Pose optimisation, mapping, and localisation techniques

摘要

A method of optimising or updating a pose comprising position and orientation of a sensor. A model of an object is received. Sensor data relating to part of the object is received from the sensor, where the sensor data is in the frame of reference of the sensor and associated with pose data representing a pose of the sensor. Part of the model is projected onto the sensor frame of reference. Corresponding feature points or lines between the sensor image and the model are identified. The pose of the sensor is optimised using the location difference between the corresponding feature points or lines. Optionally, a parameter of the model is also optimised updated, wherein the parameter defines a configuration of the object. Optionally, the sensor data is input to a convolutional neural network which generates a skeletal representation of the object form the sensor data.