A method and device for adaptively controlling Gaussian point cloud density

Abstract

The present invention discloses a method and device for adaptively controlling Gaussian point cloud density, which is applied to a handheld laser scanner. The method comprises: determining the motion characteristic parameters of the device based on the sensor data collected at the last moment; determining the current motion state classification result of the device based on the motion characteristic parameters; controlling the sensor to perform corresponding sampling operations on the target scene based on the current sampling strategy corresponding to the current motion state classification result to generate point cloud data; determining the complexity corresponding to each target point in the point cloud data and generating the point cloud complexity; and dynamically adjusting the reference point cloud density in the current sampling strategy through a PID controller based on the point cloud complexity and the current motion state classification result. Thus, the reference point cloud density in the current sampling strategy can be adaptively adjusted according to the current motion state classification result of the device, thereby balancing the rendering quality and computing load of the device and ensuring that the device can maintain smooth operation under different motion states.

Description

Technical Field

[0001] The present invention belongs to the field of image processing technology, and in particular relates to a method and device for adaptively controlling the density of Gaussian point clouds. Background Art

[0002] 3D laser scanning technology is widely used in industrial inspection, cultural heritage preservation, architectural surveying, virtual reality, and other fields. While traditional fixed 3D scanners offer high accuracy, they are limited by their fixed position and lack the flexibility to adapt to complex scenarios. With the advancement of miniaturization and mobile computing technologies, handheld 3D laser scanners have rapidly developed due to their flexibility and portability. However, they also face a series of technical challenges brought about by the movement of the equipment.

[0003] In recent years, 3D Gaussian Splatting (3DGS) technology has achieved significant breakthroughs in 3D scene representation and rendering. Through differentiable rendering and stochastic gradient descent optimization, it achieves high-quality real-time rendering. Compared to Neural Radiance Field (NeRF), this method boasts faster training and higher rendering efficiency, making real-time point cloud processing possible for handheld devices. However, high-density Gaussian point clouds can still impose a significant computational burden when the device is moving at high speeds.

[0004] The core of Gaussian point cloud rendering is to transform a 3D Gaussian ellipsoid into a 2D Gaussian distribution through projection, and then generate the final image through transparency blending. Each Gaussian point is defined by its center position, covariance matrix, and appearance attributes, and its rendering complexity is proportional to the number of points. During handheld scanning, how to dynamically adjust the point cloud density based on the motion state to balance rendering quality and computational load has become a key issue that needs to be addressed. Summary of the Invention

[0005] In response to the above-mentioned problems existing in the prior art, an embodiment of the present invention provides a method and device for adaptive control of Gaussian point cloud density; this method can perceive the motion state of a handheld laser scanner in real time, and dynamically adjust the current sampling strategy according to the perceived motion state, thereby balancing the rendering quality and computing load of the handheld laser scanner, so that the handheld laser scanner can maintain smooth operation in different motion states.

[0006] According to a first aspect of an embodiment of the present invention, there is provided an adaptive control method for Gaussian point cloud density, which is applied to a handheld laser scanner; the method comprises: determining motion characteristic parameters of the handheld laser scanner based on sensor data collected at a previous moment; classifying the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; based on a current sampling strategy corresponding to the current motion state classification result, controlling the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene to generate point cloud data; determining the complexity corresponding to each target point in the point cloud data and generating a point cloud complexity; based on the point cloud complexity and the current motion state classification result, dynamically adjusting the reference point cloud density in the current sampling strategy through a PID controller to obtain a sampling strategy corresponding to the next moment adjacent to the current moment.

[0007] Optionally, the motion characteristic parameters of the handheld laser scanner are determined based on the sensor data collected at the previous moment; including: based on the sensor data collected at the previous moment, using an unscented Kalman filter algorithm to determine the current fusion state of the handheld laser scanner; extracting the linear velocity norm, acceleration norm, and angular velocity norm of the handheld laser scanner from the current fusion state; based on the linear velocity norm and acceleration norm of the handheld laser scanner, using a fast Fourier transform algorithm to extract the main frequency component and power spectral density in a preset sliding window; based on the main frequency component and power spectral density in the preset sliding window, determining the component proportion of the spectral component with a frequency less than a preset threshold; using the component proportion, linear velocity norm, and angular velocity norm as the motion characteristic parameters of the handheld laser scanner.

[0008] Optionally, the motion characteristic parameters include at least: the component ratio of the spectral components, the linear velocity norm and the angular velocity norm; the current motion state of the handheld laser scanner is classified based on the motion characteristic parameters to obtain the current motion state classification result; including: based on the component ratio of the spectral components, the motion mode of the handheld laser scanner is detected; if the detection result indicates that the component ratio is less than a first preset component threshold, the current motion mode of the handheld laser scanner is determined to be a steady state mode; if the detection result indicates that the component ratio is not less than the first preset component threshold, the current motion mode of the handheld laser scanner is determined to be a non-steady state mode; based on the current motion mode of the handheld laser scanner, the current motion state of the handheld laser scanner is classified according to the linear velocity norm and the angular velocity norm to generate a current motion state classification result.

[0009] Optionally, based on the current motion mode of the handheld laser scanner, the current motion state of the handheld laser scanner is classified according to the linear velocity norm and the angular velocity norm to generate a current motion state classification result; including: when the current motion mode of the handheld laser scanner is a steady state mode, if the linear velocity norm is less than a first preset velocity threshold and the angular velocity norm is less than a first preset angular velocity threshold, then determining that the steady state mode corresponding to the handheld laser scanner is a stationary state; if the linear velocity norm is not less than the first preset velocity threshold and less than the second preset velocity threshold, and the angular velocity norm is less than the second preset angular velocity threshold, then determining that the handheld laser scanner is in a stationary state. The stable state mode corresponding to the scanner is a low-speed moving state; wherein, the second preset angular velocity threshold is greater than the first preset angular velocity threshold; when the current motion mode of the handheld laser scanner is a non-steady state mode, if the linear velocity norm is not less than the second preset speed threshold and less than the third preset speed threshold, or the angular velocity norm is not less than the second preset angular velocity threshold and less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a medium-speed moving state; if the linear velocity norm is not less than the fourth preset speed threshold or the angular velocity norm is not less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a high-speed moving state.

[0010] Optionally, the current sampling strategy corresponding to the current motion state classification result controls the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene to generate point cloud data; including: when the current motion state classification result is a stationary state, determining that the sampling density in the current sampling strategy is the maximum preset sampling density of the system, and controlling the sensor to perform sampling operations on all areas of the target scene according to the maximum preset sampling density to generate point cloud data; wherein the maximum preset sampling density is a preset multiple of the reference point cloud density; when the current motion state classification result is a low-speed moving state, determining that the sampling density in the current sampling strategy is a maximum preset sampling density greater than a first preset ratio and less than the maximum preset sampling density; and controlling the sensor to perform sampling operations on non-critical areas of the target scene according to the determined sampling density, and performing sampling operations on the target scene according to the maximum preset sampling density. The sampling operation is performed on the key area of ​​the target scene to generate point cloud data; when the current motion state classification result is a medium-speed moving state, the sampling density in the current sampling strategy is determined to be greater than the maximum preset sampling density of the second preset ratio and less than the maximum preset sampling density; and the sensor is controlled to perform sampling operations on the non-key areas of the target scene according to the determined sampling density, and the sampling operation is performed on the key areas of the target scene according to the maximum preset sampling density to generate point cloud data; when the current motion state classification result is a high-speed moving state, the sampling density in the current sampling strategy is determined to be greater than the maximum preset sampling density of the third preset ratio and not greater than the maximum preset sampling density of the second preset ratio; and the sensor is controlled to perform sampling operations on all areas of the target scene according to the determined sampling density to generate point cloud data; wherein, the first preset ratio>the second preset ratio>the third preset ratio.

