Point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization

Abstract

The present invention belongs to the technical field of point cloud data processing, and discloses a point cloud smooth fitting method based on local density parameter adjustment and intensity optimization. The original point cloud data is obtained and pre-processed to perform local density calculation, and the search radius strategy and the neighborhood are dynamically adjusted. After the boundary judgment, the boundary protection mechanism of the boundary point or the calculation of the joint weight and the update smooth fitting are performed to obtain the point cloud visualization result map after fitting and reconstruction. The present invention effectively solves the deficiencies of the prior art in sparse point cloud smoothing, noise robustness, intensity feature retention and boundary feature blurring, and can dynamically adapt to the sparsity and non-uniform distribution of point clouds, significantly improve the accuracy of point cloud smooth fitting and the optimization ability of intensity information, and provide a more accurate and efficient technical solution for terrain mapping, underwater target detection and environmental modeling of forward-looking three-dimensional sonar.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data processing, and in particular relates to a point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization. Background Art

[0002] Forward-looking 3D sonar is a highly efficient device, which is widely used in ocean detection and underwater environment modeling. It mainly uses sound wave echoes to obtain 3D point cloud information of target objects. Since the acquisition of sonar data is limited by underwater propagation conditions (such as sound wave attenuation, scattering, etc.) and sampling resolution, the generated point cloud is usually sparse and non-uniformly distributed, especially in a longer detection range; sonar point cloud data is often accompanied by high noise interference, especially in complex underwater environments (such as sediments, clutter interference, etc.), the uncertainty of sonar echo signals will cause isolated points or abnormal points in the point cloud; and the sonar point cloud contains echo intensity information, which can reflect the surface material, hardness and sound wave reflection characteristics of the target. Intensity information is of great significance for underwater target detection and terrain modeling, but existing technologies often fail to make full use of this feature; at the same time, due to the limited acquisition accuracy of sonar data, the boundary area of ​​the point cloud is often irregular and incomplete, and the boundary features are easily over-weakened in smoothing processing. It can be seen that 3D sonar point cloud data has the characteristics of sparseness, high noise, additional intensity information and blurred boundary features.

[0003] For the processing of 3D sonar point cloud data, current technical methods mainly focus on conventional point cloud processing algorithms, such as moving least squares (MLS) and smoothing filtering. However, these methods have shortcomings when processing 3D sonar point cloud data: the traditional MLS method uses a fixed search radius to smooth the point cloud, but for the sparsity and density non-uniformity of point clouds in forward-looking sonar data, this method is difficult to adapt to the changing local density distribution, often resulting in insufficient neighborhood or excessive smoothing errors; at the same time, forward-looking 3D sonar point clouds are affected by environmental noise and multipath effects, and there are often isolated points or outliers in the point cloud; in addition, most point cloud smoothing methods are only based on geometric distribution features, and the intensity information in the 3D sonar point cloud is not included in the optimization process, resulting in the fitting results being unable to fully retain the important physical characteristics of the sonar data; finally, the existing methods have poor protection capabilities for the boundaries of sonar point clouds, and when the neighborhood is insufficient, it is easy to cause the boundary features to be distorted or lost, reducing the accuracy of underwater environment modeling. Summary of the invention

[0004] In view of the shortcomings of the background technology, the purpose of the present invention is to propose a point cloud smooth fitting method based on local density adaptive parameter adjustment and intensity optimization, and combine it with dynamic expansion neighborhood, boundary protection or smooth fitting mechanism to improve the processing quality and robustness of sparse point clouds.

[0005] The technical solution adopted by the present invention is a point cloud smoothing fitting method based on local density parameter adjustment and strength optimization, and the steps are as follows:

[0006] Step S1, obtaining data: obtaining original point cloud data from a forward-looking 3D sonar device;

[0007] Step S2, point cloud denoising: using intensity filtering and statistical filtering algorithms to perform denoising on the original point cloud;

[0008] Step S3, calculating local density: calculating the local density of the point cloud based on the average distance from the neighborhood points to the center point;

[0009] Step S4, dynamically adjust the search radius: dynamically adjust the search radius according to the local density to adapt the algorithm to sparse and non-uniformly distributed point clouds;

[0010] Step S5, dynamically expand the neighborhood: when the number of neighborhood points is insufficient, dynamically expand the area based on the search radius to ensure the stability of the fitting process;

