Method for identification of candidate points as possible characteristic points of a calibration pattern within an image of the calibration pattern

1. A method for identification of candidate points as possible characteristic points of a calibration pattern within an image of the calibration pattern, the method comprising the steps of:

determining spots within a filtered image derived from the image of the calibration pattern, with a spot being defined as a coherent set of pixels of the filtered image having pixel values exceeding a threshold,

for each determined spot calculating a central point of the determined spot, and

identifying as candidate points all calculated central points,

wherein the step of determining the spots comprises scanning the filtered image one pixel of the filtered image after another,

wherein, if a scanned pixel has a pixel value exceeding the threshold, the scanning is paused and a spot contour tracing is executed starting with the last scanned pixel as current pixel and the second last pixel as previous pixel,

the spot contour tracing comprising as a tracing step the steps of logging the current pixel as a contour pixel,

for neighboring pixels adjacent to the current pixel, evaluating the condition, whether both the neighboring pixel has a pixel value exceeding the threshold and another neighboring pixel, which immediately precedes this neighboring pixel with respect to a sense of rotation around the current pixel, has a pixel value not exceeding the threshold,

selecting from all neighboring pixels, for which the condition is fulfilled, the neighboring pixel farthest from the previous pixel with respect to said sense of rotation, and

defining the current pixel as previous pixel and defining the selected neighboring pixel as current pixel;

the spot contour tracing further comprising the steps of repeating the tracing step until the current pixel is again equal to the last scanned pixel,

determining the spot as the set of pixels comprising the logged contour pixels and the pixels surrounded by the logged contour pixels.

2. The method in accordance with claim 1 , wherein calculating the central point of a determined spot comprises the steps of

determining a first coordinate of the central point by calculating the sum of the products of a respective first coordinate and a respective pixel value of each of the pixels of the spot and dividing this sum by the sum of the pixel values of the pixels of the spot, and

determining a second coordinate of the central point by calculating the sum of the products of a respective second co-ordinate and a respective pixel value of each of the pixels of the spot and dividing this sum by the sum of the pixel values of the pixels of the spot.

3. The method in accordance with claim 1 , wherein the filtered image is derived from a smoothed image obtained from the image of the calibration pattern by, for each pixel of the image of the calibration pattern; replacing a pixel value of the pixel by an average over the pixel values of pixels in a vicinity of the pixel.

4. The method in accordance with claim 1 , wherein the filtered image is obtained from the image of the calibration pattern or a smoothed image derived from the image of the calibration pattern by convolution of the image or the smoothed image, respectively, with a filter kernel specific to the calibration pattern or to a sub-pattern of the calibration pattern.

5. The method in accordance with claim 4 , wherein the convolution of the image or the smoothed image comprises the steps of

obtaining an integral image from the image or the smoothed image by, for each pixel of the image or the smoothed image, replacing the pixel value of the respective pixel by the sum of the pixel values of pixels having smaller first and second coordinates than the respective pixel; and

convolving the integral image with a modified filter kernel specific to the calibration pattern or to the sub-pattern of the calibration pattern.

6. The method in accordance with claim 5 , wherein the modified filter kernel comprises a field of kernel values corresponding to said sub-pattern by the kernel value of a point of the field being

equal to one, if the point corresponds to a corner of any of the squares and the sum of first and second coordinates of the corner is smaller or greater than each sum of first and second coordinates of the other corners of the same square,

equal to minus one, if the point corresponds to a corner of any of the squares and the sum of first and second coordinates of the corner is neither smaller nor greater than each sum of first and second coordinates of the other corners of the same square, and

equal to zero otherwise.

7. The method in accordance with claim 1 , wherein the calibration pattern comprises a sub-pattern formed by two equally oriented squares contrasting with a background of the calibration pattern, with a corner of one of the squares coinciding with a corner of the other of the squares.

8. The method in accordance with claim 1 , wherein the filter kernel comprises a field of kernel values corresponding to said sub-pattern by the kernel value of a point of the field being

equal to one, if the point corresponds to any of the squares of the sub-pattern, and

equal to zero otherwise.

9. The method in accordance with claim 1 , wherein the calibration pattern is an arrangement of sub-patterns comprising at least a first sub-pattern and a second sub-pattern, in particular with the second sub-pattern being an inverse of the first sub-pattern.

10. The method in accordance with claim 9 , wherein the filtered image is obtained from the image of the calibration pattern or a smoothed image derived from the image of the calibration pattern by

convolution of the image or the smoothed image with a filter kernel specific to a superposition of the first sub-pattern and the second sub-pattern; and

after convolution, for each pixel of the image or the smoothed image, replacing a pixel value of the pixel by the absolute value of the pixel value.

11. The method in accordance with claim 10 , wherein the convolution of the image or the smoothed image comprises the steps of

obtaining an integral image from the image or the smoothed image by, for each pixel of the image or the smoothed image, replacing a pixel value of the pixel by the sum of the pixel values of pixels having smaller first and second coordinates than the pixel; and

convolving the integral image with a modified filter kernel specific to the superposition of the first sub-pattern and the second sub-pattern.

