Search

EP-3631362-B1 - METHOD AND APPARATUS FOR CONSTRUCTING AN ENVIRONMENT MODEL

EP3631362B1EP 3631362 B1EP3631362 B1EP 3631362B1EP-3631362-B1

Inventors

  • DOEMLING, Maximilian
  • JIANG, WANLI
  • GRANZOW, SEBASTIAN
  • LI, Jianpeng
  • XU, TAO
  • XU, Hongshan
  • LV, Shuhan

Dates

Publication Date
20260506
Application Date
20170531

Claims (18)

  1. A computer-implemented method for constructing an environment model, comprising: - generating (120) a first Signatured Gaussian Mixture, SGM, model corresponding to a first part of the environment based on a first sensor data of a first movable object (10); - receiving (140), from a second movable object (20, 30), a second SGM model corresponding to a second part of the environment, wherein the second SGM model is generated based on a second sensor data of the second movable object (20, 30); and - constructing (160) a third SGM model comprising the first SGM model and the second SGM model, wherein the step "generating (120) a first Signatured Gaussian Mixture Model, SGMM, corresponding to a first part of the environment based on a first sensor data of a first movable object (10)" comprises: - obtaining (122) the first sensor data representing a first part of the environment; - generating (124) a Gaussian Mixture model for the first part of the environment based on the obtained first sensor data; - generating (126) at least one signature for identifying elements in the environment, wherein the signature comprises properties of the elements; and - generating (128) the first SGM model for representing the environment, wherein the first SGM Model comprises the Gaussian Mixture model of the first part of the environment and the signature.
  2. Method according to one of claims 1, wherein the first sensor data is received from: - an image recording device, especially video camera and stereo camera; and/or - a laser scanner; and/or - a radar sensor.
  3. Method according to one of claims 1 - 2, wherein the method further comprises: - obtaining (130) localization information of the first SGM model of the first movable object (10).
  4. Method according to claim 3, wherein the step "receiving (140), from a second movable object (20, 30), a second SGM model corresponding to a second part of the environment" comprises: - obtaining (144) localization information of the second SGM model from the second movable object (20, 30); and - receiving (146) the second SGM model from the second movable object (20, 30).
  5. Method according to claim 4, wherein the step "receiving (140), from a second movable object (20, 30), a second SGM model corresponding to a second part of the environment" before the step of "obtaining (144) localization information of the second SGM model from the second movable object (20, 30)" further comprises: - selecting (142) the second movable object (20, 30) according to the localization information of the first SGM model and the localization information of the second SGM model.
  6. Method according to any one of claims 1 - 5, wherein the second part of the environment corresponds to the environment of the second movable object (20, 30).
  7. Method according to any one of claims 1 - 6, wherein the second SGM model is generated based on the second sensor data from: - a further image recording device, especially video camera and stereo camera; and/or - a further laser scanner; and/or - a further radar sensor.
  8. Method according to any one of claims 1 - 7, wherein the method further comprises: - receiving (150) localization information of the second SGM model from the second movable object (20, 30).
  9. Method according to claim 8, wherein the step "constructing (160) a third SGM model comprising the first SGM model and the second SGM model" comprises: - combining the first SGM model and the second SGM model by mapping the first SGM model and the second SGM model according to the localization information of the first SGM model and the localization information of the second SGM model.
  10. Data processing device for constructing an environment model comprising: - a generating module (220) configured to generate a first Signatured Gaussian Mixture, SGM, model corresponding to a first part of the environment based on a first sensor data of a first movable object (10); - a receiving module (240) configured to receive, from a second movable object (20, 30), a second SGM model corresponding to a second part of the environment, wherein the second SGM model is generated based on a second sensor data of the second movable object (20, 30); and - a constructing module (260) configured to construct a third SGM model comprising the first SGM model and the second SGM model, wherein the generating module (240) is configured: - to obtain the first sensor data representing a first part of the environment; - to generate a Gaussian Mixture model for the first part of the environment based on the obtained first sensor data; - to generate at least one signature for identifying elements in the environment, wherein the signature comprises properties of the elements; and - to generate the first SGM model for representing the environment, wherein the first SGM Model comprises the Gaussian Mixture model of the first part of the environment and the signature.
  11. Data processing device according to claim 10, wherein the data processing device is configured to receive the first sensor data from: - an image recording device, especially video camera and stereo camera; and/or - a laser scanner; and/or - a radar sensor.
  12. Data processing device according to one of claims 10 - 11, wherein the data processing device further comprises a first localization module (230) configured to obtain localization information of the first SGM model of a first movable object (10).
  13. Data processing device according to claim 12, wherein the receiving module (240) is configured: - to obtain localization information of the second SGM model from the second movable object (20, 30); and - to receive the second SGM model from the second movable object (20, 30).
  14. Data processing device according to claim 13, wherein the receiving module (240) is further configured to select the second movable object (20, 30) according to the localization information of the first SGM model and the localization information of the second SGM model.
  15. Data processing device according to any one of claims 10 - 14, wherein the data processing device further comprises a second localization module (250) configured to receive localization information of the second SGM model from the second movable object (20, 30).
  16. Data processing device according to any one of claims 10 - 15, wherein the constructing module (260) configured to combine the first SGM model and the second SGM model by mapping the first SGM model and the second SGM model according to the localization information of the first SGM model and the localization information of the second SGM model.
  17. System for constructing an environment model comprising a data processing device according to any one of claims 10 - 16 and at least one of: - an image recording device, especially video camera and stereo camera; - a laser scanner; and - a radar sensor.
  18. Vehicle or robot comprising a system according to claim 17.

