Model construction method and device based on holder scanning identification

Abstract

The invention discloses a model construction method and device based on holder scanning identification, and relates to the technical field of three-dimensional modeling. The method comprises the following steps: performing omnibearing scanning on a target scene through a holder to obtain visible light image data and laser point cloud data; enhancing the visible light image data to obtain an enhanced image; aligning the enhanced image with the point cloud data to obtain color information of the point cloud points; registering the point cloud data at different angles to generate a preliminary three-dimensional geometric model; and according to the color information of each point cloud point, coloring the three-dimensional geometric model to obtain a final target model. The visible light image is subjected to color restoration and contrast enhancement, so that image details are clearer, and texture mapping and coloring processing of a three-dimensional model are facilitated. According to the method, the advantages of the laser point cloud and the visible light image are combined, the geometric accuracy of the model is improved, and the sense of reality of the visual effect is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a model construction method and device based on pan-tilt scanning and recognition. Background Art

[0002] With the rapid development of modern intelligent technology, the application of 3D modeling technology has become more and more extensive, covering unmanned driving, robot navigation, virtual reality (VR), augmented reality (AR), building scanning, urban digitization and other fields. These application scenarios have put forward higher requirements for the accuracy, efficiency and realism of modeling. Laser point cloud is a common and efficient way to represent 3D models, which contains the 3D coordinates of each point, or additional information such as intensity. With only 3D coordinates, the geometric structure of the 3D model can be identified, but the appearance information of the 3D model cannot be more accurately and intuitively represented. In order to make up for the limitations of a single sensor, the fusion of LiDAR and visible light cameras has become a trend in recent years. By combining the precise depth information of LiDAR with the rich color and texture information of visible light cameras, a more comprehensive and accurate 3D model can be generated.

[0003] However, in low-light environments, the camera's white balance and color reproduction capabilities will be affected, which may cause image color distortion or color cast, resulting in low image contrast and difficulty in extracting details. Especially in dark areas, the color, texture, edges and object details in the image may disappear completely, and ultimately unnatural colors or lighting effects may appear when shading and texture mapping the 3D model, resulting in an unrealistic visual effect of the model. Summary of the invention

[0004] The purpose of the present invention is to solve the problem of low quality of low-light images mentioned in the above background technology, and to propose a model construction method and device based on pan-tilt scanning recognition.

[0005] In a first aspect of the present invention, a model construction method based on gimbal scanning recognition is provided, wherein the gimbal is equipped with a laser radar and a visible light camera; the method comprises:

[0006] The target scene is scanned in all directions by the pan-tilt platform to obtain visible light image data and laser point cloud data; the laser point cloud data includes coordinate information of the point cloud points;

[0007] De-noising the laser point cloud data to obtain denoised point cloud data;

[0008] Enhance the visible light image data to obtain an enhanced image;

[0009] Align the enhanced image and point cloud data to obtain the color information of the point cloud points;

[0010] Register point cloud data from different angles to generate a preliminary 3D geometric model;

[0011] The three-dimensional geometric model is colored according to the color information of each point cloud point to obtain the final target model.

[0012] Optionally, the enhancing the visible light image data to obtain an enhanced image includes:

[0013] Convert the target visible light image into the HSV space to obtain the original hue component, saturation component and brightness component; the target visible light image is any visible light image;

[0014] According to the adaptive gamma correction algorithm, the original brightness component is enhanced to obtain the target image:

[0015]

[0016] Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the brightness mean of the brightness component;

[0017] The target image and the original brightness component are fused to obtain an enhanced brightness component;

[0018] The original hue component, saturation component and enhanced brightness component form a new HSV image;

[0019] The new HSV image is converted to RGB space to obtain the enhanced image of the target visible light image.

[0020] Optionally, the step of fusing the target image with the original brightness component to obtain an enhanced brightness component includes:

[0021] According to the target visible light image, the weight coefficient is obtained:

[0022]

[0023] Wherein, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image, respectively; a, b, and c are constants;

[0024] According to the weight coefficient, the target image and the original brightness component are fused to obtain an enhanced brightness component:

[0025] V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y);

[0026] Among them, V2 is the enhanced brightness component.

[0027] Optionally, aligning the enhanced image and the point cloud data to obtain color information of the point cloud points includes:

[0028] Performing temporal alignment on the enhanced image and point cloud data, and extracting a pair of first image and first point cloud;

[0029] According to preset camera internal and external parameters, project the target point cloud point onto the surface of the first image to obtain the pixel position of the target point cloud point in the first image; the target point cloud point is any point cloud point in the first point cloud;

[0030] According to the pixel position, bilinear interpolation is used to obtain the RGB channel value of the target point cloud point.

