Curb detection by analysis of reflection images

Abstract

Disclosed are devices, systems and methods for road feature detection using light detection and ranging (LiDAR) sensors. One example of a method for road feature detection includes obtaining a point-cloud frame that comprises a description of an intensity of a reflection of beams from an area around a vehicle. The example method further includes creating a plurality of clusters that each include (i) one or more seed points of the point-cloud frame, and (ii) additional points of the point-cloud frame based on a relationship between the additional points and the one or more seed points. The example method further includes identifying a cluster from the plurality of clusters based on a total number of points included in each of the plurality of clusters. The example method further includes detecting a road feature from the cluster.

What is claimed is: 
     
         1 . A method for detecting road features, comprising:
 obtaining a point-cloud frame that comprises a description of an intensity of a reflection of beams from an area around a vehicle;   creating a plurality of clusters that each include (i) one or more seed points of the point-cloud frame, and (ii) additional points of the point-cloud frame based on a relationship between the additional points and the one or more seed points;   identifying a cluster from the plurality of clusters based on a total number of points included in each of the plurality of clusters; and   detecting a road feature from the cluster.   
     
     
         2 . The method of  claim 1 , further comprising:
 determining that the additional points are related to the one or more seed points based on determining that the additional points are neighboring to the one or more seed points and meet a criterion associated with the one or more seed points.   
     
     
         3 . The method of  claim 2 , wherein the additional points meet the criterion based on having values that are within a tolerance of values of the one or more seed points. 
     
     
         4 . The method of  claim 1 , wherein identifying the cluster includes:
 removing certain clusters of the plurality of clusters based on a growth of a certain clusters stopping during a region-growing process in which the additional points are added to each cluster, and   identifying the cluster from remaining clusters of the plurality of clusters.   
     
     
         5 . The method of  claim 1 , the additional points included in a given cluster are determined using a k-dimensional tree constructed based on features of each point of the point-cloud frame. 
     
     
         6 . The method of  claim 1 , wherein detecting the road feature comprises detecting a drivable area for the vehicle from a portion of the area around the vehicle that is spanned by the cluster. 
     
     
         7 . The method of  claim 1 , wherein detecting the road feature comprises detecting, according to a boundary of the cluster, a curb of a road, wherein the vehicle is located on the road. 
     
     
         8 . The method of  claim 1 , wherein obtaining the point-cloud frame comprises:
 acquiring two consecutive single-frame point-clouds; and   determine an accumulated point-cloud by registering the two consecutive single-frame point clouds into a common coordinate system, wherein the point-cloud frame includes the accumulated point-cloud.   
     
     
         9 . An apparatus for detecting road features, comprising:
 a processor; and   a memory comprising executable code that, when executed by the processor, causes the apparatus to:   obtain a point-cloud frame that comprises a description of an intensity of a reflection of beams from an area around a vehicle;   create a plurality of clusters that each include (i) one or more seed points of the point-cloud frame, and (ii) additional points of the point-cloud frame based on a relationship between the additional points and the one or more seed points;   identify a cluster from the plurality of clusters based on a total number of points included in each of the plurality of clusters; and   detect a road feature from the cluster.   
     
     
         10 . The apparatus of  claim 9 , wherein the relationship between the additional points and the one or more seed points includes the additional points meeting a criterion associated with the one or more seed points. 
     
     
         11 . The apparatus of  claim 9 , wherein a given cluster includes the additional points based on the additional points having feature values that are within a tolerance of corresponding feature values of the given cluster, wherein the feature values include a curvature value and a normal value. 
     
     
         12 . The apparatus of  claim 9 , wherein identifying the cluster includes:
 including, by a region-growing process, the additional points for each of the plurality of clusters,   remove a certain cluster from the plurality of clusters based on a lack of additional points to include for the certain cluster, and   identifying the cluster from remaining clusters of the plurality of clusters.   
     
     
         13 . The apparatus of  claim 9 , wherein
 construct a k-dimensional tree from the point-cloud frame, and   include the additional points in a given cluster based on a nearest neighbor search that uses the k-dimensional tree.   
     
     
         14 . The apparatus of  claim 9 , wherein detecting the road feature comprises detecting a road surface from a portion of the area around the vehicle that is spanned by the cluster. 
     
     
         15 . The apparatus of  claim 9 , wherein detecting the road feature comprises detecting, according to one or more concave hulls of the cluster, one or more curbs of a road, wherein the vehicle is located on the road. 
     
     
         16 . The apparatus of  claim 9 , wherein obtaining the point-cloud frame comprises:
 acquiring a plurality of consecutive single-frame point-clouds using a light detection and ranging (LiDAR) sensor;   determine the point-cloud frame to include the plurality of consecutive single-frame point-clouds using a common coordinate system.   
     
     
         17 . A non-transitory computer-readable medium storing a program that causes a computer to execute a process, the process comprising:
 obtaining a point-cloud frame that comprises a description of an intensity of a reflection of beams from an area around a vehicle;   creating a plurality of clusters that each include (i) one or more seed points of the point-cloud frame, and (ii) additional points of the point-cloud frame based on a relationship between the additional points and the one or more seed points;   identifying a cluster from the plurality of clusters based on a total number of points included in each of the plurality of clusters; and   detecting a road feature from the cluster.   
     
     
         18 . The non-transitory computer-readable medium of  claim 17 , wherein the additional points are neighboring to the one or more seed points and meet a criterion associated with the one or more seed points. 
     
     
         19 . The non-transitory computer-readable medium of  claim 17 , wherein the road feature includes a drivable area for the vehicle, wherein the drivable area is captured by a span of the cluster. 
     
     
         20 . The non-transitory computer-readable medium of  claim 17 , wherein the road feature includes a curb of a road, wherein the curb is captured by a boundary of the cluster, and wherein the vehicle is located on the road.