Human height measurement method, system, electronic device and storage medium

Abstract

The application provides a human height measurement method, system, electronic equipment and storage medium, the method comprises the following steps: acquiring an RGB-D image sequence containing a human body and a camera internal parameter; extracting a key frame from the image sequence, extracting a human feature point cloud corresponding to a human main region; fitting a ground plane based on a foot region point cloud in the human feature point cloud, constructing a global coordinate system taking the ground as a reference, and transforming the point cloud to the global coordinate system through the camera internal parameter; calculating a single-frame height candidate value according to the height distribution of the transformed point cloud, and removing abnormal candidate values according to the distance between the foot region point cloud and the ground; performing cross-frame statistics on the effective single-frame height candidate values after removing the abnormal candidate values, and obtaining a human height measurement result. The application has the beneficial effects that the measurement precision of the human height is improved, the robustness of the system in a dynamic uncontrolled environment is enhanced, the robustness of the measurement system in a dynamic uncontrolled environment is enhanced, and edge noise and outlier interference of a depth sensor are filtered.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, specifically to a method, system, electronic device, and storage medium for measuring human height. Background Technology

[0002] With the rapid development of computer vision and sensor technology, non-contact human height measurement technology based on depth cameras (such as RGB-D cameras and LiDAR) is gradually being applied in fields such as smart security, human-computer interaction, smart retail, and health monitoring. Compared with traditional contact measurement tools (such as height rulers), depth vision-based solutions have significant advantages such as high automation, good user experience, and remote measurement capability.

[0003] However, in practical applications, especially in uncontrolled open environments, existing technical solutions still suffer from the following major defects and problems: If a world coordinate system based on the real physical ground (i.e., the Z=0 plane) is not established, directly calculating the vertical span of the point cloud will introduce the ground tilt angle or camera mounting angle into the height calculation, causing the measurement results to fluctuate significantly with changes in the person's or camera's position, failing to meet the requirements of high-precision measurement; the lack of an effective mechanism to identify the abnormal state of "foot separation from the ground" (i.e., floating gap) often leads to the lowest suspended point being directly misidentified as the foot, resulting in a significantly underestimation of the height or the generation of extremely large outliers, reducing the system's reliability and availability. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, electronic device, and storage medium for measuring human height, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a method for measuring human height, comprising: Acquire an image sequence containing a human body and camera intrinsic parameters, wherein the image sequence includes color image information and depth information; Keyframes are extracted from the image sequence to extract human feature point clouds; Based on the foot region point cloud in the human feature point cloud, a ground plane is fitted, a global coordinate system with the ground plane as the reference is constructed, and the human feature point cloud is transformed to the global coordinate system through camera intrinsic parameters. The candidate height value for a single frame is calculated based on the height distribution of the transformed human feature point cloud, and abnormal candidate values ​​are eliminated based on the distance between the foot region point cloud and the ground. Cross-frame statistics are performed on the single-frame height candidate values ​​after removing abnormal candidate values ​​to obtain the human height measurement results.

[0006] Preferably, the step of extracting keyframes from the image sequence and extracting human feature point clouds includes: Calculate the median depth of the central region of the depth map; Binarize the depth map based on a preset depth threshold range to generate an initial mask; The initial mask is subjected to morphological processing to obtain an optimized mask; Extract the connected components from the optimized mask, select the target connected component as the main human body region, and extract the corresponding point cloud.

[0007] Preferably, the step of fitting the ground plane based on the foot region point cloud in the human feature point cloud and constructing a global coordinate system with the ground plane as the reference includes: Select a point cloud with a preset proportion that is located in the lower region in the height dimension from the human feature point cloud as the foot region point cloud; The random sampling consensus algorithm is used to perform plane fitting on the point cloud of the foot region to obtain the ground plane equation and ground normal vector; Based on the ground normal vector, construct a rotation matrix and a displacement vector to transform the ground plane into the Z=0 plane and establish a global coordinate system; Using the rotation matrix and translation vector, the point cloud data corresponding to the main body region of the human body is transformed to the global coordinate system.

[0008] Preferably, ground reliability verification is also included: Calculate the angle between the ground normal vector and the preset upward vector. If the angle is greater than the preset angle threshold, the ground is determined to be unreliable. The proportion of points within the statistical plane fitting to the total number of points in the foot region point cloud is calculated. If the proportion is lower than a preset threshold, the ground is determined to be unreliable. If the ground is unreliable, terminate the height calculation for the current frame.

[0009] Preferably, the step of calculating candidate height values ​​for a single frame based on the height distribution of the transformed human feature point cloud includes: The vertical height distribution of human feature point cloud is statistically analyzed; based on the height distribution, the estimated values ​​of the top of the head and the bottom of the feet are determined respectively. The difference between the estimated head height and the estimated foot height is used as a candidate height value for a single frame.

[0010] Preferably, the step of removing outlier candidate values ​​based on the distance between the foot region point cloud and the ground includes: The absolute distance between the estimated foot height and the ground zero plane in the global coordinate system is calculated as the floating gap value; Compare the floating gap value with a preset floating threshold; When the floating gap value is greater than the floating threshold, the single-frame height candidate value corresponding to the current frame is determined to be an abnormal candidate value and is removed.

