METHOD AND COMPUTER SYSTEM FOR CALIBRATING RELATIVE SENSOR ORIENTATIONS

1 . A method for calibrating the relative orientation of a spatial sensor and an area sensor, wherein the measuring zones of the spatial sensor and the area sensor are overlapping, the method comprising

receiving, by a computing system, a plurality of sensor frames of the spatial sensor and the area sensor;

grouping, by the computing system, sensor frames of the spatial sensor with sensor frames of the area sensor based on temporal correlation;

selecting, by the computing system, at least one group of frames comprising a spatial sensor frame and an area sensor frame;

receiving, by the computing system, annotations for the sensor frames in the at least one selected group, wherein the annotations comprise bounding boxes for detected objects;

projecting, by the computing system, the corners of a three-dimensional bounding box in the spatial sensor frame to an image plane of the area sensor, to produce a projected rectangle;

calculating, by the computing system, an energy term based on the coordinates of the corners of the projected rectangle and the corresponding bounding box in the area sensor frame; and

optimizing, by the computing system, the relative sensor orientation by minimizing the energy term.

2 . The method according to claim 1 , wherein the plurality of sensor frames are consecutive, and wherein a limited number of groups of frames is selected for optimizing the relative sensor orientation.

3 . The method according to claim 2 , wherein the selected groups of frames comprise one group at the beginning of the sequence of consecutive frames and one group at the end of the sequence of consecutive frames.

4 . The method according to claim 1 , wherein the annotations for at least one group of frames are produced by a neural network trained on spatial sensor frames and/or a neural network trained on area sensor frames and/or a human annotator.

5 . The method according to claim 1 , wherein for a group of frames comprising multiple spatial sensor frames and area sensor frames and/or for a group of spatial sensor frame and area sensor frame comprising multiple annotated objects, optimizing the relative sensor orientation is based on the sum of the energy terms for all annotated objects in the selected group.

6 . The method according to claim 1 , wherein the intrinsic parameters of the area sensor are kept constant when optimizing the relative orientation.

7 . The method according to claim 1 , wherein for corner coordinates x0, y0, x1, y1 of the bounding box and for corner coordinates x0′, y0′, x1′, y1′ of the corresponding projected rectangle, the energy term is

E

=

1

(

x

1

-

x

0

)

⁢

(

y

1

-

y

0

)

⁢

(

(

x

0

-

x

0

′

)

2

+

(

y

0

-

y

0

′

)

2

+

(

x

1

-

x

1

′

)

2

+

(

y

1

-

y

1

′

)

2

)

.

8 . The method according to claim 1 , wherein optimizing the relative sensor orientation is performed iteratively until a convergence criterion is met.

9 . The method according to claim 8 , wherein the relative sensor orientation is described using a matrix describing rotation and translation of a sensor in three-dimensional space, and wherein during or subsequent to each optimization iteration, a factorization algorithm is applied to the rotation matrix for ensuring orthonormality.

10 . The method according to claim 1 , wherein for data sets comprising additional sensor data, additional groups of spatial sensor frames and area sensor frames are selected based on the acceleration in at least one space axis.

11 . The method according to claim 1 , wherein the plurality of sensor frames are consecutive sensor frames;

wherein the selected groups of frames represent a fraction of the received sequence of consecutive frames; and

wherein all of the consecutive spatial sensor frames are labeled;

wherein the method further comprises:

projecting the corners of a three-dimensional bounding box in the space sensor frame to the image plane of the area sensor, to produce a projected rectangle,

using the projected rectangle as two-dimensional bounding box in the area sensor frame, and

determining at least one attribute of the object.

12 . A non-transitory computer-readable medium having processor-executable instructions stored thereon for calibrating the relative orientation of a spatial sensor and an area sensor, wherein the measuring zones of the spatial sensor and the area sensor are overlapping, and wherein the processor-executable instructions, when executed, facilitate performance of the following:

receiving, by a computing system, a plurality of sensor frames of the spatial sensor and the area sensor;

grouping, by the computing system, sensor frames of the spatial sensor with sensor frames of the area sensor based on temporal correlation;

selecting, by the computing system, at least one group of frames comprising a spatial sensor frame and an area sensor frame;

receiving, by the computing system, annotations for the sensor frames in the at least one selected group, wherein the annotations comprise bounding boxes for detected objects;

projecting, by the computing system, the corners of a three-dimensional bounding box in the spatial sensor frame to an image plane of the area sensor, to produce a projected rectangle;

calculating, by the computing system, an energy term based on the coordinates of the corners of the projected rectangle and the corresponding bounding box in the area sensor frame; and

optimizing, by the computing system, the relative sensor orientation by minimizing the energy term.

13 . A computer system, comprising:

a host computer comprising a processor, a working memory, a display, an input device and a non-volatile memory;

wherein the non-volatile memory comprises instructions;

wherein the processor is configured to execute the instructions to facilitate the following being performed by the computing system:

receiving a plurality of sensor frames of the spatial sensor and the area sensor;

grouping sensor frames of the spatial sensor with sensor frames of the area sensor based on temporal correlation;

selecting at least one group of frames comprising a spatial sensor frame and an area sensor frame;

receiving annotations for the sensor frames in the at least one selected group, wherein the annotations comprise bounding boxes for detected objects;

projecting the corners of a three-dimensional bounding box in the spatial sensor frame to an image plane of the area sensor, to produce a projected rectangle;

calculating an energy term based on the coordinates of the corners of the projected rectangle and the corresponding bounding box in the area sensor frame; and

optimizing the relative sensor orientation by minimizing the energy term.