Search

CN-121998867-A - ToF imaging-based reflection cavity filling and overexposure correction method and system

CN121998867ACN 121998867 ACN121998867 ACN 121998867ACN-121998867-A

Abstract

The invention provides a reflective cavity filling and overexposure correction system and method based on ToF imaging, and aims to solve the problems of reflective cavities and overexposure common to a ToF sensor in a complex scene and improve the accuracy and reliability of depth data. After the ToF sensor collects depth data of a target area, median filtering and Gaussian filtering are performed through a depth data preprocessing module, noise is removed, and an image is smoothed. The reflection cavity filling module performs linear interpolation or texture restoration by detecting invalid values (such as 0 or 65535) in the depth data and performing proportional conversion correction on partial areas by utilizing the peripheral normal reflection data so as to effectively fill the reflection cavity. The overexposure correction module identifies an abnormal value (e.g., 0 or an abnormally small value), analyzes the boundary continuity of the overexposure region, and accordingly compensates the depth pixel value to solve the overexposure problem. The user interface module displays the corrected depth data and provides interactive functions, allowing a user to adjust correction parameters to adapt to different scene requirements. The method is excellent in traffic detection in the field of large traffic (such as airports, subways and the like), provides high-quality data support for intelligent traffic, and remarkably improves the intelligent level and robustness of the system.

Inventors

  • ZHANG WENQI
  • Request for anonymity

Assignees

  • 中科艾特(苏州)机器人有限公司

Dates

Publication Date
20260508
Application Date
20250528

Claims (10)

  1. 1. A method and a system for filling and overexposure correction of a reflective cavity based on ToF imaging are characterized in that the system comprises: a ToF sensor for acquiring depth data of a target area; A depth data preprocessing module for preprocessing the acquired depth data, including filtering and denoising; a reflective hole filling module for detecting and filling a reflective hole region in the depth data; an overexposure correction module for detecting and correcting an overexposed region in the depth data; A user interface module for displaying the modified depth data and providing a user interaction function; the corrected depth data output module is used for outputting corrected depth data for subsequent processing or display.
  2. 2. The system of claim 1, wherein the reflective cavity filling module comprises: An abnormal region detection unit for identifying invalid values (e.g., 0 or 65535) in the depth data; a neighborhood information-based shim for linear interpolation or texture restoration using surrounding normal reflection data; Scaling correction means for scaling the partially reflective cavity region to generate a corrected depth value.
  3. 3. The system of claim 1, wherein the overexposure correction module comprises: An abnormal region detection unit for identifying an abnormal value (e.g., 0 or a small value of an abnormality, such as 10, 20, etc.) in the depth data; A boundary continuity analysis unit for analyzing the boundary continuity of the overexposed region and determining the boundary range of the overexposed region; a depth pixel value correction unit for correcting the depth pixel value based on the boundary continuity analysis result.
  4. 4. The system of claim 1, wherein the depth data preprocessing module comprises: A median filtering unit for removing noise points in the depth data; A gaussian filter unit for smoothing the depth image; A denoising processing unit for further removing noise points in the depth image.
  5. 5. The system of claim 1, wherein the user interface module comprises: a data presentation unit for displaying the corrected depth data to a user; a user interaction unit for allowing a user to adjust parameters of the correction algorithm; the parameter adjustment and feedback unit is used for updating the parameters of the correction algorithm according to the parameter adjustment instruction input by the user.
  6. 6. A method for filling and overexposure correction of a reflective cavity based on ToF imaging is characterized by comprising the following steps: Acquiring depth data of the target area using a ToF sensor; preprocessing the acquired depth data, including filtering and denoising; Detecting and filling a reflective cavity region in the depth data; detecting and correcting overexposed areas in the depth data; displaying the modified depth data and providing a user interaction function; outputting the corrected depth data for subsequent processing or display.
  7. 7. The method of claim 6, wherein the step of filling the reflective cavity region comprises: Identify invalid values in the depth data (e.g., 0 or 65535); Linear interpolation or texture restoration using the surrounding normal reflection data; scaling the partially reflective cavity region to generate a corrected depth value.
  8. 8. The method of claim 6, wherein the step of correcting the overexposed region comprises: Identifying outliers (e.g., 0 or outlier small values such as 10, 20, etc.) in the depth data; Analyzing the boundary continuity of the overexposed region, and determining the boundary range of the overexposed region; based on the boundary continuity analysis results, the patch corrected depth pixel values.
  9. 9. The method of claim 6, wherein the preprocessing step comprises: median filtering the original depth data to remove noise points; performing Gaussian filtering on the depth data after the median filtering to smooth the depth image; further removing noise points in the depth image.
  10. 10. The method of claim 6, wherein the user interaction function comprises: Displaying the corrected depth data to a user; allowing the user to adjust the parameters of the correction algorithm; updating parameters of the correction algorithm based on the parameter adjustment instruction input by the user.