[0011] Preferably, the step of performing cross-frame statistics on single-frame height candidate values ​​to eliminate abnormal candidate values ​​includes: Count the number of candidate height values ​​in a single frame after removing outlier candidate values; When the number of abnormal candidate values ​​to be removed is less than the preset minimum number of frames, no measurement results are output. When the number of abnormal candidate values ​​removed is greater than or equal to the preset minimum number of frames, the single-frame height candidate values ​​of the removed abnormal candidate values ​​are sorted, and the median or preset percentile value is taken as the human height measurement result.

[0012] Secondly, embodiments of the present invention provide a human height measurement system, comprising: The acquisition module is used to acquire RGB-D image sequences containing the human body and camera intrinsic parameters; The extraction module is used to extract keyframes from image sequences and extract human feature point clouds; The transformation module is used to fit the foot region point cloud in the extracted human feature point cloud to the ground plane, construct a global coordinate system based on the ground plane, and transform the human feature point cloud to the global coordinate system through camera intrinsic parameters. The calculation module is used to calculate the candidate height value of a single frame based on the height distribution of the transformed human feature point cloud, and to remove abnormal candidate values ​​based on the distance between the foot region point cloud and the ground. The output module is used to perform cross-frame statistics on single-frame height candidate values ​​after removing abnormal candidate values ​​to obtain human height measurement results. An embodiment of the present invention provides a human height measurement method, including: Acquire RGB-D image sequences containing human bodies and camera intrinsic parameters; Keyframes are extracted from the image sequence to extract point clouds corresponding to the main human body region; Based on the foot region point cloud in the point cloud, a ground plane is fitted, a global coordinate system with the ground as the reference is constructed, and the point cloud is transformed to the global coordinate system through camera intrinsic parameters; The candidate height values ​​for a single frame are calculated based on the height distribution of the transformed point cloud, and abnormal candidate values ​​are eliminated based on the distance between the point cloud in the foot region and the ground. Valid single-frame height candidate values ​​are statistically analyzed across frames to obtain human height measurement results.

[0013] Preferably, the step of extracting keyframes from the image sequence and extracting point clouds corresponding to the main human body region includes: Calculate the median depth of the central region of the depth map; Binarize the depth map based on a preset depth threshold range to generate an initial mask; The initial mask is subjected to morphological processing to obtain an optimized mask; Extract the connected components from the optimized mask, select the target connected component as the main human body region, and extract the corresponding point cloud.

[0014] Preferably, the step of fitting the ground plane based on the foot region point cloud in the point cloud and constructing a global coordinate system includes: Select the point cloud with the lowest height value from the point cloud corresponding to the main human body area as the foot area point cloud; The random sampling consensus algorithm is used to perform plane fitting on the point cloud of the foot region to obtain the ground plane equation and ground normal vector; Based on the ground normal vector, construct a rotation matrix and a displacement vector to transform the ground plane into the Z=0 plane and establish a global coordinate system; Using the rotation matrix and translation vector, the point cloud data corresponding to the main body region of the human body is transformed to the global coordinate system.

[0015] Preferably, the method further includes ground reliability verification: calculating the angle between the ground normal vector and the preset upward vector; if the angle is greater than the preset angle threshold, the ground is determined to be unreliable; counting the proportion of the number of points in the plane fitting to the total number of point clouds in the foot region; if the proportion is lower than the preset threshold, the ground is determined to be unreliable; when the ground is unreliable, the height calculation of the current frame is terminated.

[0016] Preferably, the step of calculating the candidate height value for a single frame based on the height distribution of the transformed point cloud includes: The vertical height distribution of the point cloud of the main human body region is statistically analyzed; based on the height distribution, the estimated values ​​of the top of the head and the bottom of the feet are determined respectively. The difference between the estimated head height and the estimated foot height is used as a candidate height value for a single frame.

[0017] Preferably, the step of removing outlier candidate values ​​based on the distance between the foot region point cloud and the ground includes: The absolute distance between the estimated foot height and the ground zero plane in the global coordinate system is calculated as the floating gap value; Compare the floating gap value with a preset floating threshold; When the floating gap value is greater than the floating threshold, the single-frame height candidate value corresponding to the current frame is determined to be an abnormal candidate value and is removed.

[0018] Preferably, the cross-frame statistical analysis of valid single-frame height candidate values ​​includes: Count the number of valid candidate height values ​​in a single frame; When the number is less than the preset minimum number of valid frames, no measurement result is output; When the number is greater than or equal to the preset minimum number of valid frames, the candidate height values ​​of the valid single frames are sorted, and the median or preset percentile value is taken as the human height measurement result.

[0019] Secondly, embodiments of the present invention provide a human height measurement system, comprising: The acquisition module is used to acquire RGB-D image sequences containing the human body and camera intrinsic parameters; The extraction module is used to extract keyframes from image sequences and extract point clouds corresponding to the main human body regions; The transformation module is used to fit the ground plane to the extracted point cloud of the foot region, construct a global coordinate system based on the ground, and transform the point cloud to the global coordinate system through camera intrinsic parameters. The calculation module is used to calculate the candidate height value of a single frame based on the height distribution of the transformed point cloud, and to remove abnormal candidate values ​​based on the distance between the foot area point cloud and the ground. The output module is used to perform cross-frame statistics on valid single-frame height candidate values ​​to obtain human height measurement results.

[0020] Thirdly, embodiments of the present invention provide an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein when the processor executes the computer program, it implements the human height measurement method described in any one of the first aspects.