[0011] Optionally, the complexity corresponding to each target point in the point cloud data is determined, and the point cloud complexity is generated; including: for any target point in the point cloud data: determining the adaptive field radius of the target point; performing eigenvalue extraction processing on the field corresponding to the adaptive field radius to obtain eigenvalues; calculating the eigenvalue descriptor corresponding to the eigenvalue; determining the curvature contribution of the field based on the eigenvalue descriptor; determining the edge contribution of the field based on the edge score of the field; weighting the curvature contribution and the edge contribution to generate the complexity corresponding to the target point; generating the point cloud complexity based on the complexity corresponding to each target point in the point cloud data.

[0012] Optionally, the reference point cloud density in the current sampling strategy is dynamically adjusted through a PID controller based on the point cloud complexity and the current motion state classification result; including: taking the frame rate deviation determined by the point cloud complexity and the quality deviation determined by the edge score of the point cloud data as key indicators monitored by the PID controller; adaptively adjusting the PID parameters of the PID controller based on the current motion state classification result; determining the output result of the PID controller based on the key indicators and PID parameters monitored by the PID controller; if the output result is greater than a preset output threshold, it is determined that the system resources of the handheld laser scanner are insufficient, and a reduction operation is performed on the reference point cloud density of the current sampling strategy; if the output result is less than the negative value corresponding to the preset output threshold, it is determined that the system resources of the handheld laser scanner are sufficient, and an increase operation is performed on the reference point cloud density of the current sampling strategy; if the absolute value of the output result is not greater than the preset output threshold, no adjustment operation is performed on the reference point cloud density of the current sampling strategy.

[0013] Optionally, the method further includes: detecting the reference point cloud density corresponding to the current sampling strategy in real time; if the detection result indicates that the absolute value of the rate of change of the reference point cloud density is greater than a preset threshold, the reference point density will be readjusted until the rate of change of the reference point cloud density falls within the preset change threshold range.

[0014] Optionally, the PID parameters of the PID controller include at least: a proportional coefficient, an integral coefficient, and a differential coefficient; the adaptive adjustment of the PID parameters of the PID controller based on the current motion state classification result includes: when the current motion state classification result is a stationary state, the integral coefficient is adjusted in an increasing direction, and the proportional coefficient is adjusted in a decreasing direction; when the current motion state classification result is a low-speed moving state, no adjustment operation is performed on the proportional coefficient, the integral coefficient, and the differential coefficient; when the current motion state classification result is a medium-speed moving state, the proportional coefficient and the differential coefficient are both adjusted in their respective increasing directions; when the current motion state classification result is a high-speed moving state, the proportional coefficient and the differential coefficient are adjusted in their respective increasing directions, and the integral coefficient is adjusted in a decreasing direction.

[0015] According to the second aspect of an embodiment of the present invention, there is also provided an adaptive control device for Gaussian point cloud density, which is applied to a handheld laser scanner; the device includes: a first determination module, which is used to determine the motion characteristic parameters of the handheld laser scanner based on the sensor data collected at the previous moment; a motion state classification module, which is used to classify the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; a sampling module, which is used to control the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene based on the current sampling strategy corresponding to the current motion state classification result to generate point cloud data; a second determination module, which is used to determine the complexity corresponding to each target point in the point cloud data and generate a point cloud complexity; an adjustment module, which is used to dynamically adjust the reference point cloud density in the current sampling strategy through a PID controller based on the point cloud complexity and the current motion state classification result.

[0016] According to a third aspect of an embodiment of the present invention, an electronic device is further provided, comprising: a processor; a memory for storing instructions executable by the processor; and the processor for reading the executable instructions from the memory and executing the instructions to implement the method described in the first aspect.

[0017] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is further provided, on which a computer program is stored. When the program is executed by a processor, the method described in the first aspect is implemented.

[0018] An embodiment of the present invention provides a method and apparatus for adaptively controlling Gaussian point cloud density, which is applied to a handheld laser scanner. The method comprises: first, determining motion characteristic parameters of the handheld laser scanner based on sensor data collected at a previous moment; then, classifying the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; second, controlling the sensor to begin performing a sampling operation corresponding to the current sampling strategy on a target scene based on a current sampling strategy corresponding to the current motion state classification result to generate point cloud data; then, determining the complexity corresponding to each target point in the point cloud data and generating a point cloud complexity; and finally, dynamically adjusting the reference point cloud density in the current sampling strategy based on the point cloud complexity and the current motion state classification result using a PID controller to obtain a sampling strategy corresponding to the next moment adjacent to the current moment. Thus, the current motion state of the device is monitored in real time by multiple sensors, and a real-time feedback control algorithm is constructed using a PID controller to adaptively adjust the reference point cloud density in the current sampling strategy based on the current motion state of the device, thereby balancing the rendering quality and computational load of the device and ensuring smooth operation of the device under different motion states. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Hereinafter, some specific embodiments of the present invention will be described in detail in an exemplary and non-limiting manner with reference to the accompanying drawings. The same reference numerals in the accompanying drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the accompanying drawings:

[0020] Figure 1 A schematic flow chart of a method for adaptively controlling Gaussian point cloud density provided by one embodiment of the present invention;

[0021] Figure 2 Schematic diagram of a process for determining the complexity corresponding to each target point in point cloud data in one embodiment of the present invention;

[0022] Figure 3 A schematic structural diagram of a device for adaptively controlling Gaussian point cloud density provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purposes, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described 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 those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0024] Existing handheld laser scanners have the following disadvantages:

[0025] 1. Computing resource bottleneck: Traditional handheld laser scanners use a fixed-density Gaussian point cloud representation method, maintaining the same point cloud density regardless of the scanner's motion during the scanning process. This results in the system needing to process a large amount of redundant point cloud data when the handheld laser scanner is moving at high speeds, resulting in excessive computing resource utilization, slow device response, reduced frame rates, and even system freezes in severe cases.

[0026] 2. Unbalanced spatial sampling: Existing technologies typically use a uniformly distributed point cloud density, making it difficult to perform adaptive sampling based on differences in scene geometric complexity. This results in oversampling in simple, flat areas, wasting computing resources, and undersampling in complex geometric areas (e.g., insufficient detail).

[0027] 3. Lack of motion adaptability: When using a handheld scanner to scan the target scene, the device's motion state (for example, stillness, constant speed, acceleration, or rotation) will continue to change during the handheld scanning process. Existing technologies lack real-time perception and response mechanisms to the motion state, and therefore cannot dynamically adjust the sampling strategy and computing resource allocation based on the motion state of the handheld scanner.