[0011] Step S6, boundary judgment and obtaining point cloud result map:

[0012] Boundary judgment is based on the situation of neighboring points. Different situations of neighboring points have different judgment modes. There are two specific situations: a. Insufficient neighboring points; b. Sufficient neighboring points;

[0013] Step S6a, for the boundary points with insufficient neighborhood points, boundary protection is performed, the original geometric position and intensity value are directly retained, and the boundary feature blur or loss caused by excessive smoothing is avoided, and the point cloud result image is visualized;

[0014] Step S6b, for the boundary points with enough neighborhood points, calculate the joint weight and update the smooth fit:

[0015] Step S6b1, based on the calculation of the geometric weight and the intensity weight, a joint weight calculation method is performed to simultaneously optimize the geometric position and intensity value of the point cloud;

[0016] Step S6b2, performing smooth fitting on the geometric position and intensity value of the point cloud based on the joint weight, so as to update the geometric position and intensity value of the point cloud;

[0017] Step S6b3, generating a point cloud after fitting and reconstruction: visualizing the result of the smooth fitting to obtain a fitting result graph.

[0018] Preferably, the process steps for calculating the local density in step S3 are as follows:

[0019] For any point , the formula for defining its neighborhood with radius r is as follows:

[0020] ;

[0021] In the formula is the position vector of point i, is the position vector of point j, is the Euclidean distance, is the point set in the neighborhood with radius r and centered at point p;

[0022] Next, the local density Based on the average distance calculation from the neighborhood point to the center point, the calculation formula is:

[0023] ;

[0024] in, is the number of neighborhood points, is to prevent division by zero errors;

[0025] The larger the local density, the The smaller the distance to the neighboring points, the point is distributed in a dense area; the smaller the local density, the smaller the point The larger the distance of the neighboring points is, the more sparsely distributed the points are.

[0026] Preferably, the process steps of dynamically adjusting the search radius in step S4 are as follows:

[0027] Search Radius The adjustment formula is as follows:

[0028] ;

[0029] in, is the base search radius, is the adjustment scale factor, , are the minimum and maximum values ​​of density respectively;

[0030] In dense areas (higher density), the search radius Reduce to avoid excessive smoothing due to too many neighborhood points; in sparse areas (low density), the search radius Increase to ensure that enough neighborhood points participate in smoothing.

[0031] Preferably, the process steps of dynamically expanding the neighborhood in step S5 are as follows:

[0032] The initial search radius is Radius, each expansion increment is , the extended condition formula is as follows:

[0033] ;

[0034] In the formula is the minimum number of neighborhood points threshold, is the maximum value of the search radius;

[0035] The expansion process is based on the dynamically adjusted search radius, on which the neighborhood is expanded. The neighborhood points are recalculated each time the search radius is expanded until the number of neighborhood points meets the requirement or the search radius reaches the upper limit.

[0036] Preferably, the process steps of boundary judgment in step S6 are as follows:

[0037] Step S6a, for the boundary points with insufficient neighboring points, the boundary protection formula is as follows:

[0038] ;

[0039] Preferably, in step S6b, for boundary points with sufficient neighborhood points, the calculation method of the joint weight in step S6b1 is as follows:

[0040] Geometry Weight The definition formula is as follows:

[0041] ;

[0042] Where exp is the exponential function;

[0043] Intensity Weight The definition formula is as follows:

[0044] ;

[0045] In the formula is the standard deviation of the intensity differences, which controls for the sensitivity of the intensity;

[0046] Joint weight of geometric weight and intensity weight The definition formula is as follows:

[0047] ;

[0048] Preferably, the smooth fitting process steps of the geometric position and intensity value in step S6b2 are as follows:

[0049] The smoothing definition formula for the geometric position is as follows:

[0050] ;

[0051] The smoothing definition formula of the intensity value is as follows:

[0052] .