[0021] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the human height measurement method described in any one of the first aspects above.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: This application constructs a world coordinate system based on the physical ground by adaptive ground fitting, which can eliminate systematic errors caused by camera installation tilt and uneven ground, and improve measurement accuracy; by analyzing the distribution continuity of the human body point cloud and the foot area, it can identify and eliminate abnormal frames caused by jumping, suspended objects or noise, thereby enhancing the robustness of the system in dynamic uncontrolled environments; at the same time, it uses quantile statistical estimation to replace traditional extreme value calculation, filtering out edge noise and outlier interference from depth sensors. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating an application environment according to an exemplary embodiment; Figure 2 This is a flowchart illustrating a height measurement method according to an exemplary embodiment; Figure 3 This is a schematic diagram illustrating the operating principle of a height measurement system according to an exemplary embodiment; Figure 4 This is a schematic diagram of a height measurement system framework according to an exemplary embodiment; Figure 5 This is a schematic diagram of an electronic device structure according to an exemplary embodiment. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating the application environment of a human height measurement method according to an exemplary embodiment. For example... Figure 1 As shown, the application environment of this embodiment may include: RGB-D acquisition device 10, human height measurement system 20, and display and interaction terminal 30.

[0026] The RGB-D acquisition device 10 is used to acquire color and depth images of the human body under test in real time, forming an RGB-D image sequence and outputting camera intrinsic parameters. The RGB-D acquisition device 10 can be an image acquisition device with depth perception capabilities, such as a depth camera, 3D LiDAR, or binocular vision camera, and can be deployed in open or semi-open scenarios such as indoor spaces, corridors, stores, security checkpoints, and health monitoring stations.

[0027] The human height measurement system 20 is the core processing unit of this method. The human height measurement system may include: an image acquisition subsystem 21, a data preprocessing subsystem 22, a human point cloud extraction subsystem 23, a ground fitting and coordinate system transformation subsystem 24, a single-frame height calculation subsystem 25, a cross-frame statistics and result output subsystem 26, and a system control and interaction subsystem 27. These subsystems cooperate to complete image acquisition, point cloud processing, ground calibration, height estimation, anomaly removal, and result output, achieving high-precision and robust automatic human height measurement in dynamic, uncontrolled environments. This system can run on embedded chips, local computers, edge computing units, or cloud servers, supporting offline processing and real-time streaming processing.

[0028] The display and interactive terminal 30 is used to show users height measurement results, system status, and abnormal prompts. It may include an LCD screen, a touch screen, a mobile terminal, host computer software, etc., and supports local data display, record storage, and remote uploading.

[0029] In one alternative implementation, the RGB-D acquisition device, the human height measurement system, and the display and interactive terminal can be connected via local area network, wide area network, USB, serial port, or 4G / 5G / Wi-Fi wireless communication to achieve real-time transmission of image data, camera parameters, height results, and control commands.

[0030] In practical applications, through the collaborative work of the RGB-D acquisition device 10, the human height measurement system 20 and the display terminal 30, automated, non-contact, and high-precision human height measurement can be completed in a dynamic and uncontrolled environment, which is suitable for various scenarios such as intelligent security, smart retail, health check-up, human-computer interaction, and public facilities.

[0031] Furthermore, it should be noted that Figure 1 only shows one application environment of the human height measurement method provided in this disclosure.

[0032] It should be noted that the following diagram shows one possible sequence of steps, and it is not actually required to strictly follow the steps. In this order. Some steps can be performed in parallel without interdependence. The user information (including but not limited to user device information, user personal information, user behavior information, etc.) and data (including but not limited to data used for display, training data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0033] Figure 2 This is a flowchart illustrating a method for measuring human height according to an exemplary embodiment. Figure 2 As shown, the steps may include the following.

[0034] Please see Figure 1 , Figure 2 This invention provides a technical solution: a method for measuring human height, which is mainly completed collaboratively by a depth camera acquisition end and a back-end processing module (which can run on an embedded device, PC, or cloud server). The core idea of ​​this method is to utilize the depth information of consecutive frames, eliminate the influence of camera pose through geometric constraints, and employ a statistical quantile strategy for noise reduction, ultimately outputting a high-precision height measurement value. Specifically, the method includes the following steps: S100: Acquire RGB-D image sequences containing human figures and camera intrinsic parameters.

[0035] In the embodiments described in this specification, the RGB-D image sequence can be a continuous frame sequence generated by a depth image acquisition device, consisting of a color image and a depth image of the human body being measured. Camera intrinsic parameters may include calibration parameters used to convert pixel coordinates into three-dimensional spatial coordinates, such as camera focal length, principal point coordinates, and distortion coefficients.

[0036] In one possible implementation, RGB-D image sequences and camera intrinsic parameters can be acquired from a depth acquisition device. The depth acquisition device can include various devices with depth sensing capabilities, such as RGB-D depth cameras, 3D LiDAR, and binocular vision cameras. After receiving the image data, the image acquisition subsystem can transmit the data to the data processing subsystem for further processing.

[0037] In one alternative implementation, the data processing subsystem may need to interact with the depth acquisition device multiple times to complete steps such as camera parameter reading, image format verification, and depth data validity detection. For example, during the acquisition process, the depth acquisition device may need to provide information such as the current frame rate, exposure status, and effective depth area. These steps can be achieved through multiple interactions between the depth acquisition device and the image acquisition subsystem.

[0038] In another possible implementation, RGB-D image sequences and camera intrinsic parameters can be acquired through real-time communication with the depth acquisition device. For example, the depth acquisition device submits image data via continuous frame output, and the image acquisition subsystem can establish a connection with the depth acquisition device through a network interface or data interface to receive and verify the image data, protecting its integrity and correctness.

