Übersetzung für "Pose estimation" in Deutsch

In this context, the invention in particular relates to improving pose estimation.

This step is also called pose estimation.

Another alternative for pose estimation is based on time-of-flight (TOF) cameras.

Estimation of the pose of each tag (see Pose estimation).

Highly motivated, we accepted the challenge and applied with our technology PEAR (Comprehensive Pose Estimation for AR on Mobile Devices) in the category “Mobile Application”.

If multiple tags with the same ID are visible in the left or right image, pose estimation may compute a wrong pose if these tags have the same orientation and are approximately aligned in parallel to the image rows.

To ensure good pose-estimation quality, the frame rate should not drop significantly under 10 Hz.

We find applications for the inverse compositional algorithm in the pose estimation of unmanned aerial vehicles, so called UAV's.

Tags might be omitted from the detect response due to several reasons, e.g., if a tag is visible in only one of the cameras or if pose estimation did not succeed.

Have a look at our demo systems from Automatica 2016 including explanations of hyperspectral imaging and our new software tool for pose estimation called CVB Polimago.

In this case, the skeleton model preferably has a simple design and represents only those anatomical conditions of the human body, which are relevant to pose estimation.

However, as conventional video sequences have a frame rate of 25 Hz, a real-time-capable pose estimation at an appropriate frame rate would be desirable.

It is, therefore, the problem of the invention to provide a method for pose estimation by means of a TOF camera, which allows a continuous image analysis and pose calculation at at least 25 frames per second.

The advantage of the high speed of image segmentation by the TOF camera and the rapidly calculable SOM-updating rule allow a reliable pose estimation at up to 25 frames per second on a 2.0 GHz PC with only a minimum programming and computing expenditure.

Actually, the use of this very simple pose estimation requires that the person always holds his or her right arm clearly away from the body if he or she wants to command the machine by moving that hand, for example.

The skeleton model outlined in WO 2010/130245 A1 exhibits relatively many nodes, whose number cannot be readily reduced to 20 or less without accepting considerable errors in the pose estimation.

An aspect of the invention is thus information compression that usually takes place in two part steps: image segmentation and pose estimation.

According to the invention, pose estimation by means of SOM training is proposed that is able to work with a skeleton model that is modelled according to the anatomy of the object observed and exhibits a reduced node count, it being possible to reliably and consistently assign each model node to a predetermined body part.

Applying the learning rule for the SOM from WO 2010/130245 A1 can no longer achieve good fitting of the model into the 3D point cloud for such a skeleton model that is markedly reduced in terms of its node count, so that a method for pose estimation is described below that is adapted to the inventive use of an anatomically motivated skeleton model.

A further embodiment of the invention relates to a storage medium that stores commands that can be carried out by an arithmetic unit and that cause the arithmetic unit to carry out the calculations described here for the purpose of pose estimation.