Moving object tracking method and apparatus

1. A method for tracking an object, comprising:

obtaining a current frame captured by a camera, wherein the camera is located in a vehicle;

determining a current pose of the camera based on an inertial measurement unit IMU on the vehicle and the current frame;

determining a current pose of a target object in the current frame, wherein the target object is located around the vehicle; and

optimizing the current pose of the camera and the current pose of the target object to obtain a second pose of the camera and a second pose of the target object, wherein the second pose of the camera and the second pose of the target object are used to track the target object; and

wherein the determining a current pose of a target object in the current frame comprises:

determining a dynamic feature point in the current frame, wherein the dynamic feature point comprises a representation point, and the representation point is used as an origin of a coordinate system of the target object; and

predicting the current pose of the target object in the current frame based on an attitude of the representation point and a random motion model, to obtain the current pose of the target object.

2. The method according to claim 1 , wherein the method further comprises:

calculating a Jacobian matrix of an image feature of the current frame based on an image feature of a background part in the current frame and an image feature of the target object; and

updating the current pose of the camera and the current pose of the target object based on a measurement result of the image feature in the current frame comprises:

updating the current pose of the target object and the current pose of the camera based on the Jacobian matrix of the image feature of the current frame.

3. The method according to claim 1 , wherein the dynamic feature point further comprises a non-representation point, and the non-representation point is used to represent another dynamic feature point other than the representation point on the target object; and

the method further comprises:

when the representation point is blocked, calculating the attitude of the representation point by using an attitude of the non-representation point.

4. The method according to claim 1 , wherein the determining a dynamic feature point in the current frame comprises:

framing a target object that is moving in the current frame by using a bounding box; and

determining a feature point in the bounding box as the dynamic feature point.

5. The method according to claim 1 , wherein the method further comprises:

adding the current pose of the target object to a sliding window, to generate a Kalman state vector, wherein the sliding window comprises the current pose of the camera.

6. The method according to claim 1 , wherein the method further comprises:

if the target object in the moving state is not observed in a corresponding frame within a preset time, deleting a predicted value of a current state vector of the target object from the Kalman state vector.

7. An apparatus for tracking an object, comprising:

a memory configured to store instructions for a computer program;

a processor coupled to receive and execute the instructions which cause the apparatus to track a moving object by performing the steps of:

obtaining a current frame captured by a camera disposed in a vehicle;

determining a current pose of the camera based on an inertial measurement unit IMU on the vehicle and the current frame;

determining a current pose of a target object in the current frame, wherein the target object is located around the vehicle; and

optimizing the current pose of the camera and the current pose of the target object to obtain a second pose of the camera and a second pose of the target object, wherein the second pose of the camera and the second pose of the target object are used to track the target object; and

wherein the determining a current pose of a target object in the current frame comprises:

determining a dynamic feature point in the current frame, wherein the dynamic feature point comprises a representation point, and the representation point is used as an origin of a coordinate system of the target object; and

predicting the current pose of the target object in the current frame based on an attitude of the representation point and a random motion model, to obtain the current pose of the target object.

8. The apparatus according to claim 7 further configured to receive and execute the instructions which cause the apparatus to track the moving object by performing the steps of:

calculating a Jacobian matrix of an image feature of the current frame based on an image feature of a background part in the current frame and an image feature of the target object; and

updating the current pose of the camera and the current pose of the target object based on a measurement result of the image feature in the current frame comprises:

updating the current pose of the target object and the current pose of the camera based on the Jacobian matrix of the image feature of the current frame.

9. The apparatus according to claim 7 , wherein the dynamic feature point further comprises a non-representation point, and the non-representation point is used to represent another dynamic feature point other than the representation point on the target object; and

the method further comprises:

when the representation point is blocked, calculating the attitude of the representation point by using an attitude of the non-representation point.

10. The apparatus according to claim 7 , wherein the determining a dynamic feature point in the current frame comprises:

framing a target object that is moving in the current frame by using a bounding box; and

determining a feature point in the bounding box as the dynamic feature point.

11. The apparatus according to claim 7 , wherein the method further comprises:

adding the current pose of the target object to a sliding window, to generate a Kalman state vector, wherein the sliding window comprises the current pose of the camera.

12. The apparatus according to claim 7 further configured to receive and execute the instructions which cause the apparatus to track the moving object by performing the step of, if the target object in the moving state is not observed in a corresponding frame within a preset time, deleting a predicted value of a current state vector of the target object from the Kalman state vector.

13. A apparatus for tracking a moving object, comprising:

an input/output interface,

a processor, and

a memory, wherein the memory is configured to store program instructions, and the processor is configured to execute the program instructions from the memory to perform the steps of:

obtaining a current frame captured by a camera, wherein the camera is located in a vehicle;

determining a current pose of the camera based on an inertial measurement unit IMU on the vehicle and the current frame;

determining a current pose of a target object in the current frame, wherein the target object is located around the vehicle; and

optimizing the current pose of the camera and the current pose of the target object to obtain a second pose of the camera and a second pose of the target object, wherein the second pose of the camera and the second pose of the target object are used to track the target object; and

wherein the determining a current pose of a target object in the current frame comprises:

determining a dynamic feature point in the current frame, wherein the dynamic feature point comprises a representation point, and the representation point is used as an origin of a coordinate system of the target object; and

predicting the current pose of the target object in the current frame based on an attitude of the representation point and a random motion model, to obtain the current pose of the target object.

14. A nontransitory computer-readable storage medium, comprising a computer program, wherein when the computer program executed by a processor in an apparatus, causes the apparatus to perform the steps of:

obtaining a current frame captured by a camera, wherein the camera is located in a vehicle;

determining a current pose of the camera based on an inertial measurement unit IMU on the vehicle and the current frame;

determining a current pose of a target object in the current frame, wherein the target object is located around the vehicle; and

optimizing the current pose of the camera and the current pose of the target object to obtain a second pose of the camera and a second pose of the target object, wherein the second pose of the camera and the second pose of the target object are used to track the target object; and

wherein the determining a current pose of a target object in the current frame comprises:

determining a dynamic feature point in the current frame, wherein the dynamic feature point comprises a representation point, and the representation point is used as an origin of a coordinate system of the target object; and

predicting the current pose of the target object in the current frame based on an attitude of the representation point and a random motion model, to obtain the current pose of the target object.