CN-121982064-A - Road element tracking method, apparatus, device, storage medium, and program product

CN121982064ACN 121982064 ACN121982064 ACN 121982064ACN-121982064-A

Abstract

The embodiment of the application provides a road element tracking method, a device, equipment, a storage medium and a program product, and relates to the technical field of automatic driving. The method comprises the steps of obtaining frame image data and inertial navigation data of a target to be detected and a target identification result of the target to be detected, which is determined according to the frame image data, constructing a camera motion model according to the inertial navigation data, carrying out motion prediction processing on the target to be detected according to the camera motion model to obtain a motion prediction result of the target to be detected, and carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result to obtain a position tracking result of the target to be detected. The method can realize real-time, stable and accurate tracking of the road elements on low-power equipment.

Inventors

ZHANG CHAO
GE JUNXIA
GUO ZHAOZHONG
ZHANG WEICHAO

Assignees

西安四维图新信息技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. A method of tracking road elements, comprising: Acquiring frame image data and inertial navigation data of a target to be detected, and determining a target identification result of the target to be detected according to the frame image data; constructing a camera motion model according to the inertial navigation data, and performing motion prediction processing on the target to be detected according to the camera motion model to obtain a motion prediction result of the target to be detected; and carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result to obtain a position tracking result of the target to be detected.
2. The method of claim 1, wherein prior to performing optimal position estimation on the object to be detected based on the motion prediction result and the object recognition result, the method further comprises: performing association processing on the motion prediction result and the target recognition result; If the association is successful, carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result; If the correlation fails, updating the target recognition result according to the historical recognition result and the motion prediction result corresponding to the previous frame of image data to obtain an updated target recognition result, and carrying out optimal position estimation on the target to be detected according to the updated target recognition result and the motion prediction result.
3. The method according to claim 2, wherein performing optimal position estimation on the object to be detected according to the motion prediction result and the object recognition result to obtain a position tracking result of the object to be detected, includes: and taking the motion prediction result as a prediction value, taking the target identification result as an observation value, and carrying out Kalman filtering processing on the target to be detected so as to carry out optimal position estimation on the target to be detected, thereby obtaining a position tracking result of the target to be detected.
4. The method according to claim 2, wherein updating the target recognition result according to the history recognition result and the motion prediction result corresponding to the previous frame of image data to obtain an updated target recognition result comprises: determining a template matching range according to the motion prediction result; Performing template matching processing in the template matching range according to the historical identification result corresponding to the previous frame of image data to obtain a new target detection result; And updating the target recognition result according to the new target detection result to obtain an updated target recognition result.
5. The method of claim 4, wherein performing optimal position estimation on the object to be detected based on the updated object recognition result and the motion prediction result comprises: Determining a target fusion weight according to the first confidence coefficient corresponding to the updated target recognition result and the second confidence coefficient corresponding to the motion prediction result, wherein the target fusion weight comprises the first weight corresponding to the updated target recognition result and the second weight corresponding to the motion prediction result; according to the first weight and the second weight, carrying out weighted fusion processing on the updated target identification result and the motion prediction result to obtain a fusion detection result; and taking the fusion detection result as a predicted value, taking the updated target identification result as an observed value, and carrying out Kalman filtering processing on the target to be detected so as to carry out optimal position estimation on the target to be detected, thereby obtaining a position tracking result of the target to be detected.
6. The method of any of claims 1-5, wherein the camera motion model includes a camera depth therein, the method further comprising: And updating the camera depth included in the camera motion model according to the position tracking result of the target to be detected and the position data of the last frame of the target to be detected, so as to obtain an updated camera motion model.
7. The method of claim 3 or 5, wherein the inertial navigation data comprises heading angle data, the method further comprising: Acquiring camera parameter information; Initializing a Kalman filter according to the camera parameter information and the course angle data to obtain an initialized Kalman filter; And carrying out Kalman filtering processing on the target to be detected according to the initialized Kalman filter so as to carry out optimal position estimation on the target to be detected and obtain a position tracking result of the target to be detected.
8. A road element tracking device, comprising: The data acquisition unit is used for acquiring frame image data and inertial navigation data of a target to be detected and a target identification result of the target to be detected, which is determined according to the frame image data; the motion prediction unit is used for constructing a camera motion model according to the inertial navigation data, and performing motion prediction processing on the target to be detected according to the camera motion model to obtain a motion prediction result of the target to be detected; And the position tracking unit is used for carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result to obtain a position tracking result of the target to be detected.
9. An electronic device is characterized by comprising a memory and a processor; the memory stores computer-executable instructions; The processor executing computer-executable instructions stored in the memory, causing the processor to perform the method of any one of claims 1-7.
10. A computer readable storage medium/computer program product, characterized in that the computer readable storage medium has stored therein computer executable instructions for implementing the method according to any of claims 1-7 when executed by a processor; computer program product comprising a computer program which, when executed by a processor, implements the method of any of claims 1-7.