[0039] In one alternative implementation, camera intrinsic parameters and equipment configuration information can be managed through a database within the human height measurement system. This database can record the intrinsic parameters, distortion coefficients, and installation posture information of all calibrated depth acquisition devices. After receiving an RGB-D image sequence, the image acquisition subsystem can query this database to obtain camera intrinsic parameters to ensure the accuracy of subsequent coordinate transformations and point cloud computing.

[0040] In one possible implementation, when acquiring the RGB-D image sequence, all image frames that are currently in a normal state and have valid depth can be selected first, so that the subsequent processing is more efficient and stable.

[0041] In practical applications, by acquiring RGB-D image sequences and camera intrinsic parameters, 3D point cloud data can be quickly constructed, ensuring the accuracy of subsequent point cloud transformations and height calculations, thereby improving measurement accuracy and overall system stability. In one embodiment of this invention, an RGB-D camera (such as Intel RealSense or Kinect) is used to acquire RGB and depth images, forming an image sequence. The RGB images provide color information, and the depth images provide the distance between each pixel and the camera. The image sequence is captured at a predetermined frame rate, continuously acquiring RGB and depth data at each moment. Intrinsic parameters are obtained through a camera calibration process. Multiple images are captured using a checkerboard calibration board of known size, corner points are extracted from the images, and the camera's focal length, principal point coordinates, and distortion coefficients are calculated using a calibration algorithm. These intrinsic parameters are used for subsequent conversion from image coordinates to 3D spatial coordinates. Using the camera intrinsic parameters, the pixel coordinates and depth values ​​in the depth images are combined to convert them into 3D spatial coordinates, generating point cloud data. The point cloud represents the 3D structure of objects in the scene, providing the spatial data required for subsequent operations.

[0042] S200: Extract keyframes from the image sequence and extract the human feature point cloud corresponding to the main human body region. The specific process of this step is as follows: S201: Calculate the median depth of the central region of the depth map.

[0043] In the embodiments described in this specification, keyframes can be image frames selected from RGB-D image sequences that are clear, have a complete human body, valid depth, and are suitable for subsequent point cloud processing. The point cloud corresponding to the main human body region can be a three-dimensional point set containing only the human body under test, obtained by converting depth information and pixel coordinates through camera intrinsic parameters.

[0044] In a preferred embodiment of the present invention, the expected depth of the human body can be determined based on the median depth of the central region of the depth map. The depth map is then binarized based on a preset depth threshold to generate an initial mask. Morphological processing is performed on the initial mask to eliminate noise and holes, resulting in an optimized mask. Connected components are extracted based on the optimized mask. A comprehensive score is calculated based on area ratio, center offset, depth deviation, and edge penalty. The connected component with the highest score is selected as the main human body region, and the corresponding 3D point cloud is extracted.

[0045] In one optional implementation, the data preprocessing subsystem can perform inter-frame stability judgment on the image sequence, remove abnormal frames with motion blur, missing depth, or incomplete human body, and retain only frames whose quality meets preset conditions as key frames, so as to improve the reliability of point cloud extraction.

[0046] In another possible implementation, the human body area can be made cleaner by ignoring the row area at the bottom of the image with a preset ratio, thus suppressing ground reflection spots and near-field clutter interference.

[0047] In practical applications, by extracting keyframes and accurately extracting the point cloud of the human body, background interference and invalid data can be reduced, thereby improving the efficiency and accuracy of subsequent ground fitting and height calculation.

[0048] This step includes, in one specific implementation, firstly, defining a predefined central region of the image based on the prior assumption that the human body is typically located at the center of the field of view. Then, counting the depth values ​​of all valid depth pixels within this region and calculating their median. This median is determined as the "expected subject depth" of the current frame.

[0049] S202: Binarize the depth map based on a preset depth threshold range to generate an initial mask.

[0050] In a preferred embodiment of the present invention, step S202 specifically includes: constructing an adaptive depth gating mechanism based on the expected subject depth calculated in step S201. Specifically, a near-end depth threshold and a far-end depth threshold are set to form an effective depth range centered on the expected subject depth.

[0051] In a specific embodiment of this example, the near-end depth threshold ranges from 0.20 meters to 0.40 meters, preferably 0.30 meters; the far-end depth threshold ranges from 0.30 meters to 0.60 meters, preferably 0.50 meters. By traversing the image pixels, pixels whose depth values ​​fall within this effective range are marked as foreground candidate points, generating an initial binary mask; pixels whose depth values ​​are outside the range are marked as background points. Through this step, foreground interference that is too close and background scenes that are too far away can be effectively eliminated, retaining candidate pixels of the human body and its adjacent areas.

[0052] S203: Perform morphological processing on the initial mask to obtain an optimized mask.

[0053] In a preferred embodiment of the present invention, the purpose of this step is to perform morphological filtering on the initial mask to eliminate noise in the binary mask and optimize the connectivity of the human body region. The specific process includes: First, it is preferable to ignore the bottom row area of ​​the image (e.g., the bottom 6%) to eliminate ground reflections and near-field clutter interference from the camera lens.

[0054] Subsequently, closing and opening operations are performed sequentially on the remaining areas. The closing operation fills in gaps caused by depth loss or holes within the human body area, ensuring the integrity of the area; the opening operation removes isolated noise pixels and smooths the edge contours of the human body area. After these processes, an "optimized mask" with smooth edges, filled interior, and a clean background is generated.

[0055] Furthermore, S204: Extract connected components from the optimized mask, select the target connected component as the main human body region, and extract the corresponding point cloud.