[0053] Compared with the prior art, the present invention proposes a point cloud smoothing fitting method based on local density parameter adjustment and strength optimization. The advantages of this method are:

[0054] 1. Sparse point clouds have strong adaptability

[0055] The present invention uses local density values ​​to dynamically adjust the search radius to ensure that there are not too many neighborhood points in dense areas, and that sufficient neighborhood points can be obtained to participate in fitting in sparse areas; through a density-driven adaptive mechanism, the problem that the traditional fixed radius method cannot adapt to non-uniformly distributed point clouds is solved, especially in long-distance sparse point cloud scenarios, the fitting accuracy is significantly improved; for points with insufficient number of neighborhood points, a mechanism for dynamically expanding the neighborhood is introduced to ensure that the number of neighborhood points of each point meets the minimum requirement; this mechanism shows extremely high robustness in sparse point clouds and ensures the continuity and stability of the fitting results; whether it is a high-density local area or a low-density remote area, the present invention can flexibly process point cloud data with different density distributions by adaptively adjusting the search radius and expanding the neighborhood, and is suitable for a variety of complex scenarios.

[0056] 2. Robustness to Highly Noisy Point Cloud Data

[0057] The present invention adopts a local density calculation method based on the average distance between neighborhood points, which effectively suppresses the influence of noise points on density assessment; compared with traditional methods that only rely on the number of neighborhood points, the present invention is more robust and can accurately identify sparse and dense areas in high-noise point clouds; the present invention combines intensity filtering with statistical filtering methods to eliminate isolated points and abnormal points, and further improves the stability of fitting results when dynamically expanding the neighborhood; whether the noise in the point cloud data comes from equipment errors or environmental interference, the present invention can effectively deal with it.

[0058] 3. Feature optimization of comprehensive geometric position and intensity value

[0059] In the process of smooth fitting, the present invention introduces joint weights to simultaneously incorporate the geometric position and intensity value of the point cloud into the weight calculation; the present invention makes full use of intensity information to optimize and retain the intensity characteristics of the point cloud while smoothing the geometric features; the geometric weight is based on the Euclidean distance between points to ensure that the geometric characteristics of the point cloud remain continuous during smooth fitting; the intensity weight is based on the intensity difference to enhance the ability to express the target surface features (such as hardness, material); through the joint optimization of geometric weights and intensity weights, the present invention achieves the synchronous optimization of geometric position and intensity value, and solves the problem that intensity information is easily ignored or destroyed in traditional methods.

[0060] 4. Boundary feature protection

[0061] In the case of insufficient neighborhood points, the present invention directly retains the original coordinates and intensity values ​​of the point cloud through a boundary protection mechanism; compared with the traditional method of over-smoothing the boundary points and causing feature distortion, the present invention can significantly improve the fitting accuracy of the boundary area, making the boundary features clearer and more reliable; since the boundary area of ​​the point cloud is usually sparsely distributed and has insufficient neighborhood, it is difficult for traditional methods to handle it effectively, while the boundary protection strategy of the present invention enables the sparse boundary points to be originally retained during the fitting process, avoiding feature loss and meeting the requirements of terrain modeling and target detection for boundary clarity.

[0062] 5. Improve processing efficiency and flexibility

[0063] The present invention avoids excessive calculation of neighborhood points in dense areas by dynamically adjusting the search radius, and reduces repeated calculations in sparse areas through a neighborhood expansion mechanism, thereby significantly improving the efficiency of the algorithm in large-scale point cloud data processing.

[0064] 6. Wide range of applications

[0065] By setting parameters such as the basic search radius, the minimum number of neighborhood points, and the maximum expansion radius, the present invention can flexibly adapt to point cloud data with different resolutions and different sparsity levels; in addition, the adjustment parameters of the intensity weight (such as the standard deviation) can optimize the results for specific scenarios (such as underwater target detection) to meet the needs of multiple scenarios.

[0066] 7. Advantages of 3D Sonar Data

[0067] The present invention is particularly suitable for processing sparse point clouds generated by forward-looking three-dimensional sonar by dynamically adjusting the search radius and the boundary protection mechanism; the present invention can effectively fit both remote targets and boundary areas while optimizing the expression capability of intensity information; while optimizing the geometric fitting, the present invention retains and enhances the intensity information, so that the fitting results have higher physical reliability and interpretability; the smooth fitting point cloud generated by the present invention can be directly used for underwater terrain modeling and target detection, while meeting the high requirements for geometric accuracy and intensity characteristics, and providing efficient technical support for the application of three-dimensional sonar in the field of ocean detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The figure is an overall flow chart of the method of the present invention.

[0069] Figure 2 The results of the fitting algorithm of the embodiment of the present invention are shown in FIG. (a) and (c) are visualization results of the point cloud of the same frame of underwater slope data at different angles, (b) is the result of (a) after surface fitting, and (d) is the result of (c) after surface fitting.

