Method and device for determining pose of tracked object in image tracking process

Abstract

The invention provides a method and device for determining the pose of a tracking object in the image tracking process. The method comprises the steps that a current frame image and first point cloudinformation corresponding to the current frame image are acquired, and the first point cloud information is based on the relative relation between camera equipment and an environment space; and when the tracking object is in the static state, the pose of the tracking object in the current frame image relative to the camera device is determined according to the first point cloud information and theobject feature information of the tracking object, and the object feature information is based on the relative relationship between the tracking object and the environment space. According to the method, the relative relationship of the tracking object is converted by using the first point cloud information, so that the pose of the tracking object is convenient to determine.

Description

technical field [0001] The present disclosure relates to the technical field of image processing, and in particular, to a method and a device for determining a pose of a tracking object during an image tracking process. Background technique [0002] Image tracking technology is to locate the objects captured by the camera through image recognition, infrared, ultrasonic and other methods. Image tracking technology is widely used in the field of computer vision, such as virtual reality (VR, Virtual Reality), augmented reality (AR, Augmented Reality) and so on. [0003] At present, image tracking technology is combined with deep learning to realize image tracking, in which online learning and neural network models are mainly used, both online learning and neural network models are applied to image tracking after training. In the image tracking process, each frame of image is input to the trained online learning model or neural network model, and then the online learning model ...

AI Technical Summary

Problems solved by technology

[0004] However, there are some problems in image tracking using deep learning. For example, the deep learning model applied to image tracking needs to be trained, and model training requires a large number of training samples. It is currently difficult to provide a large number of training samples, especially online learning. Model

In addition, due to the convolution pooling effect of the neural network model, the position information output by the last layer is lost, and the position information is mainly the position information.

Images