[0056] In a preferred embodiment of the present invention, in order to select a unique human body subject region from multiple candidate regions, a comprehensive scoring model with multi-dimensional features is constructed. This scoring model includes at least four dimensions: area ratio option, center offset term, depth deviation term, and edge penalty term.

[0057] Furthermore, in a preferred embodiment of the present invention, the area ratio term can be set as a positive correlation term, prioritizing the retention of targets that occupy a proportion of the image, which conforms to the characteristic that the human body usually occupies the main area in the composition.

[0058] Furthermore, a center offset term can be set as a penalty term to calculate the Euclidean distance between the geometric center of the connected domain and the image center. The larger the distance, the higher the penalty score, thereby suppressing non-subject targets that deviate from the center of the field of view.

[0059] Furthermore, a depth deviation term can be set as a penalty term to calculate the absolute value of the difference between the average depth of pixels in the connected region and the aforementioned "expected subject depth". The larger the absolute difference, the more likely the region is to be background interference, and the lower the score.

[0060] Furthermore, the edge-touching penalty term can be set as a logical penalty term. If a connected component touches the image boundary, the penalty mechanism is triggered because the area that touches the boundary usually represents a truncated, incomplete target.

[0061] In a specific embodiment of the present invention, for the th connected component in the th frame, the weighted calculation formula for its comprehensive score based on the four features is as follows: The physical meaning of each parameter and the optimal implementation logic are as follows: This is the area ratio term: representing the proportion of connected component area to the total image area. This term is positively correlated with the score (weight). This is used to prioritize larger targets, which aligns with the characteristic that the human body usually occupies a certain proportion in the image; Center offset term: Represents the Euclidean distance between the geometric center of the connected component and the image center. This term serves as a penalty term (weight). ≥0), used to suppress non-primary targets that deviate from the center of the field of view; The depth deviation term represents the absolute value of the difference between the average depth value of pixels within the connected region and the predicted subject depth. This term serves as a penalty term (weight). This is used to remove areas whose depth distribution does not match the estimated value of the central area (such as side walls or pillars).

[0062] This is a boundary-touching penalty term: it takes a value of 1 when a connected component touches the four boundaries of the image, and 0 otherwise. This term (weight) This is used to exclude targets that are not fully displayed or are truncated by the screen.

[0063] The system calculates the comprehensive score for all connected components and selects... The connected component with the highest score is selected as the final human body feature region for the current frame. If none of the connected components reach the preset threshold, the connected component with the largest area is degenerately selected as the human body feature region.

[0064] Finally, using the pixel coordinate mask corresponding to the connected component, the corresponding human body point cloud data is extracted from the original depth map for subsequent steps of ground fitting and height calculation.

[0065] S300: Fit the ground plane based on the foot region point cloud in the human feature point cloud, construct a global coordinate system with the ground plane as the reference, and transform the human feature point cloud to the global coordinate system through camera intrinsic parameters.

[0066] In the embodiments described in this specification, the foot region point cloud can be a preset proportion of points selected from the human body point cloud, located in the lower region in the height dimension. The global coordinate system can be a world coordinate system with the ground plane as the Z=0 reference and the vertical direction as the height direction, used to eliminate systematic errors caused by the camera mounting tilt angle and ground tilt.

[0067] In one possible implementation, a predetermined proportion of points in the lower region of the human body feature point cloud are selected as the foot region point cloud in terms of height dimension. A random sampling consensus algorithm is used to perform planar fitting on the foot point cloud to obtain the ground plane equation and ground normal vector. A rotation matrix and translation vector are constructed based on the ground normal vector, and the ground plane is normalized to the Z=0 plane, thus establishing a global coordinate system. Using the rotation matrix and translation vector, the human body point cloud is transformed from the camera coordinate system to the global coordinate system.

[0068] In one optional implementation, the system may perform a ground reliability check: calculate the angle between the ground normal vector and a preset upward vector, and statistically analyze the proportion of in-plane fitting points; if the angle is too large or the proportion of in-plane points is too low, the ground is determined to be unreliable, and the calculation of the current frame is terminated. In a preferred embodiment of the present invention, this step specifically includes: In one specific implementation, the point cloud with the lowest height value in the point cloud corresponding to the main body area of ​​the human body is selected as the foot area point cloud. The random sampling consensus algorithm is used to perform plane fitting on the point cloud of the foot region to obtain the ground plane equation and ground normal vector; Based on the ground normal vector, construct a rotation matrix and a displacement vector to transform the ground plane into the Z=0 plane and establish a global coordinate system; Using the rotation matrix and translation vector, the point cloud data corresponding to the main body region of the human body is transformed to the global coordinate system.

[0069] In a specific embodiment of the present invention, the point cloud data processing procedure for the feet and the ground is as follows: S301: First, the ground estimation frames are selected. To avoid interference from accidental anomalous frames (such as severe motion blur or temporary occlusion frames) on the ground fitting accuracy, this embodiment does not use single-frame instantaneous values, but instead selects the frame corresponding to the median index from all available frame sequences as the ground estimation frame. This strategy utilizes the statistical robustness of the median to ensure that the selected frame is representative of the time series.

[0070] Furthermore, in the selected ground estimation frame, the bottom preset region of the human body detection bounding box (defined as the foot region) is locked, and pixels with valid depth values ​​within this region are extracted. Subsequently, based on the standard pinhole camera imaging model and pre-stored camera intrinsic parameters, these two-dimensional pixel coordinates and their corresponding physical depth values ​​are back-projected into the three-dimensional camera coordinate system to generate an initial three-dimensional point cloud representing the foot contact area.