[0031] Optionally, the gimbal has a horizontal rotation mechanism and a vertical rotation mechanism; the registering of point cloud data at different angles to generate a preliminary three-dimensional geometric model includes:

[0032] According to the pan / tilt rotation angle when acquiring the point cloud data, coordinate system conversion is performed, and point cloud data of different angles are projected into the preset global coordinate system to obtain the spatial coordinates of each point cloud point in the global coordinate system, thus obtaining a unified point cloud;

[0033] A triangular mesh is generated based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

[0034] A second aspect of the present invention provides a model construction device based on gimbal scanning and recognition, wherein the gimbal is equipped with a laser radar and a visible light camera; the device comprises:

[0035] A data acquisition module is used to perform an omnidirectional scan of the target scene through a pan / tilt platform to obtain visible light image data and laser point cloud data; the laser point cloud data includes coordinate information of the point cloud points;

[0036] Point cloud denoising module, used to denoise laser point cloud data to obtain denoised point cloud data;

[0037] An image enhancement module, used for enhancing visible light image data to obtain an enhanced image;

[0038] The color matching module is used to align the enhanced image and point cloud data to obtain the color information of the point cloud points;

[0039] The 3D surface generation module is used to register point cloud data from different angles and generate a preliminary 3D geometric model;

[0040] The model coloring module is used to color the three-dimensional geometric model according to the color information of each point cloud point to obtain the final target model.

[0041] Optionally, the image enhancement module includes:

[0042] A first conversion module is used to convert a target visible light image into an HSV space to obtain an original hue component, a saturation component and a lightness component; the target visible light image is any visible light image;

[0043] The gamma correction module is used to enhance the original brightness component according to the adaptive gamma correction algorithm to obtain the target image:

[0044]

[0045] Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the brightness mean of the brightness component;

[0046] An image fusion module is used to fuse the target image with the original brightness component to obtain an enhanced brightness component;

[0047] A replacement module is used to form a new HSV image by combining the original hue component, saturation component and enhanced lightness component;

[0048] The second conversion module is used to convert the new HSV image into RGB space to obtain an enhanced image of the target visible light image.

[0049] Optionally, the image fusion module includes:

[0050] The weight calculation module is used to obtain the weight coefficient according to the target visible light image:

[0051]

[0052] Wherein, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image, respectively; a, b, and c are constants;

[0053] The weighted fusion module is used to fuse the target image and the original brightness component according to the weight coefficient to obtain an enhanced brightness component:

[0054] V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y);

[0055] Among them, V2 is the enhanced brightness component.

[0056] Optionally, the color matching module includes:

[0057] A time sequence alignment module, used to perform time sequence alignment on the enhanced image and point cloud data, and extract a pair of first images and first point clouds;

[0058] A mapping module, used to project a target point cloud point onto the surface of the first image according to preset camera internal and external parameters, to obtain a pixel position of the target point cloud point in the first image; the target point cloud point is any point cloud point in the first point cloud;

[0059] The color determination module is used to obtain the RGB channel value of the target point cloud point by using bilinear interpolation according to the pixel position.

[0060] Optionally, the pan / tilt head has a horizontal rotation mechanism and a vertical rotation mechanism;

[0061] The three-dimensional surface generation module comprises:

[0062] The coordinate system conversion module is used to perform coordinate system conversion according to the pan / tilt rotation angle when acquiring point cloud data, project point cloud data of different angles into a preset global coordinate system, obtain the spatial coordinates of each point cloud point in the global coordinate system, and obtain a unified point cloud;

[0063] The meshing module is used to generate a triangular mesh based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

[0064] Beneficial effects of the present invention:

[0065] The present invention proposes a model construction method based on pan-tilt scanning and recognition, wherein the pan-tilt is equipped with a laser radar and a visible light camera; the method comprises: performing an all-round scanning of a target scene through the pan-tilt to obtain visible light image data and laser point cloud data; the laser point cloud data comprises coordinate information of point cloud points; denoising the laser point cloud data to obtain denoised point cloud data; enhancing the visible light image data to obtain an enhanced image; aligning the enhanced image and the point cloud data to obtain color information of the point cloud points; registering point cloud data at different angles to generate a preliminary three-dimensional geometric model; and coloring the three-dimensional geometric model according to the color information of each point cloud point to obtain a final target model.

