Image-based method of defining a scanning area
Image-based method of defining a scanning area with a measuring radiation-based measuring device such as a laser scanner, the method comprising roughly aligning a targeting axis of the measuring device towards an area of interest to be scanned, the area of interest having a coherent visual property, capturing with known orientation to the targeting axis an image covering the area of interest, displaying the image to a user, manually selecting by the user of at least one pixel of a group of pixels representing the area of interest in the image, automatically segmenting the group of pixels based on a similarity to the at least one selected pixel using a machine learning algorithm, and automatically defining the scanning area based on border pixels of the first group of pixels.
1 . An image-based method of defining a scanning area for, in particular metrological and/or geodetical, scanning with an optoelectronic radiation-based measuring device, the method comprising:
capturing with known orientation to a targeting axis of the measuring device a first image covering a first area of interest, the first area of interest having a coherent visual property,
displaying the first image to a user,
manually selecting at least one pixel of a first group of pixels representing the first area of interest in the image by the user, in particular by pointing on the displayed first image,
automatically segmenting the first group of pixels based on a similarity to the at least one selected pixel using a machine learning algorithm,
automatically defining the scanning area, in particular by defining vertical and horizontal outer angles for the targeting axis alignment, based on border pixels of the first group of pixels.
2 . The method according to claim 1 , wherein manually roughly aligning the targeting axis towards the first area of interest before capturing the first image.
3 . The method according to claim 2 , wherein the first image is captured in course of an automatic, in particular 360°, panorama imaging with the targeting axis being automatically aligned.
4 . The method according to claim 1 , wherein the first image is captured in course of an automatic, in particular 360°, panorama imaging with the targeting axis being automatically aligned.
5 . The method according to claim 1 , wherein the coherent visual property and the pixel similarity relate to color and/or texture.
6 . The method according to claim 1 , wherein the pixel similarity relates to color value, saturation and/or brightness.
7 . The method according to claim 1 , wherein:
computing for each pixel of the first image an embedding vector,
mapping the embedding vectors into a n-dimensional embedding space,
clustering of embedding vectors in the embedding space and
segmenting said first group of pixels according to the cluster comprising the embedding vector of the selected pixel.
8 . The method according to claim 1 , wherein automatically segmenting comprises:
computing for each pixel of the first image an embedding vector,
successively comparing embedding vectors of neighboring pixels to an initial pixel, starting with the selected pixel, and
adding pixels to said first group of pixels as long as the comparing of according embedding vectors reveals a difference below a defined threshold.
9 . The method according to claim 1 , wherein automatically centering the targeting axis before starting a scanning procedure to the defined scanning area based on a center and/or dimension of the first group of pixels.
10 . The method according to claim 1 , wherein the border pixels of the first group of pixels are defined using a geometric shape, in particular a polygon, fitted to the border of the first group of pixels.
11 . The method according to claim 1 , wherein the scanning area is defined including a buffer range which is extending a range given by the border pixels of the first group of pixels, whereby the buffer range is established by:
adding a band of pixels around the first group of pixels and/or
adding respective angle values to outer angles for the targeting axis alignment derived from the border pixels of the first group of pixels and/or
the buffer range is dependent on a property of the first group of pixels or subgroups thereof, in particular size, shape, position, saturation and/or brightness.
12 . The method according to claim 1 , wherein in case a second area of interest having a similar visual property as the first area of interest is imaged in the first image, the method comprises:
automatically segmenting a second group of pixels representing the second area of interest based on the similarity to the at least one selected pixel using a machine learning algorithm, and
defining the scanning area or a further scanning area based on border pixels of the second group of pixels.
13 . The method according to claim 1 , wherein:
storing information relating to the first group of pixels, in particular at least an embedding vector of the manually selected pixel,
capturing a second image with known orientation relation to the targeting axis, in particular whereby the first and second image are captured from different locations, of a further area of interest of similar visual property as the first area of interest,
defining a further scanning area based on the second image and the stored information without manually selecting of a pixel in the second image.
14 . The method according to claim 1 , wherein the defining of the scanning area is according to a user's choice by manual input such that an area directly attached to and/or enclosed by the first area of interest but of different visual property is either included or excluded in the scanning area, in particular wherefore according to the user's choice the defining is based either on outer border pixels or on outer and inner border pixels.
15 . A computer program product comprising program code which is stored on a non-transitory machine-readable medium, and having computer-executable instructions for performing, when run on a computer unit of a scanning measuring device, the method according to claim 14 .
16 . A computer program product comprising program code which is stored on a non-transitory machine-readable medium, and having computer-executable instructions for performing, when run on a computer unit of a scanning measuring device, the method according to claim 1 .
17 . An optoelectronic radiation-based measuring device comprising:
a targeting axis,
a camera with known orientation to the targeting axis and
a control and evaluation unit, the control and evaluation unit being configured to:
automatically segment, using a machine learning algorithm, a group of pixels within an image of an area of interest of coherent visual property taken with the camera based on a similarity to an at least one pixel selected by a user within the image,
automatically define a scanning area, in particular by defining pairs of vertical and horizontal outer angles for the targeting axis alignment, based on border pixels of the group of pixels, and
automatically initiate a measuring of the scanning area.
18 . A computer program product comprising program code which is stored on a non-transitory machine-readable medium, and having computer-executable instructions for performing a method when run on a computer unit of the scanning measuring device according to claim 17 .