[0071] Furthermore, to eliminate outliers that significantly deviate from the main distribution, the initial point cloud needs to be quantile-clipping before planar fitting. Specifically, along the optical axis of the camera's intrinsic coordinate system, the lower and upper quantiles of the point cloud depth are calculated, retaining only point cloud data within this range. Preferably, the pruning range is set between the lower 2% and upper 98% of the data distribution, thereby obtaining a clean foot region point cloud for subsequent fitting.

[0072] Furthermore, the RANSAC (Random Sample Consensus RANSAC) algorithm is used to perform plane fitting on the preprocessed foot region point cloud. Through iterative calculation, the optimal ground plane parameters are obtained. The ground plane is uniquely determined by the ground normal vector and the plane bias term, and its mathematical expression satisfies the plane equation: ,in It is the normal vector. For point cloud coordinates, For plane offset terms, This is a transpose operation.

[0073] Furthermore, to ensure the consistency of the coordinate system construction, this embodiment introduces a normal vector direction constraint mechanism. According to the coordinate system conventions of this application, the theoretical upward direction in the camera coordinate system is used as the reference vector. The dot product of the fitted ground normal vector and this reference vector is calculated: if the dot product is less than zero, it indicates that the current normal vector points away from the sky direction, and the normal vector and its corresponding plane offset term are inverted; if the dot product is greater than or equal to zero, the original parameters remain unchanged. Through the above constraints, the finally determined ground normal vector is forced to point in the vertical upward direction of the world coordinate system, avoiding coordinate system confusion caused by the uncertainty of the normal vector direction. According to the coordinate system conventions of this application, the theoretical upward direction in the camera coordinate system is defined as the reference vector. The system calculates the dot product of the fitted normal vector and this reference vector: If the dot product result is less than zero, it indicates that the current normal vector points downward (i.e., the direction of the ground). The system performs an inversion operation, inverting both the normal vector and its corresponding plane offset term. If the dot product result is greater than or equal to zero, then keep the parameters unchanged.

[0074] This step ensures that the final determined ground normal vector always points strictly upward (towards the zenith) in the world coordinate system, eliminating the risk of coordinate system confusion caused by uncertainty in the direction of the normal vector.

[0075] Secondly, S302: After obtaining the ground normal vector with unified orientation, construct the rigid body transformation relationship from the camera coordinate system to the world coordinate system, including the rotation matrix and translation vector, so that the vertical axis (Z-axis) of the world coordinate system is aligned with the calibrated ground normal vector.

[0076] In a specific embodiment of the present invention, the specific construction process is as follows: Furthermore, based on the ground normal vector after unifying the direction, a rotation matrix is ​​calculated to align the upward vector of the camera coordinate system to the opposite direction of the ground normal vector (i.e., perpendicular to the ground upward).

[0077] Furthermore, a reference point on the ground plane is calculated, which is equal to the product of the negative plane offset term and the ground normal vector. A translation vector is constructed based on this reference point, such that the reference point is mapped to the origin of the world coordinate system after transformation, thereby normalizing the ground plane equation to a plane with Z equal to zero.

[0078] Finally, in step S303: using the rotation matrix and translation vector constructed above, the point cloud feature data corresponding to the human body's main body region, originally belonging to the camera coordinate system, is transformed into the global world coordinate system through rigid body transformation equations. After the transformation, the ground height in the global world coordinate system is defined as zero, eliminating the influence of camera installation tilt angle and position deviation.

[0079] Furthermore, to prevent incorrect ground fitting results from causing height measurement failure, this embodiment includes a ground reliability verification step after constructing the global coordinate system: Angle verification: Calculate the angle between the ground normal vector and the preset gravity upward vector. If the angle is greater than the preset angle threshold (e.g., alignment is less than 0.70), the ground is determined to be unreliable. Interior point verification: Statistically calculate the proportion of interior points participating in plane fitting to the total number of point clouds in the foot region. If this proportion is lower than the preset interior point threshold (e.g., 0.35), the ground is determined to be unreliable.

[0080] If any of the above checks fail, the system will terminate the height calculation process for the current frame and will not output the measurement results, thereby ensuring the reliability of the system's engineering delivery.

[0081] In practical applications, ground fitting and coordinate system normalization can significantly reduce measurement bias caused by camera pose and uneven ground, thereby improving the consistency and accuracy of height results.

[0082] S400: Calculate candidate height values ​​for a single frame based on the height distribution of the transformed human feature point cloud, and remove abnormal candidate values ​​based on the distance between the foot region point cloud and the ground.

[0083] In the embodiments of this specification, the candidate height value for a single frame can be the difference between the height of the top of the head and the height of the feet, obtained based on the height distribution of the human body point cloud in the current frame. Abnormal candidate values ​​can be calculation results that do not conform to the true height due to human jumping, tiptoeing, hanging in the air, or abnormal posture.

[0084] In one possible implementation, the vertical height distribution of the human body point cloud in the global coordinate system is statistically analyzed to determine the estimated head height corresponding to the high quantile and the estimated foot height corresponding to the low quantile. The difference between the two is used as a candidate height value for a single frame. The absolute distance between the foot height and the ground zero plane is calculated as the floating gap value. If the floating gap value is greater than a preset threshold, the current frame is determined to be an abnormal frame and the corresponding candidate value is removed.

