Scene analysis using image and range data
View Patent ↗Image and range data associated with an image can be processed to estimate planes within the 3D environment in the image. By utilizing image segmentation techniques, image data can identify regions of visible pixels having common features. These regions can be used to candidate regions for fitting planes to the range data based on a RANSAC technique.
1. A method comprising:
determining, respectively, a feature vector for each pixel of a plurality of pixels of a color image;
determining a feature space comprising the plurality of feature vectors;
randomly determining a plurality of first planes;
randomly dividing the feature space with the plurality of first planes to form a plurality of regions of the feature space;
determining a respective hash code for each region of the plurality of regions, wherein a hash code is indicative of a location of an associated region with respect to each first plane of the plurality of first planes;
determining a number of feature vectors that map to each hash code;
selecting a region of the plurality of regions based on a number of feature vectors that map to each hash code;
associating each feature vector of the selected region with a respective three dimensional location to generate a plurality of three dimensional locations;
partitioning the plurality of three dimensional locations into a plurality of partitions;
selecting a partition of the plurality of partitions;
identifying a first set of three dimensional locations associated with the selected partition; and
estimating a second plane that describes the three dimensional locations associated with the selected partition.
2. The method of claim 1 , wherein the estimating comprises a RANdom SAmple Consensus to estimate a best fitting plane.
3. The method of claim 1 , further comprising iteratively estimating the second plane.
4. The method of claim 1 , wherein the partition is selected based on a size of a segment associated with a partition.
5. The method of claim 1 , wherein the partition is selected based on a salient feature of the color image.
6. An apparatus comprising:
a processor; and
memory coupled to the processor, the memory comprising executable instructions that when executed by the processor cause the processor to effectuate operations comprising:
determining, respectively, a feature vector for each pixel of a plurality of pixels of a color image;
determining a feature space comprising the plurality of feature vectors;
randomly determining a plurality of first planes;
randomly dividing the feature space with the plurality of first planes to form a plurality of regions of the feature space;
determining a respective hash code for each region of the plurality of regions, wherein a hash code is indicative of a location of an associated region with respect to each first plane of the plurality of first planes;
determining a number of feature vectors that map to each hash code;
selecting a region of the plurality of regions based on a number of feature vectors that map to each hash code;
associating each feature vector of the selected region with a respective three dimensional location to generate a plurality of three dimensional locations;
partitioning the plurality of three dimensional locations into a plurality of partitions;
selecting a partition of the plurality of partitions;
identifying a first set of three dimensional locations associated with the selected partition; and
estimating a second plane that describes the three dimensional locations associated with the selected partition.
7. The apparatus of claim 6 , wherein the estimating comprises a RANdom SAmple Consensus to estimate a best fitting plane.
8. The apparatus of claim 6 , the operations further comprising iteratively estimating the second plane.
9. The apparatus of claim 6 , wherein the partition is selected based on a size of a segment associated with a partition.
10. The apparatus of claim 6 , wherein the partition is selected based on a salient feature of the color image.
11. Memory comprising executable instructions that when executed by a processor cause the processor to effectuate operations comprising:
determining, respectively, a feature vector for each pixel of a plurality of pixels of a color image;
determining a feature space comprising the plurality of feature vectors;
randomly determining a plurality of first planes;
randomly dividing the feature space with the plurality of first planes to form a plurality of regions of the feature space;
determining a respective hash code for each region of the plurality of regions, wherein a hash code is indicative of a location of an associated region with respect to each first plane of the plurality of first planes;
determining a number of feature vectors that map to each hash code;
selecting a region of the plurality of regions based on a number of feature vectors that map to each hash code;
associating each feature vector of the selected region with a respective three dimensional location to generate a plurality of three dimensional locations;
partitioning the plurality of three dimensional locations into a plurality of partitions;
selecting a partition of the plurality of partitions;
identifying a first set of three dimensional locations associated with the selected partition; and
estimating a second plane that describes the three dimensional locations associated with the selected partition.
12. The memory of claim 11 , wherein the-estimating comprises a RANdom SAmple Consensus to estimate a best fitting plane.
13. The memory of claim 11 , the operations further comprising iteratively estimating the second plane.
14. The memory of claim 11 , wherein the partition is selected based on a size of a segment associated with a partition.
15. The memory of claim 11 , wherein the partition is selected based on a salient feature of the color image.