Back to results
US 11,378,398 B2Granted
Surveying device with automatic training of locked object or person for camera based target tracking
AssigneeLeica Geosystems AG
InventorsUlrich Hornung, Zoltán Török, Veroljub Maksimovic, Bernhard Metzler
Life of this patent
Estimating the term…
Abstract
The invention generally relates to an industrial or geodetic surveying device having a tracking unit configured to receive over a tracking channel a tracking signal which is used to track a cooperative target. According to the invention, the surveying device comprises a camera to generate image data, wherein based on the additional knowledge about the position of the cooperative target in an image generated in the locked state, the visual appearance of the target carrier is learned and the result is stored in the form of a detector for use in a backup tracking mode.
Claims14
An industrial or geodetic surveying device comprising:
- a base;
- a tracking unit, which is rotatable about two axes relative to the base and defines an aiming axis, the tracking unit being configured to:
- receive over a tracking channel a tracking signal which is indicative of a change in angular location of a cooperative target of a target carrier with respect to the industrial or geodetic surveying device,
- to determine the change in angular location of the cooperative target based on the tracking signal, and
- to generate control data for adjusting an alignment of the aiming axis based on the determined change in angular location,
- angle measuring sensors configured to generate angle data indicative of an orientation of the tracking unit with respect to the base;
- and
- a camera configured to generate image data,
- wherein the industrial or geodetic surveying device is configured for operation in a locked state, wherein the tracking channel is undisturbed, namely such that the tracking signal is receivable without unscheduled interruption, and the target carrier is continuously tracked based on the control data,
- wherein the industrial or geodetic surveying device is configured to:
- generate a series of images by the camera in the locked state and provide the series of images to a training algorithm based on computer vision and/or machine learning, wherein each image of the series of images comprises an image of the cooperative target and an image of the target carrier,
- train a detector for the target carrier, wherein the training algorithm is configured to provide as the detector a detection model for detecting the target carrier by
- assigning a common imaging area to the images of the series of images by making use of a known position of the image of the cooperative target provided by the locked state, wherein the common imaging area provides an estimated position of the image of the target carrier relative to the known position of the image of the cooperative target, and
- analyzing the common imaging area of the series of images and, based thereof, learning on the fly recognition indicia indicative of a visual appearance of the target carrier with respect to different relative point of views of the surveying device onto the target carrier, and
- perform a backup tracking mode, wherein
- image data is generated,
- a position of the target carrier within an image of the image data is determined by finding the image of the target carrier within the image data using the trained detection model to detect and locate the target carrier, and
- the aiming axis is adjusted based on the determined position of the target carrier.
15 more elements
A computer program product comprising a non-transitory machine-readable medium having computer-executable instructions stored thereon that when executed by a computer cause performance of a method comprising:
- generating a series of images by a camera in a locked state of an industrial or geodetic surveying device and providing the series of images to a training algorithm based on computer vision and/or machine learning, wherein each image of the series of images comprises an image of a cooperative target and an image of a target carrier,
- training a detector for the target carrier, wherein the training algorithm is configured to provide as the detector a detection model for detecting the target carrier by
- assigning a common imaging area to the images of the series of images by making use of a known position of the image of the cooperative target provided by the locked state, wherein the common imaging area provides an estimated position of the image of the target carrier relative to the known position of the image of the cooperative target, and
- training a detector for the target carrier by analyzing the common imaging area of the series of images and, based thereof, learning recognition indicia indicative of a visual appearance of the target carrier with respect to different relative point of views of the surveying device onto the target carrier,
- generating image data,
- determining a position of the target carrier within an image of the image data by finding the image of the target carrier within the image data using the trained detection model to detect and locate the target carrier, and
- adjusting an aiming axis based on the determined position of the target carrier.
4 more elements
Dependencies are inferred from the claim text.