[0085] In one alternative implementation, the point cloud height distribution can be quantile-clipping to filter out extreme outliers, making the top and bottom estimations more stable and reducing noise interference from the depth sensor.

[0086] In practical applications, by identifying the floating state of the feet and removing abnormal frames, it is possible to avoid measurement errors or jumps caused by the dynamic posture of the human body, thereby improving the robustness of the system in uncontrolled environments. In a preferred embodiment of the present invention, step S400 specifically includes: statistically analyzing the height distribution of the point cloud of the human body's main region in the vertical direction; determining the estimated height of the top of the head and the estimated height of the feet based on the height distribution; using the difference between the estimated height of the top of the head and the estimated height of the feet as a candidate height value for a single frame; calculating the absolute distance between the estimated height of the feet and the zero plane of the ground in the global coordinate system as a floating gap value; comparing the floating gap value with a preset floating threshold; when the floating gap value is greater than the floating threshold, determining that the candidate height value for the current frame is an abnormal candidate value and removing it.

[0087] S500: Perform cross-frame statistics on valid single-frame height candidate values ​​after removing outlier candidate values ​​to obtain human height measurement results.

[0088] In the embodiments of this specification, a valid single-frame height candidate value can be a reliable height data that meets the conditions of ground reliability and foot contact after anomaly removal. Cross-frame statistics can be used to estimate the central tendency of valid results from multiple frames to output a stable and interference-resistant final height.

[0089] In one possible implementation, the total number of valid single-frame height candidate values ​​is counted. If the number of valid values ​​is less than the preset minimum number of frames, the measurement result is not output. If the frame count requirement is met, the valid candidate values ​​are sorted, and the median or preset percentile value is taken as the final human height measurement result.

[0090] In one alternative implementation, the system can perform sliding window statistics on results from multiple consecutive frames, update and output a stable height in real time, and support dynamic measurement and rapid response.

[0091] In practical applications, the multi-frame statistical fusion strategy can further suppress depth noise, posture fluctuations, and transient interference, achieving high-precision and high-stability automatic measurement of human height. In a preferred embodiment of the present invention, step S500 specifically includes: counting the number of valid single-frame height candidate values; when the number is less than the preset minimum number of valid frames, not outputting the measurement result; when the number is greater than or equal to the preset minimum number of valid frames, sorting the valid single-frame height candidate values, and taking the median or preset quantile as the human height measurement result.

[0092] In addition, please see Figure 3 Embodiments of the present invention provide a human height measurement system 30, including: FIG3 is a block diagram of a human height measurement system according to an exemplary embodiment. Embodiments of the present disclosure provide a human height measurement system 40, including: an acquisition module 101, an extraction module 102, a conversion module 103, a calculation module 104, and an output module 105.

[0093] The acquisition module 101 is used to acquire an RGB-D image sequence containing a human body and camera intrinsic parameters; the RGB-D image sequence contains color image information and depth image information, and the camera intrinsic parameters include camera focal length, principal point coordinates and distortion coefficients.

[0094] The extraction module 102 is used to extract keyframes from the image sequence and extract human feature point clouds. The extraction module determines the main human body region and extracts the corresponding point clouds based on the median depth of the center region of the depth map, depth threshold binarization, morphological processing and comprehensive scoring of connected components.

[0095] The conversion module 103 is used to fit the ground plane to the foot region point cloud in the extracted human feature point cloud, construct a global coordinate system based on the ground plane, and transform the human feature point cloud to the global coordinate system through camera intrinsic parameters; the conversion module is also configured to perform ground reliability verification, and terminate the calculation of the current frame when the verification does not meet the conditions.

[0096] The calculation module 104 is used to calculate the candidate height value of a single frame based on the height distribution of the transformed human feature point cloud, and to remove abnormal candidate values ​​based on the distance between the foot area point cloud and the ground; the calculation module realizes abnormal frame identification and filtering by calculating the floating gap value.

[0097] The output module 105 is used to perform cross-frame statistics on the single-frame height candidate values ​​after removing abnormal candidate values ​​to obtain the human height measurement result; when the number of effective frames meets the preset conditions, the output module outputs the final height using the median or preset quantile value.

[0098] In one alternative implementation, the above modules can be implemented by software, hardware, firmware, or a combination thereof. The modules interact with each other via a bus or interface to collaboratively complete human height measurement in a dynamic, uncontrolled environment.

[0099] The acquisition module 101 is used to acquire RGB-D image sequences containing human bodies and camera intrinsic parameters; Extraction module 102 is used to extract keyframes from image sequences and extract point clouds corresponding to the main human body region; The transformation module 103 is used to fit the ground plane to the extracted point cloud of the foot region, construct a global coordinate system based on the ground, and transform the point cloud to the global coordinate system through camera intrinsic parameters. The calculation module 104 is used to calculate the candidate height value of a single frame based on the height distribution of the transformed point cloud, and to remove abnormal candidate values ​​based on the distance between the foot area point cloud and the ground. The output module 105 is used to perform cross-frame statistics on valid single-frame height candidate values ​​to obtain human height measurement results.

[0100] Please see Figure 4 , Figure 4 A schematic diagram of the mechanism of an electronic device 20 that can implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of control devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or claimed herein.

[0101] Electronic device 20 includes at least one processor 21 and a memory, such as read-only memory (ROM) 22 and random access memory (RAM) 23, communicatively connected to at least one processor 21. The memory stores computer programs executable by at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 22 or loaded from storage unit 28 into the RAM 13. The RAM 23 may also store various programs and data required for the operation of electronic device 20. The processor 21, ROM 22, and RAM 23 are interconnected via bus 24. Input / output (I / O) interface 25 is also connected to bus 24.