Description

ToF imaging-based reflection cavity filling and overexposure correction method and system Technical Field The invention relates to the technical fields of intelligent traffic, machine vision and artificial intelligence, in particular to a ToF imaging data error correction technology for traffic detection in the field of large traffic (such as airports, subways and the like). Background A ToF (Time-of-Flight) sensor is a commonly used three-dimensional imaging device capable of rapidly acquiring depth information of a scene. However, in practical applications, toF imaging may generate reflective cavities and overexposure, which may result in inaccurate imaging data, affecting subsequent target detection and identification. The specific problems are described as follows: The reflective cavity phenomenon is defined as that a ToF camera cannot receive enough reflected light signals, so that a depth value cannot be calculated, and a region with a null or invalid depth value appears in a depth map. Common reasons are: Low reflectivity objects such as black objects, light absorbing materials, etc., the reflected light intensity is too low to be effectively detected by the camera. An occluded or shadowed area where light cannot reach or reflected light is occluded, resulting in the camera not receiving a valid signal. Beyond the measuring range, the target object is too far from the camera, beyond the effective measuring range of the camera, the reflected light signal is too weak. Multipath interference, namely, reflected light reaches a camera after multiple reflections, so that signal confusion is caused, and the depth cannot be accurately calculated. Depth value range in ToF imaging, reflective cavities are typically represented by a specific invalid value, a common invalid depth value may be 0 (representing no depth information) or other special values defined by the camera system, e.g. some cameras may represent an invalid depth value by 65535 (maximum value of unsigned 16 bit integer), but not all cameras use this value, depending on the hardware and software design of the camera. And (II) overexposure is defined as that the reflected light signal received by the ToF camera is too strong and exceeds the dynamic range of a camera sensor, so that the depth calculation is wrong or the depth value is abnormal. Common reasons are: high reflectivity objects such as white objects, specular objects, etc., reflected light intensities that are too high, beyond the processing power of the camera sensor. Close-range objects, objects too close to the camera, reflected light intensities too high, which can easily lead to overexposure. Ambient light interference-strong ambient light (e.g., direct sunlight) may enhance reflected light intensity, resulting in overexposure. Depth value range the overexposed depth value is typically an outlier, the specific value depending on the way the camera is handled. Some cameras may set the depth value of the overexposed region to 0, indicating an invalid depth, and some may give erroneous depth values, which may be much lower than the actual depth, e.g. negative or very small positive values. Disclosure of Invention The invention provides a method and a system for filling and overexposure correction of a reflective cavity based on ToF imaging, which are used for filling or correcting depth pixel values by analyzing boundary continuity of a reflective abnormal region, so that the problems of the reflective cavity and overexposure in a ToF 3D point cloud picture are solved. The specific technical scheme is as follows: Reflective cavity filling module 1. Abnormal region detection, detecting regions of depth value 0 or other special values defined by the camera system (e.g., 65535), which typically correspond to reflective cavities. 2. Filling algorithm: filling based on neighborhood information, namely filling depth values of the reflection cavity area by analyzing normal reflection data around the abnormal area and adopting a linear interpolation or texture-based restoration algorithm. Scaling the abnormal region by calculating the scaling relation of the surrounding normal reflection data with respect to the partial reflection cavity, thereby generating a corrected depth pixel value. (II) overexposure correction module 1. Abnormal region detection, detecting regions of depth value 0 or abnormally small values (e.g., 10, 20, etc.), which generally correspond to overexposure. 2. Correction algorithm: and (3) analyzing the boundary continuity of the reflection abnormal region, and supplementing and correcting the depth pixel value. And (3) repairing the depth image, namely continuously filling the corrected depth pixel values in the boundary of the abnormal reflection area in the depth image, so as to ensure the integrity and the accuracy of the depth data. (III) depth data preprocessing module 1. And filtering, namely filtering the original depth data, removing noise and smoothing the dept