[0066] Through the omnidirectional scanning and point cloud registration technology of the gimbal, point cloud data from different angles can be effectively integrated to generate a more complete and accurate 3D model. By enhancing the visible light image, the color reproduction and contrast in low-light environments can be effectively improved, making the image details clearer, which is helpful for the texture mapping and coloring of the 3D model. This method combines the advantages of laser point cloud and visible light image, which not only improves the geometric accuracy of the model, but also enhances the realism of the visual effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The present invention will be further described below in conjunction with the accompanying drawings.

[0068] Figure 1 A flowchart of a model construction method based on PTZ scanning recognition is provided for an embodiment of the present invention;

[0069] Figure 2 A structural diagram of a model construction device based on pan-tilt scanning recognition is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] The embodiment of the present invention provides a model construction method based on PTZ scanning recognition. Figure 1 , Figure 1 A flow chart of a model construction method based on PTZ scanning recognition provided by an embodiment of the present invention. The method comprises the following steps:

[0072] S101, perform an all-round scan of the target scene through the gimbal to obtain visible light image data and laser point cloud data.

[0073] S102, denoising the laser point cloud data to obtain denoised point cloud data.

[0074] S103, enhancing the visible light image data to obtain an enhanced image.

[0075] S104, aligning the enhanced image and the point cloud data to obtain color information of the point cloud points.

[0076] S105, registering the point cloud data at different angles to generate a preliminary three-dimensional geometric model.

[0077] S106, coloring the three-dimensional geometric model according to the color information of each point cloud point to obtain the final target model.

[0078] Among them, the gimbal is equipped with a laser radar and a visible light camera; the laser point cloud data includes point clouds at different angles and the coordinate information of their point cloud points; the visible light image data includes visible light images at different angles.

[0079] Based on a model construction method based on pan-tilt scanning and recognition provided by an embodiment of the present invention, through the pan-tilt omnidirectional scanning and point cloud registration technology, point cloud data from different angles can be effectively integrated to generate a more complete and accurate three-dimensional model. By enhancing the visible light image, the color restoration and contrast in low-light environments can be effectively improved, making the image details clearer, which is helpful for texture mapping and coloring of the three-dimensional model. This method combines the advantages of laser point cloud and visible light image, which not only improves the geometric accuracy of the model, but also enhances the realism of the visual effect.

[0080] In one embodiment, step S103 includes:

[0081] Step 1: Convert the target visible light image to the HSV space to obtain the original hue component, saturation component, and brightness component.

[0082] Step 2: According to the adaptive gamma correction algorithm, the original brightness component is enhanced to obtain the target image:

[0083]

[0084] Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the mean value of the lightness components.

[0085] Step three, fuse the target image and the original luminance component to obtain an enhanced luminance component.

[0086] Step 4: The original hue component, saturation component and enhanced lightness component are combined into a new HSV image.

[0087] Step 5: Convert the new HSV image to RGB space to obtain the enhanced image of the target visible light image.

[0088] The target visible light image is any visible light image.

[0089] In one implementation, by enhancing the brightness component, the visibility of dark areas and detail areas in the image can be effectively improved, the contrast of the image can be increased, and the details can be made more prominent, especially in areas with low light, which is crucial for subsequent 3D model generation, shading, and texture mapping.

[0090] In one implementation, the adjustment coefficient can be set to 0.5. The gamma parameter is adjusted according to the mean value of the brightness component to avoid the over-enhancement problem that may occur in traditional gamma correction. In adaptive gamma correction, the selection of gamma value is dynamic and can be flexibly adjusted according to the brightness of the specific image, making the brightness enhancement more natural and avoiding artifacts or distortion caused by over-processing.

[0091] In one embodiment, the target image and the original brightness component are fused to obtain an enhanced brightness component including:

[0092] Step 1: Get the weight coefficient based on the target visible light image:

[0093]

[0094] Among them, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image respectively; a, b, and c are constants.

[0095] Step 2: According to the weight coefficient, the target image and the original brightness component are fused to obtain the enhanced brightness component:

[0096] V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y);

[0097] Among them, V2 is the enhanced brightness component.