[0070] Figure 3The results of the fitting algorithm of the embodiment of the present invention are shown in FIG. (a) and (c) are visualization results of two different frames of underwater flat terrain data, (b) is the result of (a) after surface fitting, and (d) is the result of (c) after surface fitting. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present application to further clearly and completely describe the technical solutions in the embodiments of the present application. It should be noted that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0072] In order to make the invention objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application are further described in detail in conjunction with the drawings in the specification: In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the advantages of the present invention will be further illustrated by comparing the embodiments in conjunction with the drawings and specific implementation methods.

[0073] The present invention proposes a point cloud smoothing fitting method based on local density parameter adjustment and strength optimization. The overall flow chart of the method is as follows: Figure 1 As shown, the steps of this method are described in detail below:

[0074] Step S1, obtaining data: obtaining original point cloud data from a forward-looking 3D sonar device;

[0075] Step S2, point cloud denoising: using intensity filtering and statistical filtering algorithms to perform denoising on the original point cloud;

[0076] Furthermore, in order to improve the accuracy and robustness of surface fitting, it is often necessary to process the original point cloud data, which can be processed by combining intensity filtering and statistical filtering;

[0077] S2a, intensity filtering formula is defined as follows:

[0078] ;

[0079] in, is a point in the point cloud, Yes The strength value of is the intensity threshold;

[0080] Points with low intensity values ​​are usually noise points or invalid points, which are caused by sensor measurement errors or environmental interference. Through intensity filtering, the signal-to-noise ratio of point cloud data can be significantly improved, and points with strong target structure information can be retained, providing higher quality input data for subsequent surface fitting. Intensity filtering is particularly suitable for processing point cloud data with obvious signal reflection characteristics. By eliminating low-intensity points, points with reliable geometric characteristics are retained, reducing the interference of noise points in surface fitting, making the fitting results smoother and more accurate.

[0081] S2b, perform statistical filtering, by analyzing the local density distribution of points in the point cloud, remove points with abnormal density; statistical filtering requires first calculating each point To its neighboring points All points The Euclidean distance formula is as follows:

[0082] ;

[0083] in, is the current point, Represents the nearest neighbor point set of a point, which contains k neighboring points. Indicate point With point The Euclidean distance of

[0084] By calculating the distance, the midpoint of the point cloud can be measured The local distribution characteristics of , that is, its geometric relationship with neighboring points;

[0085] For each point , the formula for calculating the average distance between it and its neighbors is as follows:

[0086] ;

[0087] in Yes The average neighborhood distance of , k is the number of points contained in the neighborhood. Average neighborhood distance Characterizes the point The local density of

[0088] if The value is large, indicating that The surrounding points are sparse, which may be an outlier;

[0089] Statistical filtering constructs global statistical characteristics by analyzing the local average distance of all points in the point cloud, including the mean and standard deviation. The average value formula is as follows:

[0090] ;

[0091] in, represents the average neighborhood distance of all points in the point cloud, and n represents the total number of points in the point cloud;

[0092] The standard deviation formula is as follows:

[0093] ;

[0094] in, represents the standard deviation of the average neighborhood distance of all points;

[0095] Mean and standard deviation It reflects the local density distribution of all points in the point cloud and is used to determine the threshold of abnormal points;

[0096] Next, whether a point is an outlier is determined by the following conditional formula:

[0097] ;

[0098] in is a hyperparameter that is usually used to control the strictness of the detection;

[0099] The decision rule is point The average neighborhood distance Above average and standard deviation The weighted sum of For outliers.

[0100] Step S3, calculate local density: obtain the local density of the point cloud by calculating the average distance from the neighborhood points to the center point; the density value reflects the sparsity of the point cloud, avoiding misjudgment caused by relying solely on the number of neighborhood points;

[0101] Furthermore, the local density calculation process in step S3 is as follows:

[0102] For any point , the formula for defining its neighborhood with radius r is as follows:

[0103] ;

[0104] In the formula is the position vector of point i, is the position vector of point j, is the Euclidean distance, is the point set in the neighborhood with radius r and centered at point p;

[0105] Next, the local density Based on the average distance calculation from the neighborhood point to the center point, the calculation formula is:

[0106] ;

[0107] in, is the number of neighborhood points, is to prevent division by zero errors;

[0108] The larger the local density, the The smaller the distance to the neighboring points, the point is distributed in a dense area; the smaller the local density, the smaller the point The larger the distance of the neighboring points is, the more sparsely distributed the points are.