[0028] like Figure 1 , which is a flow chart of a method for adaptively controlling Gaussian point cloud density provided by one embodiment of the present invention.

[0029] A method for adaptively controlling the density of a Gaussian point cloud; applied to a handheld laser scanner; the method comprising at least the following steps:

[0030] S101, determining motion characteristic parameters of a handheld laser scanner based on sensor data collected at a previous moment;

[0031] S102, classifying the current motion state of the handheld laser scanner based on the motion feature parameters to obtain a current motion state classification result;

[0032] S103, based on the current sampling strategy corresponding to the current motion state classification result, controlling the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene to generate point cloud data;

[0033] S104, determining the complexity corresponding to each target point in the point cloud data, and generating the point cloud complexity;

[0034] S105 , based on the point cloud complexity and the current motion state classification result, dynamically adjust the reference point cloud density in the current sampling strategy through a PID controller.

[0035] The handheld laser scanner carries several sensors, each of which is used to collect sensor data of the target scene.

[0036] In S101 , motion characteristic parameters of the handheld laser scanner are determined based on the sensor data collected at the last moment according to a preset rule or model.

[0037] Exemplarily, the sensor data collected at the previous moment is used to determine the current fusion state of the handheld laser scanner using an unscented Kalman filter algorithm; the linear velocity norm, acceleration norm, and angular velocity norm of the handheld laser scanner are extracted from the current fusion state; based on the linear velocity norm and acceleration norm of the handheld laser scanner, a fast Fourier transform algorithm is used to extract the main frequency components and power spectrum density in a preset sliding window; based on the main frequency components and power spectrum density in the preset sliding window, the component proportion of the spectral components with a frequency less than a preset threshold is determined; the component proportion, linear velocity norm, and angular velocity norm are used as motion characteristic parameters of the handheld laser scanner.

[0038] Further exemplarily, the sensor data collected at the previous moment is used to determine the current fusion state of the handheld laser scanner using an unscented Kalman filter algorithm; including: constructing a state vector, a nonlinear state transition model, and a nonlinear measurement model of the handheld laser scanner; obtaining the previous state vector corresponding to the handheld laser scanner at the previous moment; using the nonlinear state transition model to perform time updates on the previous state vector to generate a current state vector corresponding to the current moment adjacent to the previous moment; when the handheld laser scanner receives new measurement data from the ToF sensor; based on the new measurement data, the current state vector is updated using the nonlinear measurement model to generate an updated state; here, the updated state is determined as the current fusion state of the handheld laser scanner.

[0039] For example: (1) Create the state vector X of a handheld laser scanner

[0040] The state vector X includes the complete information of the device motion and the sensor error term, as shown in the following equation (1).

[0041] Formula (1);

[0042] in: is a three-dimensional space position vector, indicating the position of the device in the world coordinate system;

[0043] is the velocity vector, which indicates the movement rate of the device in three directions;

[0044] is the posture represented by the quaternion, satisfying the normalization constraint of ||q||=1;

[0045] is the accelerometer bias vector, which is used to correct the raw acceleration measurement;

[0046] is the gyroscope bias vector, which is used to correct the raw angular velocity measurement;

[0047] It is the ToF measurement scale factor, which is used to correct the proportional error of depth measurement.

[0048] (2) Time update of state vector X

[0049] Aiming at the motion characteristics of the handheld laser scanner, a nonlinear state transfer model including rotational dynamics was constructed. The nonlinear state transfer model is used to accurately describe the temporal evolution relationship of different variables in the state vector, as shown in the following equation (2).

[0050] Formula (2);

[0051] in: is the 3×3 rotation matrix corresponding to the quaternion q, which converts the acceleration from the sensor coordinate system to the world coordinate system; g is the gravitational acceleration vector [0, 0, 9.81]ᵀ, in m / s²;

[0052] Represents quaternion multiplication operation to ensure the correctness of state vector update;

[0053] represents the exponential mapping function, which converts the angular velocity vector into a quaternion representing the rotation;

[0054] are the process noises of each state variable, modeled as Gaussian white noise with zero mean;

[0055] a(k-1) represents the accelerometer measurement at time k-1; W(k-1) represents the gyroscope measurement at time k-1. The current state vector is shown below.

[0056]

[0057] 3) Measurement update of state vector X

[0058] Construct a nonlinear measurement model that associates the ToF distance z(k) with the state vector X(k), as shown in the following equation (3);

[0059] Formula (3);

[0060] in, The measurement noise is modeled as Gaussian white noise with zero mean.

[0061] Whenever the system receives a new measurement from the ToF sensor, it performs a measurement update step to modify the state vector and covariance matrix.

[0062] Therefore, the handheld laser scanner realizes continuous tracking of the device's motion state by alternating between time updates and measurement updates, which is conducive to the real-time adjustment of the point cloud sampling density according to the device's motion state in the later stage.

[0063] This system uses the unscented Kalman filter algorithm to detect the motion state of the handheld laser scanner, which solves the limitations of the Kalman filter in dealing with nonlinear systems.

[0064] (4) Extract motion feature parameters

[0065] The linear velocity norm, acceleration norm, and angular velocity norm of the handheld laser scanner are extracted from the current fusion state.

[0066] The linear velocity norm reflects the overall movement rate of the device, and its calculation formula is:

[0067] Formula (4);

[0068] in, 、 、 are the velocity components in the directions of the three coordinate axes, in m / s.

[0069] The acceleration norm reflects the rate of change of device speed, and the calculation formula is:

[0070] Formula (5);

[0071] in, The acceleration components in the three coordinate axis directions are in m / s².

[0072] The angular velocity norm reflects the speed of the device's rotation. The calculation formula is the same as formula (5), the only difference is that acceleration is replaced by angular velocity, which is not presented here.

[0073] The system uses Fast Fourier Transform (FFT) to analyze the speed data within a 500ms time window and extract the main frequency components. And the power spectrum density P(f) is used to determine the current motion mode of the handheld laser scanner; the specific formula is shown below:

[0074]

[0075] Formula (6);

[0076] Where: P(f<2Hz) represents the sum of the powers of all spectral components with frequencies below 2Hz; Represents the sum of the powers of all spectral components; Indicates the proportion of spectral components with frequencies less than 2 Hz.

[0077] The component ratio is used to qualitatively analyze the motion of the handheld laser scanner and obtain the device motion mode. After determining the device motion mode, the motion intensity of the handheld laser scanner is quantitatively analyzed based on the linear velocity norm and angular velocity norm to obtain the device motion state classification result.

[0078] In S102 and S103 , the current motion state of the handheld laser scanner is classified based on the motion feature parameters using a preset rule or model algorithm to obtain a current motion state classification result.

[0079] Exemplarily, the motion characteristic parameters include at least: the component ratio of the spectral components, the linear velocity norm and the angular velocity norm; the current motion state of the handheld laser scanner is classified based on the motion characteristic parameters to obtain the current motion state classification result; including: based on the component ratio of the spectral components, the motion mode of the handheld laser scanner is detected; if the detection result indicates that the component ratio is less than a first preset component threshold, the current motion mode of the handheld laser scanner is determined to be a steady state mode; if the detection result indicates that the component ratio is not less than the first preset component threshold, the current motion mode of the handheld laser scanner is determined to be a non-steady state mode; based on the current motion mode of the handheld laser scanner, the current motion state of the handheld laser scanner is classified according to the linear velocity norm and the angular velocity norm to generate a current motion state classification result.

