Search

CN-121999055-A - Road monitoring camera calibration recovery method and device based on image feature matching

CN121999055ACN 121999055 ACN121999055 ACN 121999055ACN-121999055-A

Abstract

The invention relates to a road monitoring camera calibration recovery method and device based on image feature matching, wherein the method comprises the steps of obtaining a current image; extracting a plurality of feature points and descriptors of each feature point in a current image, matching the feature points of the current image with reference images according to the feature points and the descriptors of each feature point of the current image to obtain matching point pairs, determining world coordinates corresponding to each matching point pair according to all the matching point pairs and a pre-constructed static scene feature library, estimating target external parameters of the current camera according to the world coordinates corresponding to each matching point pair, and updating a projection matrix of the image to the world coordinates according to the target external parameters so as to realize measurement and analysis of target positions in a monitored scene according to the updated projection matrix. By the method, unmanned calibration maintenance of the road monitoring camera is realized, the capability of high robustness and adaptation to complex environments is achieved, and the labor cost is remarkably reduced.

Inventors

  • CHEN ZHIQIANG
  • YAN JINYU
  • WU KEWEI
  • HE XIAOGANG

Assignees

  • 北京卓视智通科技有限责任公司

Dates

Publication Date
20260508
Application Date
20251219

Claims (10)

  1. 1. The road monitoring camera calibration recovery method based on image feature matching is characterized by comprising the following steps of: acquiring a current image from a monitoring video stream acquired by a current camera; extracting a plurality of feature points and descriptors of each feature point in the current image; According to the multiple feature points of the current image and the descriptors of each feature point, performing feature point matching on the current image and a reference image to obtain matching point pairs; determining world coordinates corresponding to each matching point pair according to all the matching point pairs and a pre-constructed static scene feature library; estimating a target external parameter of the current camera according to the world coordinates corresponding to each matching point pair; And updating a projection matrix from the image to world coordinates according to the target external parameters so as to realize measurement and analysis of the target position in the monitored scene according to the updated projection matrix.
  2. 2. The method according to claim 1, wherein the method further comprises: extracting a plurality of feature points from a reference image and describing a descriptor of each feature point; Determining world coordinates corresponding to each feature point in the reference image based on a mapping relation between pre-constructed image pixel coordinates and world coordinates; and storing all the feature points of the reference image and the corresponding world coordinates to form a static scene feature library.
  3. 3. The method according to claim 1, wherein the method further comprises: Detecting a dynamic target in the current image; shielding the dynamic target from the current image to obtain a target image; the extracting a plurality of feature points and descriptors of each feature point in the current image comprises: extracting a plurality of feature points and descriptors of each feature point in the target image.
  4. 4. A method according to any one of claims 1 to 3, wherein said estimating the target outliers of the current camera from the corresponding world coordinates of each of the matching pairs of points comprises: s1, randomly selecting a set number of point pairs from all the matched point pairs; S2, performing external parameter estimation according to the set number of point pairs to obtain a preliminary external parameter; s3, calculating to obtain the re-projection error of each point according to all the matching point pairs and the preliminary external parameters; s4, screening out interior points conforming to the model according to all the reprojection errors; s5, based on the inner points of the coincidence model, determining new primary outer parameters, repeating S3 to S5, and determining the corresponding primary outer parameters when the number of the inner points of the coincidence model is maximized as target outer parameters.
  5. 5. A method according to any one of claims 1 to 3, wherein the target outliers comprise a rotation matrix and a translation vector, and wherein updating the projection matrix of the image to world coordinates based on the target outliers comprises: and updating a projection matrix of the image to world coordinates according to the rotation matrix, the translation vector and the internal parameters of the current camera.
  6. 6. The utility model provides a road monitoring camera calibration recovery unit based on image feature matches which characterized in that includes: the acquisition module is used for acquiring a current image from the monitoring video stream acquired by the current camera; the feature extraction module is used for extracting a plurality of feature points in the current image and descriptors of each feature point; the matching module is used for matching the characteristic points of the current image with the reference image according to the plurality of characteristic points of the current image and the descriptors of each characteristic point to obtain matching point pairs; the world coordinate determining module is used for determining world coordinates corresponding to each matching point pair according to all the matching point pairs and the pre-constructed static scene feature library; the target external parameter determining module is used for estimating the target external parameter of the current camera according to the world coordinates corresponding to each matching point pair; And the updating module is used for updating the projection matrix from the image to the world coordinates according to the target external parameters so as to realize the measurement and analysis of the target position in the monitored scene according to the updated projection matrix.
  7. 7. The apparatus of claim 6, wherein the apparatus further comprises: The static scene feature library construction module is used for extracting a plurality of feature points and descriptors of the feature points from a reference image, determining world coordinates corresponding to the feature points in the reference image based on a mapping relation between pre-constructed image pixel coordinates and world coordinates, and storing all the feature points and the corresponding world coordinates of the reference image to form the static scene feature library.
  8. 8. The apparatus of claim 6, wherein the apparatus further comprises: the preprocessing module is used for detecting a dynamic target in the current image, shielding the dynamic target from the current image and obtaining a target image; The feature extraction module is specifically configured to, when extracting a plurality of feature points and a description of each feature point in the current image: extracting a plurality of feature points and descriptors of each feature point in the target image.
  9. 9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1-5 when the computer program is executed.
  10. 10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the method of any of claims 1-5.