12. The method in accordance with claim 11 , wherein the modified filter kernel comprises a field of kernel values corresponding to the superposition of the first sub-pattern and the second sub-pattern by the kernel value of a point of the field being

equal to one, if the point corresponds to a corner of any of the squares of the first sub-pattern and the sum of first and second coordinates of the corner is smaller or greater than each sum of first and second coordinates of the other corners of the same square,

equal to minus one, if the point corresponds to a corner of any of the squares of the first sub-pattern and the sum of first and second coordinates of the corner is neither smaller nor greater than each sum of first and second coordinates of the other corners of the same square,

equal to minus one, if the point corresponds to a corner of any of the squares of the second sub-pattern and the sum of first and second coordinates of the corner is smaller or greater than each sum of first and second coordinates of the other corners of the same square,

equal to one, if the point corresponds to a corner of any of the squares of the second sub-pattern and the sum of first and second coordinates of the corner is neither smaller nor greater than each sum of first and second coordinates of the other corners of the same square, and

equal to zero otherwise.

13. The method in accordance with claim 10 , wherein the filter kernel comprises a field of kernel values corresponding to the superposition of the first sub-pattern and the second sub-pattern by the kernel value of a point of the field being

equal to one, if the point corresponds to any of the squares of the first sub-pattern,

equal to minus one, if the point corresponds to any of the squares of the second sub-pattern, and

equal to zero otherwise.

14. The method in accordance with claim 9 , wherein each of the first and second sub-patterns is formed by two equally oriented squares contrasting with a back-ground of the calibration pattern, with a corner of one of the squares coinciding with a corner of the other of the squares,

wherein the relative orientation of the two squares of the first sub-pattern is perpendicular to the relative orientation of the two squares of the second sub-pattern.

15. A method for identification of candidate points as possible characteristic points of a calibration pattern within an image of the calibration pattern, the method comprising the steps of:

determining spots within a filtered image derived from the image of the calibration pattern, with a spot being defined as a coherent set of pixels of the filtered image having pixel values exceeding a threshold;

calculating for each determined spot a central point of the determined spot; and

identifying all calculated central points as candidate points; wherein

the step of determining the spots comprises scanning the filtered image one pixel of the filtered image after another, and wherein;

if a scanned pixel has a pixel value exceeding the threshold, the scanning is paused and a spot contour tracing is executed starting with the last scanned pixel as current pixel and the second last pixel as previous pixel.

16. The method in accordance with claim 15 , wherein the spot contour tracing comprising as a tracing step the steps of:

logging the current pixel as a contour pixel;

for neighboring pixels adjacent to the current pixel, evaluating the condition, whether both the neighboring pixel has a pixel value exceeding the threshold and another neighboring pixel, which immediately precedes this neighboring pixel with respect to a sense of rotation around the current pixel, has a pixel value not exceeding the threshold;

selecting from all neighboring pixels, for which the condition is fulfilled, the neighboring pixel farthest from the previous pixel with respect to said sense of rotation; and

defining the current pixel as previous pixel and defining the selected neighboring pixel as current pixel.

17. The method in accordance with claim 15 , wherein the spot contour tracing further comprising the steps of:

repeating the tracing step until the current pixel is again equal to the last scanned pixel; and

determining the spot as the set of pixels comprising the logged contour pixels and the pixels surrounded by the logged contour pixels.

18. A system for identification of candidate points as possible characteristic points of a calibration pattern within an image of the calibration pattern, the system comprising:

an image of a calibration pattern captured by a camera mounted to a vehicle; and

a microcontroller in communication with the camera, the microcontroller configured to:

determine spots within a filtered image derived from the image of the calibration pattern, with a spot being defined as a coherent set of pixels of the filtered image having pixel values exceeding a threshold;

calculate for each determined spot a central point of the determined spot; and

identify all calculated central points as candidate points; wherein

the microcontroller is further configured to determine the spots by scanning the filtered image one pixel of the filtered image after another; and wherein

if a scanned pixel has a pixel value exceeding the threshold, the microcontroller is further configured to pause the scanning and execute a spot contour tracing starting with the last scanned pixel as current pixel and the second last pixel as previous pixel.

19. The system in accordance with claim 18 , wherein the microcontroller is further configured to:

log the current pixel as a contour pixel;

for neighboring pixels adjacent to the current pixel, evaluate the condition, whether both the neighboring pixel has a pixel value exceeding the threshold and another neighboring pixel, which immediately precedes this neighboring pixel with respect to a sense of rotation around the current pixel, has a pixel value not exceeding the threshold;

select from all neighboring pixels, for which the condition is fulfilled, the neighboring pixel farthest from the previous pixel with respect to said sense of rotation; and

define the current pixel as previous pixel and defining the selected neighboring pixel as current pixel.

20. The system in accordance with claim 18 , wherein the microcontroller is further configured to:

repeat the tracing step until the current pixel is again equal to the last scanned pixel; and

determine the spot as the set of pixels comprising the logged contour pixels and the pixels surrounded by the logged contour pixels.