement method includes: Acquire three-dimensional point cloud data and two-dimensional pixel data of a to-be-measured area, wherein the three-dimensional point cloud data is acquired by a laser radar, and the two-dimensional pixel data is acquired by an RGB lens, wherein a calibration relationship is established between the laser radar and the RGB lens; Preprocessing the three-dimensional point cloud data and the two-dimensional pixel data of the area to be measured to obtain target point cloud data and target pixel data of the target to be measured in the area to be measured; Acquire three-dimensional calculation data of the target to be measured in the target pixel data according to the calibration relationship and the target point cloud data; The target point cloud data is calibrated using the three-dimensional calculation data to obtain measurement data of the target to be measured; the actual measurement method includes: Performing semantic segmentation on the three-dimensional point cloud data of the area to be measured, dividing the three-dimensional point cloud data of the area to be measured into a plurality of point cloud plane models, wherein the preprocessing includes the semantic segmentation; Obtain the point cloud plane model where the target to be measured is located; Obtaining a preprocessing target, wherein the preprocessing target is two-dimensional pixel data corresponding to the point cloud plane model where the target to be measured is located; Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship; Acquiring the three-dimensional calculation data according to the pre-processing target projected on the point cloud plane model; The actual measurement method includes: Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data; Obtain target point cloud data of the target to be measured; Obtaining the moment when the laser radar is aimed at the target to be measured; Acquire two-dimensional initial data matching the moment as the two-dimensional pixel data; The actual measurement method includes: identifying edges of preprocessed objects in two-dimensional pixel data; Projecting the preprocessed target onto the point cloud plane model according to the calibration relationship; Obtain whether the lengths of the two edge points projected on the point cloud plane model match the lengths of the corresponding points on the point cloud plane model. If not, ignore the current edge pixel and perform the step of identifying the edge of the pre-processed target in the two-dimensional pixel data again. If so, stretch the pre-processed target and cover it on the point cloud plane model. Acquiring the three-dimensional calculation data from the pre-processed target after stretching; Determine whether the 3D calculation data matches the target point cloud data. If not, use the 3D calculation data as the measurement data of the target to be measured.

2. The actual measurement method according to claim 1, characterized in that: The obtaining of the preprocessing target comprises: Selecting or identifying target pixel data of the target to be measured from the two-dimensional pixel data; Obtaining the wall edge of the wall where the target pixel data is located in the two-dimensional pixel data; Pixel data of the wall where the target pixel data is located is obtained according to the wall edge as the preprocessing target.

3. The actual measurement method according to claim 2, characterized in that: The actual measurement method includes: In the two-dimensional pixel data, the wall edge is identified in all directions based on the texture information and starting from the target pixel data.

4. The actual measurement method according to claim 2, characterized in that: The target to be measured is a door frame or a window frame, and obtaining the three-dimensional calculation data according to the pre-processed target projected on the point cloud plane model includes: Identify corner feature points of the target to be measured after projection and preprocessing the target; The length between the corner feature points is obtained according to the correspondence between the point cloud plane model and the preprocessing target and the spatial coordinates of the point cloud plane model, and the three-dimensional calculation data includes the length between the corner feature points.

5. The actual measurement method according to claim 1, characterized in that: The actual measurement method includes: Obtaining three-dimensional point cloud data of a region to be measured and a plurality of two-dimensional initial data; Align the 3D point cloud data and 2D initial data of the area to be measured; The three-dimensional point cloud data of the area to be measured is matched with the two-dimensional initial data of all the areas to be measured, and the optimal two-dimensional initial data is selected as the two-dimensional pixel data by minimizing the weighted energy error.

6. A high-precision laser radar system, characterized in that: The laser radar system includes a laser radar, an RGB lens, a processing module and a server, and the laser radar system is used to implement the actual measurement method as described in any one of claims 1 to 5.

7. A laser radar, characterized in that: The laser radar is used in the laser radar system as claimed in claim 6.

Images