Method and apparatus for determining driving information

1. A method of determining driving information of a vehicle, the method comprising:

generating a probability map corresponding to a target object to be detected from a driving image, the probability map being generated based on determining a probability value for a unit of at least one pixel;

extracting representative regions from a candidate region of the target object detected from the driving image;

extracting at least two of the extracted representative regions, based on the generated probability map, the at least two of the extracted representative regions having highest probability values among all probability values of the extracted representative regions;

calculating a confidence value for the candidate region of the target object, to be an average value of the highest probability values of the extracted at least two of the representative regions; and

determining the driving information of the vehicle, based on the calculated confidence value,

wherein the target object comprises lane lines including boundary lines of a driving lane of the vehicle, and

wherein the determining of the driving information of the vehicle comprises:

comparing each of confidence values of the boundary lines of the driving lane with a preset threshold value; and

based on each of the confidence values of the boundary lines of the driving lane being compared to be greater than the preset threshold value, determining a center between the boundary lines of the driving lane as a center line of the driving lane.

2. The method of claim 1 , wherein the highest probability values are a top threshold percentage among the all probability values, the top threshold percentage being determined based on any one or any combination of a traffic situation, a country-specific road traffic law, and a road type.

3. The method of claim 1 , wherein the extracting of the representative regions comprises extracting the representative regions from the candidate region of the target object, at equal intervals.

4. The method of claim 1 , wherein the extracting of the representative regions comprises extracting the representative regions from the candidate region of the target object, based on either one or both of a preset distance range corresponding to a region of interest (ROI) and a number of preset representative regions.

5. The method of claim 1 , wherein the extracting of the representative regions comprises:

converting the candidate region of the target object from an image domain to a world domain; and

extracting the representative regions from the candidate region converted to the world domain, at equal intervals.

6. The method of claim 1 , wherein the generating of the probability map comprises generating the probability map corresponding to the target object by calculating a plurality of probability values in units of one pixel, based on the driving image, each of the plurality of probability values comprising a probability value that a corresponding pixel will correspond to the target object.

7. The method of claim 1 , wherein the generating of the probability map comprises:

determining pixel-by-pixel classes of the driving image, using a scene segmentation algorithm; and

generating a class-by-class probability map by calculating a pixel-by-pixel probability value of each of the pixel-by-pixel classes.

8. The method of claim 1 , wherein the determining of the driving information of the vehicle comprises, based on one of the confidence values of one of the boundary lines of the driving lane being compared to be greater than the preset threshold value and another one of the confidence values of another one of the boundary lines of the driving lane being compared to be less than the preset threshold value:

generating a virtual lane boundary line, based on the one of the boundary lines of the driving lane and a width of the vehicle; and

determining a center between the generated virtual lane boundary line and the one of the boundary lines of the driving lane as the center line of the driving lane.

9. The method of claim 1 , further comprising adjusting a lateral position of the vehicle, based on information of the determined center line of the driving lane.

10. A method of determining driving information of a vehicle, the method comprising:

generating a probability map corresponding to a lane line to be detected from a driving image, based on pixel-by-pixel segmentation, the probability map being generated based on determining a probability value for a unit of at least one pixel;

detecting a lane line region, based on line fitting on the generated probability map;

extracting representative regions from the detected lane line region;

extracting at least two of the extracted representative regions, based on the generated probability map, the at least two of the extracted representative regions corresponding to highest probability values of a predetermined percentage among all probability values of the extracted representative regions;

calculating a confidence value for the lane line region, to be an average value of the highest probability values of the extracted at least two of the representative regions; and

determining the driving information of the vehicle, based on the calculated confidence value,

wherein the target object comprises lane lines including boundary lines of a driving lane of the vehicle, and

wherein the determining of the driving information of the vehicle comprises:

comparing each of confidence values of the boundary lines of the driving lane with a preset threshold value; and

based on each of the confidence values of the boundary lines of the driving lane being compared to be greater than the preset threshold value, determining a center between the boundary lines of the driving lane as a center line of the driving lane.

11. An apparatus for determining driving information of a vehicle, the apparatus comprising:

a sensor configured to obtain a driving image;

a processor configured to:

generate a probability map corresponding to a target object to be detected from the driving image, the probability map being generated based on determining a probability value for a unit of at least one pixel;

extract representative regions from a candidate region of the target object detected from the driving image;

extract at least two of the extracted representative regions, based on the generated probability map, the at least two of the extracted representative regions having highest probability values among all probability values of the extracted representative regions;

calculating a confidence value for the candidate region of the target object, to be an average value of the highest probability values of the extracted at least two the representative regions; and

determine the driving information of the vehicle, based on the calculated confidence value; and

a communication interface configured to output the determined driving information,

wherein the target object comprises lane lines including boundary lines of a driving lane of the vehicle, and

wherein the processor is further configured to:

compare each of confidence values of the boundary lines of the driving lane with a preset threshold value; and

based on each of the confidence values of the boundary lines of the driving lane being compared to be greater than the preset threshold value, determine a center between the boundary lines of the driving lane as a center line of the driving lane.

12. The apparatus of claim 11 , wherein the highest probability values are a top threshold percentage among the all probability values, the top threshold percentage being determined based on any one or any combination of a traffic situation, country-specific road traffic law, and a road type.

13. The apparatus of claim 11 , wherein the processor is further configured to extract the representative regions from the candidate region of the target object, at equal intervals.

14. The apparatus of claim 11 , wherein the processor is further configured to extract the representative regions from the candidate region of the target object, based on either one or both of a preset distance range corresponding to a region of interest (ROI) and a number of preset representative regions.