[0102] Multiple components in electronic device 20 are connected to I / O interface 25, including: input unit 26, such as keyboard, mouse, etc.; output unit 27, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 29, such as network card, modem, wireless transceiver, etc. Communication unit 29 allows electronic device 20 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0103] Processor 21 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 21 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 21 performs the various methods and processes described above.

[0104] In some embodiments, the methods described above can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 28. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 20 via ROM 22 and / or communication unit 29. When the computer program is loaded into RAM 23 and executed by processor 21, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, processor 21 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0105] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0106] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0107] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0108] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0109] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0110] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for measuring human height, characterized in that, include: Acquire an RGB-D image sequence containing a human body and camera intrinsic parameters, wherein the image sequence contains color image information and depth information; Keyframes are extracted from the image sequence to extract point clouds of human features corresponding to the main human body region; Based on the foot region point cloud in the human feature point cloud, a ground plane is fitted, a global coordinate system with the ground plane as the reference is constructed, and the human feature point cloud is transformed to the global coordinate system through camera intrinsic parameters. The candidate height value for a single frame is calculated based on the height distribution of the transformed human feature point cloud, and abnormal candidate values ​​are eliminated based on the distance between the foot region point cloud and the ground. Cross-frame statistics are performed on the single-frame height candidate values ​​that have been effectively eliminated to obtain the human height measurement results.

2. The method for measuring human height according to claim 1, characterized in that, The step of extracting keyframes from the image sequence and extracting human body feature point clouds includes: Calculate the median depth of the central region of the depth map; Binarize the depth map based on a preset depth threshold range to generate an initial mask; The initial mask is subjected to morphological processing to obtain an optimized mask; Extract the connected components from the optimized mask, select the target connected component as the main human body region, and extract the corresponding point cloud.

3. The method for measuring human height according to claim 1, characterized in that, The process of fitting a ground plane based on the foot region point cloud in the human feature point cloud and constructing a global coordinate system with the ground plane as the reference includes: Select the point cloud with the lowest height value in the lower region from the feature point cloud corresponding to the main human body region as the foot region point cloud. The random sampling consensus algorithm is used to perform plane fitting on the point cloud of the foot region to obtain the ground plane equation and ground normal vector; Based on the ground normal vector, construct a rotation matrix and a displacement vector to transform the ground plane into the Z=0 plane and establish a global coordinate system; Using the rotation matrix and translation vector, the point cloud data corresponding to the features of the human body's main body region is transformed to the global coordinate system.

4. The method for measuring human height according to claim 3, characterized in that, It also includes ground reliability verification: Calculate the angle between the ground normal vector and the preset upward vector. If the angle is greater than the preset angle threshold, the ground is determined to be unreliable. The proportion of points within the statistical plane fitting to the total number of points in the foot region point cloud is calculated. If the proportion is lower than a preset threshold, the ground is determined to be unreliable. If the ground is unreliable, terminate the height calculation for the current frame.

5. The method for measuring human height according to claim 1, characterized in that, The calculation of candidate height values ​​for a single frame based on the height distribution of the transformed human feature point cloud includes: The height distribution of the feature point cloud of the human body's main body region in the vertical direction is statistically analyzed; based on the height distribution, the estimated values ​​of the top of the head and the bottom of the feet are determined respectively. The difference between the estimated head height and the estimated foot height is used as a candidate height value for a single frame.

6. The method for measuring human height according to claim 5, characterized in that, The step of removing outlier candidate values ​​based on the distance between the foot region point cloud and the ground includes: The absolute distance between the estimated foot height and the ground zero plane in the global coordinate system is calculated as the floating gap value; Compare the floating gap value with a preset floating threshold; When the floating gap value is greater than the floating threshold, the single-frame height candidate value corresponding to the current frame is determined to be an abnormal candidate value and is removed.

7. The method for measuring human height according to claim 6, characterized in that, The method of performing cross-frame statistics on single-frame height candidate values ​​to remove outlier candidate values ​​includes: Count the number of valid single-frame height candidate values ​​after removing outlier candidate values; When the number of abnormal candidate values ​​to be removed is less than the preset minimum number of valid frames, no measurement result is output; When the number of abnormal candidate values ​​removed is greater than or equal to the preset minimum number of valid frames, the valid single-frame height candidate values ​​of the removed abnormal candidate values ​​are sorted, and the median or preset percentile value is taken as the human height measurement result.

8. A human height measurement system, characterized in that, include: The acquisition module is used to acquire RGB-D image sequences containing the human body and camera intrinsic parameters; The extraction module is used to extract keyframes from image sequences and extract feature point clouds corresponding to the main human body region; The transformation module is used to fit the foot region point cloud in the extracted human feature point cloud to the ground plane, construct a global coordinate system based on the ground plane, and transform the human feature point cloud to the global coordinate system through camera intrinsic parameters. The calculation module is used to calculate the candidate height value of a single frame based on the height distribution of the transformed human feature point cloud, and to remove abnormal candidate values ​​based on the distance between the foot region point cloud and the ground. The output module is used to perform cross-frame statistics on valid single-frame height candidate values ​​after removing abnormal candidate values, and obtain the human height measurement results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, When the processor executes the computer program, it implements the human height measurement method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the human height measurement method according to any one of claims 1 to 7.

Images