[0080] For example: If If the absolute value of the difference between the value and 1 is less than 0.01, the current motion mode of the handheld laser scanner is determined to be the steady state mode. In the current motion mode, the device's motion energy is mainly concentrated in the low-frequency component (<2Hz). The steady state mode is used to indicate that the handheld laser scanner moves relatively smoothly and slowly.

[0081] like If the absolute value of the difference between 0 and 1 is not less than 0.01, it is determined that the current motion mode of the handheld laser scanner is a non-stationary state mode. In the current motion mode, the device motion energy is mainly concentrated in the high-frequency components. The non-stationary state mode is used to indicate that the motion of the handheld laser scanner contains rapid changes, jitter or vibration.

[0082] Further illustratively, classifying the current motion state of the handheld laser scanner according to the linear velocity norm and the angular velocity norm to generate a current motion state classification result includes:

[0083] S1. When the current motion mode of the handheld laser scanner is a steady state mode, if the linear velocity norm is less than a first preset velocity threshold and the angular velocity norm is less than a first preset angular velocity threshold, it is determined that the steady state mode corresponding to the handheld laser scanner is a stationary state; if the linear velocity norm is not less than the first preset velocity threshold and less than the second preset velocity threshold, and the angular velocity norm is less than the second preset angular velocity threshold, it is determined that the steady state mode corresponding to the handheld laser scanner is a low-speed moving state; wherein, the second preset angular velocity threshold is greater than the first preset angular velocity threshold.

[0084] S2. When the current motion mode of the handheld laser scanner is a non-steady state mode, if the linear velocity norm is not less than the second preset speed threshold and less than the third preset speed threshold, or the angular velocity norm is not less than the second preset angular velocity threshold and less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a medium-speed moving state; if the linear velocity norm is not less than the fourth preset speed threshold or the angular velocity norm is not less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a high-speed moving state.

[0085] For example, based on the linear velocity norm ||v|| and the angular velocity norm ||ω||, the exercise intensity is divided into four levels:

[0086] Static state (S0): ;

[0087] The static state corresponds to the situation where the device maintains a relatively fixed position. =0.05m / s, =0.1rad / s. In this state, the system prioritizes rendering quality, maximizing point cloud density and geometric detail. The sampling density is increased to the system's highest level, resulting in the maximum preset sampling density.

[0088] Low-speed movement state (S1): ;

[0089] The low-speed movement state corresponds to the situation where the device moves slowly and steadily. =0.3m / s, =0.5rad / s. Sampling is performed on the key areas of the target scene at the maximum preset sampling density, and sampling is performed on the non-key areas of the target scene at 80% of the maximum preset sampling density.

[0090] Medium-speed movement state (S2):

[0091] ;

[0092] The medium-speed movement state corresponds to the situation where the device moves at a medium speed or rotates at a medium rate. =0.8m / s, =1.2rad / s. Sampling is performed on the key areas of the target scene at the maximum preset sampling density, and sampling is performed on the non-key areas of the target scene at 60% of the maximum preset sampling density.

[0093] High-speed movement state (S3): ;

[0094] High-speed motion corresponds to the situation where the device moves or rotates rapidly. In this case, the system performs sampling operations on all areas of the target scene at a maximum preset sampling density of 30-60%.

[0095] In S104, for any target point in the point cloud data: determine the complexity corresponding to each target point based on a preset rule or model algorithm to generate the target point complexity; and generate the point cloud complexity based on the complexity of each target point in the point cloud data.

[0096] In S105, exemplarily, the frame rate deviation determined by the point cloud complexity and the quality deviation determined by the edge score of the point cloud data are used as key indicators monitored by the PID controller; the PID parameters of the PID controller are adaptively adjusted based on the current motion state classification result; the output result of the PID controller is determined based on the key indicators and PID parameters monitored by the PID controller; if the output result is greater than a preset output threshold, it is determined that the system resources of the handheld laser scanner are insufficient, and a reduction operation is performed on the benchmark point cloud density of the current sampling strategy; if the output result is less than the negative value corresponding to the preset output threshold, it is determined that the system resources of the handheld laser scanner are sufficient, and an increase operation is performed on the benchmark point cloud density of the current sampling strategy; if the absolute value of the output result is not greater than the preset output threshold, no adjustment operation is performed on the benchmark point cloud density of the current sampling strategy.

[0097] Specifically, (1) the two key indicators monitored by the PID controller are frame rate deviation and quality deviation.

[0098] The current frame rate is determined based on the point cloud complexity; the edge score of the point cloud data is determined, and the current quality index of the point cloud data is determined based on the edge score. The frame rate deviation of the point cloud data is then determined based on the difference between the preset target frame rate and the current frame rate; and the quality deviation of the point cloud data is determined based on the difference between the current quality index and the preset minimum quality threshold.

[0099] Frame rate deviation The formula is as follows:

[0100] Formula (7);

[0101] in, is the preset target frame rate, F(t) is the current frame rate;

[0102] Quality deviation The formula is as follows:

[0103] Formula (8);

[0104] Among them, Q(t) is the current quality index, which is used to evaluate the overall rendering quality and measure the degree of preservation of key geometric features. It is a preset minimum quality threshold used to ensure that the system always meets basic visual quality requirements.

[0105] (2) Adaptively adjusting the PID parameters of the PID controller based on the current motion state classification result. Exemplarily, the PID parameters of the PID controller include at least a proportional coefficient, an integral coefficient, and a differential coefficient; the adaptively adjusting the PID parameters of the PID controller based on the current motion state classification result includes: when the current motion state classification result is a stationary state, adjusting the integral coefficient in an increasing direction and adjusting the proportional coefficient in a decreasing direction; when the current motion state classification result is a low-speed moving state, not adjusting the proportional coefficient, the integral coefficient, and the differential coefficient; when the current motion state classification result is a medium-speed moving state, adjusting the proportional coefficient and the differential coefficient in their respective increasing directions; when the current motion state classification result is a high-speed moving state, adjusting the proportional coefficient and the differential coefficient in their respective increasing directions and adjusting the integral coefficient in a decreasing direction.

[0106] It should be noted that the proportional coefficient, differential coefficient and integral coefficient all have preset threshold ranges. For example: The calculation formula is as follows:

[0107] Formula (9);

[0108] in, is the base scale factor; The adjustment function is based on the current motion state classification result of the handheld laser scanner. When the current motion state classification result is a high-speed moving state, the adjustment function controls the proportional coefficient to be adjusted in an increasing direction to increase the response speed of the PID controller. When the current motion state classification result is a stationary state, the adjustment function controls the proportional coefficient to be adjusted in a decreasing direction to decrease the response speed of the PID controller.