Description

FIELD OF THE INVENTION The present invention relates in general to the field of vehicle or robot, and in more particular, to a method and an apparatus for constructing the environment model for vehicle or robot. BACKGROUND The autonomous driving technology has been researched for many years, and many of the proposed benefits have been demonstrated in varied applications. Many categories of maps have been developed and used in vehicle/robot, such as a point map which consists of laser points, a grid map which separates the environment into a grid with each grid cell recording whether it is occupied by something as well as the probability of the occupancy, a geometric primitive map which uses one or more types of geometric primitives to represent entities in the environment, and a feature map which mainly consists of feature points and their corresponding descriptors extracted from other types of data (e.g., a point cloud, a camera image, etc.), and the Normal distribution transform (NDT) map which uses uni-weighted Gaussian Mixture Model to represent the environment, with each Gaussian distribution modeling a unique grid cell of the environment, a Normal distribution transform Occupancy (NDT-OM) map which separates the environment into grid, within each grid cell of which a Gaussian distribution is calculated among the data points in the cell and a weight which represents the occupancy probability of this cell is maintained for the Gaussian distribution. However, according to the existing technology, a vehicle/robot may only use its own sensor data (received from sensors mounted on the vehicle/robot) to construct the environment model, which is limited to the surrounding region of the vehicle/robot. An example of a vehicle/ robot constructing Signatured Gaussian Mixture Models for map elements within a real-time point cloud or image acquired by the vehicle/robot is provided in WO 2016/201670 A1. The task of the present invention is to provide a method and a device that can enlarge the range of the environmental model. The above mentioned task is solved by claim 1, as well as claim 10. Advantageous features are also defined in dependent claims. SUMMARY Embodiments of the present invention provide a method, a device, a system and a vehicle for constructing an environment model, which enable a wide range of the environmental model constructed according to sensor data. Accordingly, a computer-implemented method for constructing an environment model is provided. The method comprises: generating a first Signatured Gaussian Mixture, SGM, model corresponding to a first part of the environment based on a first sensor data; receiving a second SGM model corresponding to a second part of the environment; and constructing a third SGM model comprising the first SGM model and the second SGM model. The step "generating a first Signatured Gaussian Mixture Model, SGMM, corresponding to a first part of the environment based on a first sensor data" comprises: obtaining the first sensor data representing a first part of the environment; generating a Gaussian Mixture model for the first part of the environment based on the received sensor data; generating at least one signature for identifying elements in the environment, wherein the signature comprises properties of the elements; and generating the first SGM model for representing the environment, wherein the SGM Model comprises the Gaussian Mixture model of the first part of the environment and the signature. In a further possible implementation manner, the sensor data is received from: an image recording device, especially video camera and stereo camera; and/or a laser scanner; and/or a radar sensor. In another further possible implementation manner, the method further comprises: obtaining localization information of the first SGM model. In another further possible implementation manner, wherein the step "receiving a second SGM model corresponding to a second part of the environment" comprises: obtaining localization information of the second SGM model from the second movable object; and receiving the second SGM model from the second movable object. In another further possible implementation manner, wherein the step "receiving a second SGM model corresponding to a second part of the environment" further comprises: selecting the second movable object according to the localization information of the first SGM model and the localization information of the second SGM model. In another further possible implementation manner, the second part of the environment corresponds to the environment of the second movable object. In another further possible implementation manner, the second SGM model is generated based on a second sensor data from: a further image recording device, especially video camera and stereo camera; and/or a further laser scanner; and/or a further radar sensor. In another further possible implementation manner, the method further comprises: receiving localization info