Description

Road element tracking method, apparatus, device, storage medium, and program product Technical Field The present application relates to the field of automatic driving technology, and in particular, to a method, apparatus, device, storage medium and program product for tracking road elements. Background In an autopilot system, real-time, stable tracking of road elements (e.g., traffic signs, traffic lights, shafts, and delineators, etc.) is a key link to achieve path planning, obstacle avoidance, and driving decisions. In the related art, the tracking of the road elements is realized mainly based on a visual tracking method of the feature points. Specifically, feature points of target contour screening are used, contour prediction is carried out on the traffic plate of the next frame through an optical flow tracking module, and target contour parameters of the traffic plate in the next frame are obtained. In addition, the visual tracking method needs larger calculation force requirements and cannot be applied to low calculation force equipment, and on the other hand, complex scenes such as brightness change, rapid movement, shielding, repeated textures and the like are difficult to adapt, so that the accuracy and reliability of the road element tracking are affected. Disclosure of Invention The embodiment of the application provides a road element tracking method, a device, equipment, a storage medium and a program product, which can realize real-time, stable and accurate tracking of road elements on low-power equipment. In a first aspect, an embodiment of the present application provides a road element tracking method, including: Acquiring frame image data and inertial navigation data of a target to be detected, and determining a target identification result of the target to be detected according to the frame image data; constructing a camera motion model according to the inertial navigation data, and performing motion prediction processing on the target to be detected according to the camera motion model to obtain a motion prediction result of the target to be detected; and carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result to obtain a position tracking result of the target to be detected. In a possible implementation manner, before performing optimal position estimation on the object to be detected according to the motion prediction result and the object identification result, the method further includes: performing association processing on the motion prediction result and the target recognition result; If the association is successful, carrying out optimal position estimation on the target to be detected according to the motion prediction result and the target identification result; If the correlation fails, updating the target recognition result according to the historical recognition result and the motion prediction result corresponding to the previous frame of image data to obtain an updated target recognition result, and carrying out optimal position estimation on the target to be detected according to the updated target recognition result and the motion prediction result. In a possible implementation manner, according to the motion prediction result and the target identification result, performing optimal position estimation on the target to be detected to obtain a position tracking result of the target to be detected, including: and taking the motion prediction result as a prediction value, taking the target identification result as an observation value, and carrying out Kalman filtering processing on the target to be detected so as to carry out optimal position estimation on the target to be detected, thereby obtaining a position tracking result of the target to be detected. In one possible implementation manner, according to the historical recognition result and the motion prediction result corresponding to the previous frame of image data, updating the target recognition result to obtain an updated target recognition result includes: determining a template matching range according to the motion prediction result; Performing template matching processing in the template matching range according to the historical identification result corresponding to the previous frame of image data to obtain a new target detection result; And updating the target recognition result according to the new target detection result to obtain an updated target recognition result. In a possible implementation manner, performing optimal position estimation on the object to be detected according to the updated object identification result and the motion prediction result includes: Determining a target fusion weight according to the first confidence coefficient corresponding to the updated target recognition result and the second confidence coefficient corresponding to the motion prediction result, wherein the target fusion wei