[0109] The calculation formula of the integral coefficient is shown as follows:

[0110] Formula (10);

[0111] in, is the base integral coefficient; The adjustment function is based on the current motion state classification result of the handheld laser scanner. When the current motion state classification result is a high-speed moving state, the adjustment function controls the integral coefficient to be adjusted in a decreasing direction to prevent over-adjustment of the PID controller. When the current motion state classification result is a stationary state, the adjustment function controls the proportional coefficient to be adjusted in an increasing direction to reduce the control accuracy of the PID controller.

[0112] The calculation formula of the differential coefficient is shown below.

[0113] Formula (11);

[0114] in, is the base differential coefficient, The adjustment function is based on the current motion state classification result of the handheld laser scanner. When the current motion state classification result is a high-speed moving state, the adjustment function controls the differential coefficient to perform an adjustment operation in an increasing direction to improve the PID controller's ability to predict motion changes.

[0115] The PID parameters are adjusted according to the classification results of different motion states of the handheld laser scanner, as follows: Static state: Integral-dominated, focusing on precise control and stability;

[0116] Low speed state: three-way balance, balanced control effect;

[0117] Medium speed state: proportional and differential are enhanced to improve response speed;

[0118] High-speed state: Proportional and differential are significantly enhanced, while integral is weakened, giving priority to ensuring fast system response.

[0119] Output of PID controller As shown in the following formula:

[0120] Formula (12);

[0121] In the output results In the expression, the proportional term is used to directly respond to the deviation between the current frame rate and the target frame rate; the integral term is used to eliminate the system steady-state error and ensure long-term stable operation; the differential term is used to predict the system change trend and adjust the parameters in advance.

[0122] Standard PID control: The first three items constitute the traditional PID controller, which mainly optimizes the system frame rate;

[0123] Quality penalty term: μ·max(0, -e Q (t)) ensuring that minimum quality requirements are not compromised;

[0124] The quality penalty factor μ=2.0 generates strong negative feedback when the quality index is lower than the threshold.

[0125] If the output is greater than u min , it is determined that the system resources of the handheld laser scanner are insufficient, and a reduction operation is performed on the reference point cloud density of the current sampling strategy;

[0126] If the output result is less than 0 and less than u min If the corresponding negative value is obtained, it is determined that the handheld laser scanner has sufficient system resources and the baseline point cloud density of the current sampling strategy is increased;

[0127] If the absolute value of the output result is not less than 0 and not greater than u min , then no adjustment operation is performed on the benchmark point cloud density of the current sampling strategy to reduce frequent adjustments caused by small fluctuations.

[0128] Here, the reference point cloud density indicates the initial reference value for adaptively adjusting the point cloud density. This embodiment dynamically adjusts the adjustment coefficient of the current sampling strategy relative to the reference point cloud density using a PID controller. A sampling strategy for the next moment adjacent to the current moment is generated to guide the sensor in performing sampling operations corresponding to the next sampling strategy for the target scene at the next moment. The time interval between the current moment and the next moment is typically determined based on the actual scenario.

[0129] This embodiment uses a Gaussian point cloud density adaptive control method. When the device is moving rapidly, the system automatically reduces the overall point cloud density to maintain a smooth rendering frame rate. When the device is stationary or moving at low speeds, the system increases the point cloud sampling density to maximize detail restoration, thereby improving the motion adaptability of the handheld laser scanner. The PID feedback control of this embodiment ensures that the system automatically achieves the optimal performance balance across different hardware platforms and target scenarios.

[0130] The core innovation of this embodiment's adaptive Gaussian point cloud density control method lies in upgrading a single fixed density model to a dynamic, multi-dimensional intelligent control system. Using IMU and ToF sensors, the system accurately perceives the device's current motion state. Combining the feedback control of a PID controller with the mathematical tunability of the Gaussian distribution, it establishes a complete chain of "environmental perception - intelligent decision-making - real-time optimization." This systematic solution not only addresses the three major pain points of existing technologies (e.g., computing resource bottlenecks, uneven spatial sampling, and lack of motion adaptive capabilities), but also achieves a technological leap from "passive data processing" to "active optimization of resource allocation," significantly improving the system performance ceiling and application universality of handheld laser scanners.

[0131] In a preferred implementation of this embodiment, the method further includes: detecting the reference point cloud density corresponding to the sampling strategy in real time; if the detection result indicates that the absolute value of the rate of change of the reference point cloud density is greater than a preset threshold, readjusting the reference point density until the rate of change of the reference point cloud density falls within the preset change threshold.

[0132] Specifically, the difference between the adjusted reference point cloud density and the pre-adjustment reference point cloud density is calculated. This difference is then compared with the pre-adjustment reference point cloud density to determine the rate of change of the reference point cloud density. If the absolute value of the rate of change of the reference point cloud density exceeds 0.2, the reference point density is readjusted until the rate of change falls within the range of -0.2 to 0.2. This prevents oscillation in the PID controller's output, ensuring smooth system operation under various motion conditions.

[0133] In another preferred embodiment of this embodiment, the method further includes analyzing the output results corresponding to each of a plurality of consecutive moments; if the analysis results indicate that the plurality of output results conform to periodic variations, determining that the PID controller exhibits periodic oscillations, and adjusting the PID parameters of the PID controller. This prevents the output results of the PID controller from exhibiting periodic oscillations, thereby ensuring that the system maintains smooth operation under various motion conditions.

[0134] like Figure 2 FIG. 1 is a flow chart of determining the complexity corresponding to each target point in point cloud data according to an embodiment of the present invention.

[0135] Determine the complexity corresponding to each target point in the point cloud data and generate the point cloud complexity; including at least the following steps:

[0136] S201, for any target point in the point cloud data: determining the adaptive domain radius of the target point; performing eigenvalue extraction processing on the domain corresponding to the adaptive domain radius to obtain an eigenvalue; calculating an eigenvalue descriptor corresponding to the eigenvalue; determining the curvature contribution of the domain based on the eigenvalue descriptor; determining the edge contribution of the domain based on the edge score of the domain; weighting the curvature contribution and the edge contribution to generate the complexity corresponding to the target point;

[0137] S202: Generate point cloud complexity based on the complexity corresponding to each target point in the point cloud data.

[0138] Specifically, the average point spacing near the target point is determined; the adaptive domain radius of the target point is determined based on the average point spacing; the domain corresponding to the target point is determined according to the adaptive domain radius; a covariance matrix is ​​constructed based on the domain centroid corresponding to the domain and the point cloud within the domain; the covariance matrix is ​​decomposed by the Jacobi method to obtain several eigenvalues; eigenvalue descriptors corresponding to the several eigenvalues ​​are calculated; the curvature contribution of the domain is determined based on the eigenvalue descriptors; the edge contribution of the domain is determined based on the edge score of the domain; the curvature contribution and the edge contribution are weighted according to preset rules to generate the complexity corresponding to the target point.

[0139] For example: Average point spacing The calculation formula is as follows:

[0140] Formula (13);

[0141] in are the coordinates of the K nearest neighbor points of the target point P, and P is the coordinate of the target point.

[0142] Based on the local average point spacing, the system calculates the adaptive neighborhood radius :

[0143] Formula (14);

[0144] Where β is a scaling factor that can be dynamically adjusted based on the overall density of the point cloud:

[0145] Formula (15);

[0146] is the base scaling factor, is the adjustment coefficient, is the average density of the current point cloud, is the reference density value.