[0109] Step S4, dynamically adjust the search radius: dynamically adjust the search radius according to the local density to adapt the algorithm to sparse and non-uniformly distributed point clouds;

[0110] Furthermore, the process steps of dynamically adjusting the search radius in step S4 are as follows:

[0111] Search Radius The adjustment formula is as follows:

[0112] ;

[0113] in, is the base search radius, is the adjustment scale factor, , are the minimum and maximum values ​​of density respectively;

[0114] In dense areas (higher density), the search radius Reduce to avoid excessive smoothing due to too many neighborhood points; in sparse areas (low density), the search radius Increase to ensure that enough neighborhood points participate in smoothing.

[0115] Step S5, dynamically expand the neighborhood: when the number of neighborhood points is insufficient, dynamically expand the area based on the search radius to ensure the stability of the fitting process;

[0116] Furthermore, the steps of the dynamic neighborhood expansion process in step S5 are as follows:

[0117] The initial search radius is Radius, each expansion increment is , the expansion conditions are as follows:

[0118] ;

[0119] In the formula is the minimum number of neighborhood points threshold, is the maximum value of the search radius;

[0120] The expansion process is based on the dynamically adjusted search radius, on which the neighborhood is expanded. The neighborhood points are recalculated each time the search radius is expanded until the number of neighborhood points meets the requirement or the search radius reaches the upper limit.

[0121] Step S6, boundary judgment and obtaining point cloud result map:

[0122] Boundary judgment is based on the situation of neighborhood points. Different neighborhood points have different judgment modes. There are two specific situations: 1. Insufficient neighborhood points; 2. Sufficient neighborhood points.

[0123] 1. For boundary points with insufficient neighborhood points, boundary protection is performed to directly retain the original geometric position and intensity value to avoid blurring or loss of boundary features due to excessive smoothing, and obtain a point cloud result map;

[0124] Furthermore, the process steps of the boundary protection in step S6a are as follows:

[0125] The rules for border protection are as follows:

[0126] .

[0127] 2. For boundary points with enough neighborhood points, calculate the joint weight and update the smooth fit:

[0128] (1) Based on the calculation of geometric weight and intensity weight, a joint weight calculation method is performed to simultaneously optimize the geometric position and intensity value of the point cloud;

[0129] Furthermore, the calculation method of the joint weight in step S6b1 is as follows:

[0130] Geometry Weight The definition formula is as follows:

[0131] ;

[0132] Where exp is the exponential function;

[0133] Intensity Weight The definition formula is as follows:

[0134] ;

[0135] In the formula is the standard deviation of the intensity differences, which controls for the sensitivity of the intensity;

[0136] Joint weight of geometric weight and intensity weight The definition formula is as follows:

[0137] ;

[0138] (2) Smoothly fitting the geometric position and intensity value of the point cloud based on the joint weight to update the geometric position and intensity value of the point cloud;

[0139] Furthermore, the smooth fitting process steps of the geometric position and intensity value in step S6b2 are as follows:

[0140] The coordinate smoothing definition formula is as follows:

[0141] ;

[0142] The intensity smoothing definition formula is as follows:

[0143] .

[0144] (3) Generate the fitted reconstructed point cloud: Visualize the result of the smooth fitting to obtain the fitted result graph.

[0145] The present invention is compared with the traditional method, as shown in Table 1:

[0146] Table 1 Comparison between the improved method and the traditional method

[0147] characteristic Traditional least squares method Traditional MLS Improved MLS in the present invention Neighborhood selection Global Fixed search radius Dynamically adjust search radius Weight calculation none Geometry Weight Geometry weight + strength weight Adaptability to sparse point clouds Difference generally excellent Boundary point protection none none Keep original features Intensity information utilization none none Jointly optimize strength features Noise robustness Difference Strong powerful

[0148] Embodiment 1

[0149] Take a frame of underwater slope terrain point cloud data, after noise reduction, the unfitted point cloud data and the point cloud data calculated by the fitting method of the present invention are compared. Figure 2 shown. Figure 2 (a) and (c) are the visualization results of the point cloud of the same frame of underwater slope terrain data at different angles, where (b) is the result of (a) after surface fitting, and (d) is the result of (c) after surface fitting.