[0098] In one implementation, a, b, and c can select the weight values ​​of the standard brightness conversion formula, namely 0.299, 0.587, and 0.114. As can be seen from the above formula, after the brightness value is calculated, it is normalized so that the weight of the original image is 0 at the location where the brightness is 0; the weight of the original image is 1 at the location where the brightness is 255 (normalized to 1). Through the fused brightness components, different brightness areas can be enhanced in a targeted manner, so that the dark details in the image are highlighted, while retaining the edge details of the high-brightness area, and effectively avoiding overexposure or distortion in the highlight area. This enhancement of local adaptability can achieve a balance between the overall brightness and local details of the image, improving the overall image quality.

[0099] In one embodiment, step S104 includes:

[0100] Step 1: Temporally align the enhanced image and point cloud data to extract a pair of first image and first point cloud.

[0101] Step 2: Project the target point cloud point onto the surface of the first image according to the preset camera internal and external parameters to obtain the pixel position of the target point cloud point in the first image.

[0102] Step 3: According to the pixel position, bilinear interpolation is used to obtain the RGB channel value of the target point cloud point.

[0103] The target point cloud point is any point cloud point in the first point cloud.

[0104] In one implementation, by performing temporal alignment on the enhanced image and point cloud data, it is possible to ensure that the image and point cloud data are scanned or photographed at the same time, thus avoiding errors or inconsistencies caused by time differences. This is critical for subsequent point cloud coloring and texture mapping, ensuring that the generated 3D model has a high degree of matching in color and detail.

[0105] In one implementation, bilinear interpolation is used to calculate the RGB channel values ​​of the target point cloud point as follows:

[0106]

[0107] Among them, P i (X,Y) is the i-th channel value of the target point cloud point, i = 1, 2, 3, representing R, G, B channels respectively; (X,Y) is the pixel position of the target point cloud point; (X1,Y1), (X2,Y1), (X1,Y2), (X2,Y2) are the upper left pixel position, upper right pixel position, lower left pixel position and lower pixel position respectively. By projecting the point cloud data onto the image and using bilinear interpolation to obtain the RGB value, each point of the point cloud can obtain accurate color information. In this way, each point of the 3D model will have corresponding color data, and the generated model is not only accurate in geometric structure, but also more realistic in visual effect, especially in the presentation of details and lighting effects, the color transition of the model is smoother and more natural.

[0108] In one embodiment, the gimbal has a horizontal rotation mechanism and a vertical rotation mechanism; step S105 includes:

[0109] Step 1: According to the pan-tilt rotation angle when acquiring the point cloud data, coordinate system conversion is performed, and point cloud data of different angles are projected into a preset global coordinate system to obtain the spatial coordinates of each point cloud point in the global coordinate system, thereby obtaining a unified point cloud;

[0110] Step 2: Generate a triangular mesh based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

[0111] In one implementation, the center position of the gimbal can be used as the origin, and a global coordinate system can be established based on the posture when the horizontal rotation angle is 0 and the vertical pitch angle is 0. The transformation matrix of the coordinate system is obtained through the gimbal rotation angle, so that point cloud data at different angles can be included in the same coordinate system. By projecting point cloud data at different angles into a preset global coordinate system, it can be ensured that all data from different directions can be processed uniformly. This avoids conflicts and inconsistencies between multiple local coordinate systems, and ensures that the spatial relationship of all data is correct when the 3D model is generated.

[0112] The embodiment of the present invention provides a model construction method based on PTZ scanning recognition. Figure 2 , Figure 2 A structural diagram of a model construction device based on PTZ scanning recognition provided by an embodiment of the present invention. The device includes:

[0113] The data acquisition module is used to perform an all-round scan of the target scene through the pan-tilt platform to obtain visible light image data and laser point cloud data.

[0114] The point cloud denoising module is used to denoise the laser point cloud data to obtain denoised point cloud data.

[0115] The image enhancement module is used to enhance the visible light image data to obtain an enhanced image.

[0116] The color matching module is used to align the enhanced image and point cloud data to obtain the color information of the point cloud points.

[0117] The 3D surface generation module is used to register point cloud data from different angles and generate a preliminary 3D geometric model.

[0118] The model coloring module is used to color the three-dimensional geometric model according to the color information of each point cloud point to obtain the final target model.

[0119] Among them, the gimbal is equipped with a laser radar and a visible light camera; the laser point cloud data includes the coordinate information of the point cloud points.