[0147] The centroid of the domain is determined based on the coordinates of each point in the domain. The covariance matrix is ​​constructed based on the centroid of the domain, the coordinates of each point in the domain, and the total number of points in the domain. The specific calculation formula is shown below:

[0148] Formula (16);

[0149] Among them, q is the coordinate of the point in the field, is the centroid of the area, is the set of points in the domain, C( p ) is a 3×3 real symmetric matrix containing the spatial distribution information of the neighborhood points. The Jacobi method is used to calculate C( p ) performs eigenvalue decomposition:

[0150]

[0151] Eigenvalues ​​are sorted in descending order , the corresponding eigenvector is .

[0152] Eigenvalues ​​and eigenvectors have clear geometric meanings: Approximate the local surface normal vector; The basis that forms the local tangent plane.

[0153] Based on the eigenvalue decomposition results, the system calculates a series of feature descriptors that describe local geometric characteristics:

[0154] Curvature index calculation:

[0155] Formula (17);

[0156] K( p ) range is [0,1 / 3], the larger the K(p) value is, the higher the local curvature is. p) Close to 0, uniform spherical area K( p ) is close to 1 / 3.

[0157] Complexity Focus on the local geometric features of a single point and its immediate surroundings, mainly based on curvature and edge strength evaluation:

[0158] The curvature contribution term is defined as:

[0159] Formula (18);

[0160] Where K( p ) is the curvature index calculated above, is the average curvature value of the point cloud, is the standard deviation of curvature, and the sigmoid function maps the curvature to the [0,1] interval and enhances the response to outliers. Contribution weight to curvature.

[0161] Edge score E( p ), the score measures whether the target point p is located in the geometric edge area by the change of the normal vector in the neighborhood of the target point p.

[0162] Formula (19);

[0163] Among them, n i and n p are the normal vectors of point p and its neighboring points, and N is the number of neighboring points. By calculating the difference between the normal vectors, we can determine whether the point is in the edge area. If E( p ) is larger, indicating that the point is in the edge area and the change is more drastic; if it is smaller, it means that the point is in the flat area.

[0164] The marginal contribution term is defined as:

[0165] Formula (20);

[0166] Where E(p) is the edge score calculated above, and are the mean and standard deviation of edge scores, Contribution weight to the edge.

[0167] Microcomplexity Integrate the above two terms by weighted summation:

[0168] Formula (21);

[0169] Complexity Reflects the complexity of the local geometric structure in the point cloud. Higher values ​​indicate richer geometric details at the target point, which requires a higher sampling density to accurately represent it.

[0170] Traverse all target points in the point cloud data and independently calculate the adaptive neighborhood and complexity of each target point; determine the point cloud complexity based on the complexity corresponding to each target point.

[0171] Therefore, the adaptive domain construction strategy is adopted to improve the accuracy of feature extraction at the target point, solve the problem of feature extraction instability caused by the fixed radius neighborhood construction strategy that cannot adapt to the spatial changes in point cloud density, improve the accuracy of feature extraction at the target point, and thus improve the accuracy of target point complexity calculation, which is conducive to adaptive sampling according to the complexity differences of the target scene.

[0172] like Figure 3 , which is a schematic structural diagram of a device for adaptively controlling Gaussian point cloud density provided by one embodiment of the present invention.

[0173] A device for adaptively controlling Gaussian point cloud density, characterized in that it is applied to a handheld laser scanner; the device 300 includes: a first determination module 301, used to determine the motion characteristic parameters of the handheld laser scanner based on the sensor data collected at the previous moment; a motion state classification module 302, used to classify the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; a sampling module 303, used to control the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene based on the current sampling strategy corresponding to the current motion state classification result, to generate point cloud data; a second determination module 304, used to determine the complexity corresponding to each target point in the point cloud data, and generate a point cloud complexity; an adjustment module 305, used to dynamically adjust the reference point cloud density in the current sampling strategy through a PID controller based on the point cloud complexity and the current motion state classification result.

[0174] In a preferred implementation manner of this embodiment, the first determination module includes: a prediction unit, used to determine the current fusion state of the handheld laser scanner using an unscented Kalman filter algorithm based on the sensor data collected at the previous moment; an extraction unit, used to extract the linear velocity norm, acceleration norm, and angular velocity norm of the handheld laser scanner from the current fusion state; an extraction unit, used to extract the main frequency component and power spectral density in a preset sliding window based on the linear velocity norm and acceleration norm of the handheld laser scanner using a fast Fourier transform algorithm; a first determination unit, used to determine the component proportion of spectral components with a frequency less than a preset threshold based on the main frequency component and power spectral density in the preset sliding window; a second determination unit, used to use the component proportion, linear velocity norm, and angular velocity norm as motion characteristic parameters of the handheld laser scanner.

[0175] In a preferred implementation manner of this embodiment, the motion characteristic parameters include at least: the component ratio of the spectral component, the linear velocity norm and the angular velocity norm; the motion state classification module includes: a detection unit, which is used to detect the motion mode of the handheld laser scanner based on the component ratio of the spectral component; a first determination unit, which is used to determine that the current motion mode of the handheld laser scanner is a steady state mode if the detection result indicates that the component ratio is less than a first preset component threshold; a second determination unit, which is used to determine that the current motion mode of the handheld laser scanner is a non-steady state mode if the detection result indicates that the component ratio is not less than the first preset component threshold; a classification unit, which is used to classify the current motion state of the handheld laser scanner based on the current motion mode of the handheld laser scanner and the linear velocity norm and the angular velocity norm, and generate a current motion state classification result.

[0176] In a preferred implementation manner of this embodiment, the classification unit includes: a first determining subunit, which is used to, when the current motion mode of the handheld laser scanner is a steady state mode, determine that the steady state mode corresponding to the handheld laser scanner is a stationary state if the linear velocity norm is less than a first preset velocity threshold and the angular velocity norm is less than a first preset angular velocity threshold; if the linear velocity norm is not less than the first preset velocity threshold and less than a second preset velocity threshold, and the angular velocity norm is less than the second preset angular velocity threshold, determine that the steady state mode corresponding to the handheld laser scanner is a low-speed moving state; wherein, the second preset angular velocity norm is less than the second preset angular velocity threshold. The speed threshold is greater than the first preset angular velocity threshold; the second determination subunit is used to, when the current motion mode of the handheld laser scanner is a non-steady state mode, if the linear velocity norm is not less than the second preset speed threshold and less than the third preset speed threshold, or the angular velocity norm is not less than the second preset angular velocity threshold and less than the third preset angular velocity threshold, then determine that the non-steady state mode of the handheld laser scanner is a medium-speed moving state; if the linear velocity norm is not less than the fourth preset speed threshold or the angular velocity norm is not less than the third preset angular velocity threshold, then determine that the non-steady state mode of the handheld laser scanner is a high-speed moving state.