[0150] Embodiment 2

[0151] Take two frames of point cloud data of underwater flat terrain, and after noise reduction, compare the unfitted point cloud data with the point cloud data calculated by the fitting method of the present invention. Figure 3 shown. Figure 3 (a) and (c) are visualization results of two different frames of underwater flat terrain data, (b) is the result of (a) after surface fitting, and (d) is the result of (c) after surface fitting.

[0152] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0153] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization, characterized in that: include: Step S1, obtaining data: obtaining original point cloud data from a forward-looking 3D sonar device; Step S2, point cloud denoising: using intensity filtering and statistical filtering algorithms to perform denoising on the original point cloud; Step S3, calculating local density: calculating the local density of the point cloud based on the average distance from the neighborhood points to the center point; Step S4, dynamically adjust the search radius: dynamically adjust the search radius according to the local density to adapt the algorithm to sparse and non-uniformly distributed point clouds; Step S5, dynamically expand the neighborhood: when the number of neighborhood points is insufficient, dynamically expand the area based on the search radius to ensure the stability of the fitting process; Step S6, boundary judgment and obtaining point cloud result map: Boundary judgment is based on the situation of neighboring points. Different situations of neighboring points have different judgment modes. There are two specific situations: a. Insufficient neighboring points; b. Sufficient neighboring points; Step S6a, for the boundary points with insufficient neighborhood points, boundary protection is performed, the original geometric position and intensity value are directly retained, and the boundary feature blur or loss caused by excessive smoothing is avoided, and the point cloud result image is visualized; Step S6b, for the boundary points with enough neighborhood points, calculate the joint weight and update the smooth fit: Step S6b1, based on the calculation of the geometric weight and the intensity weight, a joint weight calculation method is performed to simultaneously optimize the geometric position and intensity value of the point cloud; Step S6b2, performing smooth fitting on the geometric position and intensity value of the point cloud based on the joint weight, so as to update the geometric position and intensity value of the point cloud; Step S6b3, visualizing the result of the smooth fitting to obtain a fitting result graph.

2. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: The process steps for calculating the local density in step S3 are as follows: For any point , the formula for defining its neighborhood with radius r is as follows: ; In the formula is the position vector of point i, is the position vector of point j, is the Euclidean distance, For point The point set in the neighborhood with a radius of r as the center; Next, the local density Based on the average distance calculation from the neighborhood point to the center point, the calculation formula is: ; in, is the number of neighborhood points, is to prevent division by zero errors.

3. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: The process steps of dynamically adjusting the search radius in step S4 are as follows: Search Radius The adjustment formula is as follows: ; in, is the base search radius, is the adjustment scale factor, is the minimum value of density, is the maximum value of density, is the local density.

4. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: The process steps of dynamically expanding the neighborhood in step S5 are as follows: The initial search radius is Radius, each expansion increment is , the extended condition formula is as follows: ； In the formula is the minimum number of neighborhood points threshold, is the maximum value of the search radius, It is a point is the center and the radius is The point set in the neighborhood of is the position vector of point i, is the search radius.

5. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: The formula for boundary protection in step S6a is as follows: ； in, is the number of neighborhood points, is the minimum number of neighborhood points threshold, is the position vector of point i, Yes The intensity value of .

6. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: The steps of the method for calculating the joint weight in step S6b1 are as follows: Geometry Weight The definition formula is as follows: ； Where exp is the exponential function, is the position vector of point i, is the position vector of point j, is the search radius; Intensity Weight The definition formula is as follows: ； In the formula is the standard deviation of the intensity differences, which controls the sensitivity of the intensity, Yes The strength value of Yes The strength value of Joint weight of geometric weight and intensity weight The definition formula is as follows: 。 7. The point cloud smoothing fitting method based on local density parameter adjustment and intensity optimization according to claim 1, characterized in that: In step S6b2, the smooth fitting process of the geometric position and the intensity value is as follows: The smoothing definition formula for the geometric position is as follows: ； in, is the position vector of point i, is the position vector of point j, Yes The strength value of The smoothing definition formula of the intensity value is as follows: ； in, is the position vector of point i, is the position vector of point j, Yes The intensity value of .

Images