[0120] Based on a model construction device based on pan-tilt scanning and recognition provided by an embodiment of the present invention, through the pan-tilt omnidirectional scanning and point cloud registration technology, point cloud data from different angles can be effectively integrated to generate a more complete and accurate three-dimensional model. By enhancing the visible light image, the color reproduction and contrast in low-light environments can be effectively improved, making the image details clearer, which is helpful for texture mapping and coloring processing of the three-dimensional model. This method combines the advantages of laser point cloud and visible light image, which not only improves the geometric accuracy of the model, but also enhances the realism of the visual effect.

[0121] In one embodiment, the image enhancement module includes:

[0122] The first conversion module is used to convert the target visible light image into the HSV space to obtain the original hue component, saturation component and brightness component.

[0123] The gamma correction module is used to enhance the original brightness component according to the adaptive gamma correction algorithm to obtain the target image:

[0124]

[0125] Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the mean value of the lightness components.

[0126] The image fusion module is used to fuse the target image with the original brightness component to obtain an enhanced brightness component.

[0127] The replacement module is used to compose a new HSV image from the original hue component, saturation component and enhanced lightness component.

[0128] The second conversion module is used to convert the new HSV image into RGB space to obtain an enhanced image of the target visible light image.

[0129] The target visible light image is any visible light image.

[0130] In one embodiment, the image fusion module includes:

[0131] The weight calculation module is used to obtain the weight coefficient according to the target visible light image:

[0132]

[0133] Among them, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image respectively; a, b, and c are constants.

[0134] The weighted fusion module is used to fuse the target image and the original brightness component according to the weight coefficient to obtain the enhanced brightness component:

[0135] V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y);

[0136] Among them, V2 is the enhanced brightness component.

[0137] In one embodiment, the color matching module includes:

[0138] The time sequence alignment module is used to perform time sequence alignment on the enhanced image and point cloud data, and extract a pair of first images and first point clouds.

[0139] The mapping module is used to project the target point cloud point onto the surface of the first image according to the preset camera internal and external parameters to obtain the pixel position of the target point cloud point in the first image.

[0140] The color determination module is used to obtain the RGB channel value of the target point cloud point according to the pixel position by using bilinear interpolation.

[0141] The target point cloud point is any point cloud point in the first point cloud.

[0142] In one embodiment, the gimbal has a horizontal rotation mechanism and a vertical rotation mechanism; the three-dimensional surface generation module includes:

[0143] The coordinate system conversion module is used to perform coordinate system conversion according to the pan / tilt rotation angle when acquiring point cloud data, project point cloud data of different angles into a preset global coordinate system, obtain the spatial coordinates of each point cloud point in the global coordinate system, and obtain a unified point cloud.

[0144] The meshing module is used to generate a triangular mesh based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

[0145] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A model construction method based on PTZ scanning recognition, characterized in that: The gimbal is equipped with a laser radar and a visible light camera; the method comprises: The target scene is scanned in all directions by the pan-tilt platform to obtain visible light image data and laser point cloud data; the laser point cloud data includes coordinate information of the point cloud points; De-noising the laser point cloud data to obtain denoised point cloud data; Enhance the visible light image data to obtain an enhanced image; Align the enhanced image and point cloud data to obtain the color information of the point cloud points; Register point cloud data from different angles to generate a preliminary 3D geometric model; The three-dimensional geometric model is colored according to the color information of each point cloud point to obtain the final target model.

2. The model construction method based on PTZ scanning recognition according to claim 1, characterized in that: The enhanced image obtained by enhancing the visible light image data includes: Convert the target visible light image into the HSV space to obtain the original hue component, saturation component and brightness component; the target visible light image is any visible light image; According to the adaptive gamma correction algorithm, the original brightness component is enhanced to obtain the target image: Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the brightness mean of the brightness component; The target image and the original brightness component are fused to obtain an enhanced brightness component; The original hue component, saturation component and enhanced brightness component form a new HSV image; The new HSV image is converted to RGB space to obtain the enhanced image of the target visible light image.

3. The model construction method based on PTZ scanning recognition according to claim 2, characterized in that: The target image and the original brightness component are fused to obtain an enhanced brightness component, including: According to the target visible light image, the weight coefficient is obtained: Wherein, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image, respectively; a, b, and c are constants; According to the weight coefficient, the target image and the original brightness component are fused to obtain an enhanced brightness component: V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y); Among them, V2 is the enhanced brightness component.