[0177] In a preferred implementation manner of this embodiment, the sampling module includes: a first sampling unit, which is used to determine that the sampling density in the current sampling strategy is the maximum preset sampling density of the system when the current motion state classification result is a stationary state, and control the sensor to perform sampling operations on all areas of the target scene according to the maximum preset sampling density to generate point cloud data; wherein the maximum preset sampling density is a preset multiple of the reference point cloud density; a second sampling unit, which is used to determine that the sampling density in the current sampling strategy is the maximum preset sampling density greater than the first preset ratio and less than the maximum preset sampling density when the current motion state classification result is a low-speed moving state; and control the sensor to perform sampling operations on non-critical areas of the target scene according to the determined sampling density, and perform sampling operations on critical areas of the target scene according to the maximum preset sampling density to generate point cloud data; a third sampling unit A sampling unit is used to determine, when the current motion state classification result is a medium-speed moving state, the sampling density in the current sampling strategy to be greater than the maximum preset sampling density of the second preset ratio and less than the maximum preset sampling density; and control the sensor to perform sampling operations on non-critical areas of the target scene according to the determined sampling density, and perform sampling operations on critical areas of the target scene according to the maximum preset sampling density to generate point cloud data; a fourth sampling unit is used to determine, when the current motion state classification result is a high-speed moving state, the sampling density in the current sampling strategy to be greater than the maximum preset sampling density of the third preset ratio and not greater than the maximum preset sampling density of the second preset ratio; and control the sensor to perform sampling operations on all areas of the target scene according to the determined sampling density to generate point cloud data; wherein, the first preset ratio>the second preset ratio>the third preset ratio.

[0178] In a preferred implementation of this embodiment, the second determination module includes: a first generation unit, which is used to: determine the adaptive field radius of the target point for any target point in the point cloud data; perform eigenvalue extraction processing on the field corresponding to the adaptive field radius to obtain eigenvalues; calculate the eigenvalue descriptor corresponding to the eigenvalue; determine the curvature contribution of the field based on the eigenvalue descriptor; determine the edge contribution of the field based on the edge score of the field; weight the curvature contribution and the edge contribution to generate the complexity corresponding to the target point; a second generation unit is used to generate point cloud complexity based on the complexity corresponding to each target point in the point cloud data.

[0179] In a preferred implementation of this embodiment, the adjustment module includes: a first determination unit, which is used to use the frame rate deviation determined by the point cloud complexity and the quality deviation determined by the edge score of the point cloud data as key indicators monitored by the PID controller; an adjustment unit, which is used to adaptively adjust the PID parameters of the PID controller based on the current motion state classification result; a second determination unit, which is used to determine the output result of the PID controller based on the key indicators and PID parameters monitored by the PID controller; a first execution unit, which is used to determine that the system resources of the handheld laser scanner are insufficient if the output result is greater than a preset output threshold, and to perform a reduction operation on the benchmark point cloud density of the current sampling strategy; a second execution unit, which is used to determine that the system resources of the handheld laser scanner are sufficient if the output result is less than the negative value corresponding to the preset output threshold, and to perform an increase operation on the benchmark point cloud density of the current sampling strategy; and a third execution unit, which is used to not perform an adjustment operation on the benchmark point cloud density of the current sampling strategy if the absolute value of the output result is not greater than the preset output threshold.

[0180] In a preferred implementation of this embodiment, the method further includes: a detection module for detecting in real time the reference point cloud density corresponding to the current sampling strategy; an adjustment module for readjusting the reference point density until the reference point cloud density change rate falls within a preset change threshold range if the detection result indicates that the absolute value of the change rate of the reference point cloud density is greater than a preset threshold.

[0181] In a preferred implementation manner of this embodiment, the PID parameters of the PID controller include at least: a proportional coefficient, an integral coefficient, and a differential coefficient; the adjustment unit includes: a first adjustment subunit, which is used to perform an adjustment operation on the integral coefficient in an increasing direction and an adjustment operation on the proportional coefficient in a decreasing direction when the current motion state classification result is a stationary state; a second adjustment subunit, which is used to not perform an adjustment operation on the proportional coefficient, the integral coefficient and the differential coefficient when the current motion state classification result is a low-speed moving state; a third adjustment subunit, which is used to perform an adjustment operation on both the proportional coefficient and the differential coefficient in their respective increasing directions when the current motion state classification result is a medium-speed moving state; and a fourth adjustment subunit, which is used to perform an adjustment operation on the proportional coefficient and the differential coefficient in their respective increasing directions and an adjustment operation on the integral coefficient in a decreasing direction when the current motion state classification result is a high-speed moving state.

[0182] The above-described device can implement the adaptive control method for Gaussian point cloud density provided in one embodiment of the present invention, and possesses the corresponding functional modules and beneficial effects of executing the adaptive control method for Gaussian point cloud density. For technical details not fully described in this embodiment, please refer to the Adaptive Control Method for Gaussian Point Cloud Density Provided in One Embodiment of the Present Invention.

[0183] The present invention also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; the processor is used to read the executable instructions from the memory and execute the instructions to implement the adaptive control method for Gaussian point cloud density described in the present invention.

[0184] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0185] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0186] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enable the processor to execute the steps of the method according to the following embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0187] The computer-readable storage medium may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0188] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0189] The block diagrams of the devices, devices, equipment, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0190] It should also be noted that in the apparatus, device, and method of the present application, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0191] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0192] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless otherwise inconsistent.

[0193] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0194] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for adaptively controlling the density of Gaussian point clouds, characterized in that: Applied to a handheld laser scanner; the method comprises: Determining motion characteristic parameters of the handheld laser scanner based on the sensor data collected at the last moment; Classifying the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; Based on the current sampling strategy corresponding to the current motion state classification result, controlling the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene to generate point cloud data; For any target point in the point cloud data: determining an adaptive domain radius of the target point; performing eigenvalue extraction processing on a domain corresponding to the adaptive domain radius to obtain an eigenvalue; calculating an eigenvalue descriptor corresponding to the eigenvalue; determining a curvature contribution of the domain based on the eigenvalue descriptor; determining an edge contribution of the domain based on an edge score of the domain; weighting the curvature contribution and the edge contribution to generate a complexity corresponding to the target point; generating a point cloud complexity based on the complexity corresponding to each target point in the point cloud data; Based on the point cloud complexity and the current motion state classification result, the reference point cloud density in the current sampling strategy is dynamically adjusted through a PID controller.

2. The method according to claim 1, characterized in that The method of determining the motion characteristic parameters of the handheld laser scanner based on the sensor data collected at the last moment includes: Determining a current fusion state of the handheld laser scanner using an unscented Kalman filter algorithm based on the sensor data collected at the last moment; extracting a linear velocity norm, an acceleration norm, and an angular velocity norm of the handheld laser scanner from the current fusion state; Based on the linear velocity norm and acceleration norm of the handheld laser scanner, a fast Fourier transform algorithm is used to extract the main frequency components and power spectrum density in a preset sliding window; Determining the proportion of spectral components with frequencies less than a preset threshold based on the main frequency components and the power spectral density in the preset sliding window; The component ratio, linear velocity norm and angular velocity norm are used as motion characteristic parameters of the handheld laser scanner.