Description

Road monitoring camera calibration recovery method and device based on image feature matching Technical Field The invention relates to the technical fields of traffic video analysis, urban Intelligent Transportation (ITS) systems and computer vision and multi-vision geometry, in particular to a road monitoring camera calibration recovery method and device based on image feature matching. Background The fixed monocular monitoring camera in the traffic road scene is required to complete the mapping relation calibration between the pixel coordinates and the space meter coordinates in the mode of manual target distribution or manual point selection in the initial stage of engineering deployment. However, in the long-term outdoor operation process, the camera often causes fine position offset due to uncontrollable factors such as strong wind, rod looseness, construction maintenance, vehicle collision, vibration and the like, so that the original calibration parameters are invalid. The traditional recalibration mode needs manual arrival, target arrangement, road sealing operation and repeated measurement, and huge labor cost, traffic interference risk and unstable precision are brought. Specifically, the prior art mainly includes: Manual target distribution calibration, which depends on targets such as checkerboard, chArUco and the like, requires professional field operation and cannot cope with frequent camera offset. Based on the calibration of the lane lines, the camera pose is estimated by using the geometric constraint of the lane lines, but is limited by the visibility, shielding and abrasion of the lane lines. Based on the calibration of the vehicle size, the camera height and angle are reversely pushed by the vehicle size priori, but the camera is sensitive to the vehicle type, the vehicle speed and the shielding, and the accuracy is limited. Aiming at the prior art, the prior art has the following technical problems: the method is highly dependent on manual work, and has high target arrangement, road sealing shooting and safety risk. The drift monitoring mechanism is absent, the calibration quality cannot be estimated in real time, and the recovery flow is difficult to trigger in time. And the dynamic interference is serious, namely the interference characteristics of the dynamic targets such as vehicles, pedestrians and the like in the traffic scene are matched and the pose is estimated. And the version management is lacking, and the calibration parameters are difficult to trace back, roll back and multiplex. The system is difficult to maintain for a long time, lacks a closed loop optimization mechanism, and gradually reduces the calibration precision along with the environmental change. Disclosure of Invention The invention aims to solve at least one technical problem by providing a road monitoring camera calibration recovery method and device based on image feature matching. In a first aspect, the technical scheme for solving the technical problems is as follows, the method for calibrating and recovering the road monitoring camera based on image feature matching comprises the following steps: acquiring a current image from a monitoring video stream acquired by a current camera; extracting a plurality of feature points and descriptors of each feature point in the current image; According to a plurality of feature points of the current image and descriptors of each feature point, matching the feature points of the current image with the reference image to obtain matching point pairs; determining world coordinates corresponding to each matching point pair according to all the matching point pairs and a pre-constructed static scene feature library; estimating a target external parameter of the current camera according to the world coordinates corresponding to each matching point pair; And updating a projection matrix from the image to world coordinates according to the external parameters of the target, so as to realize measurement and analysis of the target position in the monitored scene according to the updated projection matrix. The method has the advantages that the matching of the characteristic points of the current image and the reference image is realized by acquiring the current image from the monitoring video stream and extracting the characteristic points and the descriptors thereof, so that the world coordinates corresponding to the matching point pairs can be accurately determined, and a reliable basis is provided for the subsequent external parameter estimation. The object external parameters of the current camera, which are obtained based on the world coordinate estimation of the matching point pairs, can accurately reflect the gesture change of the camera in the monitored scene. And further updating a projection matrix from the image to world coordinates according to the target external parameters, so that the monitoring system can accurately measure and analyze the target position in the