4. The model construction method based on PTZ scanning recognition according to claim 1, characterized in that: The enhanced image and point cloud data are aligned to obtain the color information of the point cloud points, including: Performing temporal alignment on the enhanced image and point cloud data, and extracting a pair of first image and first point cloud; According to preset camera internal and external parameters, project the target point cloud point onto the surface of the first image to obtain the pixel position of the target point cloud point in the first image; the target point cloud point is any point cloud point in the first point cloud; According to the pixel position, bilinear interpolation is used to obtain the RGB channel value of the target point cloud point.

5. The model construction method based on PTZ scanning recognition according to claim 1, characterized in that: The pan / tilt head is provided with a horizontal rotation mechanism and a vertical rotation mechanism; The step of registering point cloud data at different angles to generate a preliminary three-dimensional geometric model includes: According to the pan / tilt rotation angle when acquiring the point cloud data, coordinate system conversion is performed, and point cloud data of different angles are projected into the preset global coordinate system to obtain the spatial coordinates of each point cloud point in the global coordinate system, thus obtaining a unified point cloud; A triangular mesh is generated based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

6. A model construction device based on PTZ scanning recognition, characterized in that: The gimbal is equipped with a laser radar and a visible light camera; the device includes: A data acquisition module is used to perform an omnidirectional scan of the target scene through a pan / tilt platform to obtain visible light image data and laser point cloud data; the laser point cloud data includes coordinate information of the point cloud points; Point cloud denoising module, used to denoise laser point cloud data to obtain denoised point cloud data; An image enhancement module, used for enhancing visible light image data to obtain an enhanced image; The color matching module is used to align the enhanced image and point cloud data to obtain the color information of the point cloud points; The 3D surface generation module is used to register point cloud data from different angles and generate a preliminary 3D geometric model; The model coloring module is used to color the three-dimensional geometric model according to the color information of each point cloud point to obtain the final target model.

7. The model construction device based on PTZ scanning and recognition according to claim 6, characterized in that: The image enhancement module comprises: A first conversion module is used to convert a target visible light image into an HSV space to obtain an original hue component, a saturation component and a lightness component; the target visible light image is any visible light image; The gamma correction module is used to enhance the original brightness component according to the adaptive gamma correction algorithm to obtain the target image: Where V1 is the target image; (x, y) is the pixel position; V0 is the original brightness component; γ is the gamma parameter; α is the preset adjustment coefficient; V mean is the brightness mean of the brightness component; An image fusion module is used to fuse the target image with the original brightness component to obtain an enhanced brightness component; A replacement module is used to form a new HSV image by combining the original hue component, saturation component and enhanced lightness component; The second conversion module is used to convert the new HSV image into RGB space to obtain an enhanced image of the target visible light image.

8. The model construction device based on PTZ scanning and recognition according to claim 7, characterized in that: The image fusion module comprises: The weight calculation module is used to obtain the weight coefficient according to the target visible light image: Wherein, k is the required weight coefficient; (x, y) is the pixel position; g is the brightness value obtained by the RGB channel; R, G, B are the red, green, and blue channel values ​​of the target visible light image, respectively; a, b, and c are constants; The weighted fusion module is used to fuse the target image and the original brightness component according to the weight coefficient to obtain an enhanced brightness component: V2(x,y)=k(x,y)*V0(x,y)+(1-k(x,y))*V1(x,y); Among them, V2 is the enhanced brightness component.

9. The model construction device based on PTZ scanning and recognition according to claim 6, characterized in that: The color matching module comprises: A time sequence alignment module, used to perform time sequence alignment on the enhanced image and point cloud data, and extract a pair of first images and first point clouds; A mapping module, used to project a target point cloud point onto the surface of the first image according to preset camera internal and external parameters, to obtain a pixel position of the target point cloud point in the first image; the target point cloud point is any point cloud point in the first point cloud; The color determination module is used to obtain the RGB channel value of the target point cloud point by using bilinear interpolation according to the pixel position.

10. The model construction device based on PTZ scanning and recognition according to claim 6, characterized in that: The pan / tilt head is provided with a horizontal rotation mechanism and a vertical rotation mechanism; The three-dimensional surface generation module comprises: The coordinate system conversion module is used to perform coordinate system conversion according to the pan / tilt rotation angle when acquiring point cloud data, project point cloud data of different angles into a preset global coordinate system, obtain the spatial coordinates of each point cloud point in the global coordinate system, and obtain a unified point cloud; The meshing module is used to generate a triangular mesh based on the unified point cloud to obtain a three-dimensional plane as a preliminary three-dimensional geometric model.

Images