3. The method according to claim 1, characterized in that The motion characteristic parameters include at least: a component ratio of a spectral component, a linear velocity norm, and an angular velocity norm; the current motion state of the handheld laser scanner is classified based on the motion characteristic parameters to obtain a current motion state classification result; including: detecting a motion pattern of the handheld laser scanner based on a proportion of the spectral components; If the detection result indicates that the absolute value of the difference between the component ratio and 1 is less than a first preset component threshold, determining that the current motion mode of the handheld laser scanner is a steady state mode; If the detection result indicates that the absolute value of the difference between the component ratio and 1 is not less than a first preset component threshold, determining that the current motion mode of the handheld laser scanner is a non-stationary state mode; Based on the current motion mode of the handheld laser scanner, the current motion state of the handheld laser scanner is classified according to the linear velocity norm and the angular velocity norm, and a current motion state classification result is generated.

4. The method according to claim 3, characterized in that The method of classifying the current motion state of the handheld laser scanner based on the current motion mode of the handheld laser scanner and according to the linear velocity norm and the angular velocity norm to generate a current motion state classification result includes: When the current motion mode of the handheld laser scanner is a steady state mode, if the linear velocity norm is less than a first preset velocity threshold and the angular velocity norm is less than the first preset angular velocity threshold, it is determined that the steady state mode corresponding to the handheld laser scanner is a stationary state; if the linear velocity norm is not less than the first preset velocity threshold and less than a second preset velocity threshold, and the angular velocity norm is less than the second preset angular velocity threshold, it is determined that the steady state mode corresponding to the handheld laser scanner is a low-speed moving state; wherein the second preset angular velocity threshold is greater than the first preset angular velocity threshold; When the current motion mode of the handheld laser scanner is a non-steady state mode, if the linear velocity norm is not less than the second preset speed threshold and less than the third preset speed threshold, or the angular velocity norm is not less than the second preset angular velocity threshold and less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a medium-speed moving state; if the linear velocity norm is not less than the fourth preset speed threshold or the angular velocity norm is not less than the third preset angular velocity threshold, then the non-steady state mode of the handheld laser scanner is determined to be a high-speed moving state.

5. The method according to claim 1, wherein The method comprises: controlling the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene based on the current sampling strategy corresponding to the current motion state classification result to generate point cloud data; When the current motion state classification result is a stationary state, the sampling density in the current sampling strategy is determined to be the maximum preset sampling density of the system, and the sensor is controlled to perform sampling operations on all areas of the target scene according to the maximum preset sampling density to generate point cloud data; wherein the maximum preset sampling density is a preset multiple of the reference point cloud density; When the current motion state classification result is a low-speed moving state, determining that the sampling density in the current sampling strategy is greater than a maximum preset sampling density of a first preset ratio and less than the maximum preset sampling density; and controlling the sensor to perform a sampling operation on a non-critical area of ​​the target scene according to the determined sampling density, and to perform a sampling operation on a critical area of ​​the target scene according to the maximum preset sampling density, to generate point cloud data; When the current motion state classification result is a medium-speed moving state, determining that the sampling density in the current sampling strategy is greater than the maximum preset sampling density of the second preset ratio and less than the maximum preset sampling density; and controlling the sensor to perform a sampling operation on a non-critical area of ​​the target scene according to the determined sampling density, and to perform a sampling operation on a critical area of ​​the target scene according to the maximum preset sampling density, to generate point cloud data; When the current motion state classification result is a high-speed moving state, the sampling density in the current sampling strategy is determined to be a maximum preset sampling density greater than the third preset ratio and not greater than the maximum preset sampling density of the second preset ratio; and the sensor is controlled to perform sampling operations on all areas of the target scene according to the determined sampling density to generate point cloud data; wherein, the first preset ratio>the second preset ratio>the third preset ratio.

6. The method according to claim 1, characterized in that The method of dynamically adjusting the density of the reference point cloud in the current sampling strategy by a PID controller based on the point cloud complexity and the current motion state classification result comprises: Using the frame rate deviation determined by the point cloud complexity and the quality deviation determined by the edge score of the point cloud data as key indicators monitored by the PID controller; Adaptively adjusting the PID parameters of the PID controller based on the current motion state classification result; Determining an output result of the PID controller based on key indicators and PID parameters monitored by the PID controller; If the output result is greater than a preset output threshold, it is determined that the system resources of the handheld laser scanner are insufficient, and a reference point cloud density of the current sampling strategy is reduced; If the output result is less than a negative value corresponding to a preset output threshold, it is determined that the system resources of the handheld laser scanner are sufficient, and an increase operation is performed on the reference point cloud density of the current sampling strategy; If the absolute value of the output result is not greater than the preset output threshold, no adjustment operation is performed on the reference point cloud density of the current sampling strategy.

7. The method according to claim 6, characterized in that Also includes: Real-time detection of the reference point cloud density corresponding to the current sampling strategy; If the detection result indicates that the absolute value of the change rate of the reference point cloud density is greater than a preset threshold, the reference point density will be readjusted until the change rate of the reference point cloud density falls within the preset change threshold range.

8. The method according to claim 6, characterized in that The PID parameters of the PID controller include at least a proportional coefficient, an integral coefficient, and a differential coefficient; the PID parameters of the PID controller are adaptively adjusted based on the current motion state classification result; including: When the current motion state classification result is a stationary state, an adjustment operation is performed on the integral coefficient in an increasing direction, and an adjustment operation is performed on the proportional coefficient in a decreasing direction; When the current motion state classification result is a low-speed moving state, no adjustment operation is performed on the proportional coefficient, the integral coefficient, and the differential coefficient; When the current motion state classification result is a medium-speed moving state, adjusting operations are performed on both the proportional coefficient and the differential coefficient along respective increasing directions; When the current motion state classification result is a high-speed moving state, the proportional coefficient and the differential coefficient are adjusted along their respective increasing directions, and the integral coefficient is adjusted along a decreasing direction.

9. An adaptive control device for Gaussian point cloud density, characterized in that: For use with a handheld laser scanner; the device comprises: A first determining module, configured to determine motion characteristic parameters of the handheld laser scanner based on the sensor data collected at a previous moment; A motion state classification module, configured to classify the current motion state of the handheld laser scanner based on the motion characteristic parameters to obtain a current motion state classification result; A sampling module, configured to control the sensor to start performing a sampling operation corresponding to the current sampling strategy on the target scene based on the current sampling strategy corresponding to the current motion state classification result, thereby generating point cloud data; a second determination module, configured to determine the complexity corresponding to each target point in the point cloud data and generate a point cloud complexity; An adjustment module, configured to dynamically adjust the density of the reference point cloud in the current sampling strategy through a PID controller based on the point cloud complexity and the current motion state classification result; The second determination module includes: a first generation unit, which is used to: determine the adaptive field radius of the target point for any target point in the point cloud data; perform eigenvalue extraction processing on the field corresponding to the adaptive field radius to obtain eigenvalues; calculate the eigenvalue descriptor corresponding to the eigenvalue; determine the curvature contribution of the field based on the eigenvalue descriptor; determine the edge contribution of the field based on the edge score of the field; weight the curvature contribution and the edge contribution to generate the complexity corresponding to the target point; and a second generation unit, which is used to generate point cloud complexity based on the complexity corresponding to each target point in the point cloud data.

10. A computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images