Traffic obstruction detection

Abstract

One or more techniques and/or systems are provided for training and/or utilizing a traffic obstruction identification model for identifying traffic obstructions based upon vehicle location point data. For example, a training dataset, comprising sample vehicle location points (e.g., global positioning system location points of vehicles) and traffic obstruction identification labels (e.g., locations of known traffic obstructions such as stop signs, crosswalks, stop lights, etc.), may be evaluated to extract a set of training features indicative of traffic flow patterns. The set of training features and the traffic obstruction identification labels may be used to train a traffic obstruction identification model to create a trained traffic obstruction identification model. The trained traffic obstruction identification model may be used to determine whether a road segment has a traffic obstruction or not.

What is claimed is: 
     
         1 . A method for training a traffic obstruction identification model, comprising:
 obtaining a training dataset comprising sample vehicle location points and traffic obstruction identification labels, wherein the traffic obstruction identification labels correspond to traffic obstructions configured to control a flow of traffic and disposed at fixed locations along road segments;   extracting a set of training features from the training dataset based upon the sample vehicle location points, the set of training features indicative of traffic flow patterns, the traffic flow patterns indicative of a flow of traffic along a first set of one or more road segments; and   training the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create a trained traffic obstruction identification model for classifying uncategorized traffic obstructions along a second set of one or more road segments different than the first set of one or more road segments into one or more categories of traffic obstructions based upon traffic flow patterns of vehicles encountering the uncategorized traffic obstructions along the second set of one or more road segments.   
     
     
         2 . The method of  claim 1 , the traffic obstructions comprising at least one of a stop light, a stop sign, a crosswalk, or a railroad crossing. 
     
     
         3 . The method of  claim 1 , the extracting a set of training features comprising:
 evaluating the sample vehicle location points to identify a count of vehicles having speeds below a speed threshold; and   comparing the count of vehicles to a total count of vehicles to determine a vehicle speed feature for inclusion within the set of training features.   
     
     
         4 . The method of  claim 1 , the extracting a set of training features comprising:
 evaluating the sample vehicle location points to determine a median speed; and   identifying a standard deviation from the median speed to determine a median average deviation feature for inclusion within the set of training features.   
     
     
         5 . The method of  claim 1 , the extracting a set of training features comprising:
 identifying a first count of vehicle location points within a first road segment; and   comparing the first count of vehicle location points to counts of vehicle location points within one or more neighboring road segments to determine a relative point density feature for inclusion within the set of training features.   
     
     
         6 . The method of  claim 1 , the training the traffic obstruction identification model comprising:
 identifying one or more parameters for use by the trained traffic obstruction identification model based upon the training dataset and the set of training features.   
     
     
         7 . The method of  claim 1 , the extracting a set of training features comprising:
 extracting a first set of training features for a first road segment of the first set of one or more road segments and associated with a first traffic obstruction of the traffic obstructions; and   extracting a second set of training features for a second road segment of the first set of one or more road segments and associated with a second traffic obstruction of the traffic obstructions, the second traffic obstruction different than the first traffic obstruction.   
     
     
         8 . The method of  claim 1 , wherein the traffic flow patterns indicated by the set of training features comprise a stopping pattern along at least one road segment of the first set of one or more road segments. 
     
     
         9 . The method of  claim 1 , wherein the traffic flow patterns indicated by the set of training features comprise an acceleration pattern along at least one road segment of the first set of one or more road segments. 
     
     
         10 . The method of  claim 1 , wherein the traffic flow patterns indicated by the set of training features comprise a speed pattern along at least one road segment of the first set of one or more road segments. 
     
     
         11 . The method of  claim 1 , wherein the traffic flow patterns indicated by the set of training features comprise a length of a queued line of vehicles along at least one road segment of the first set of one or more road segments. 
     
     
         12 . The method of  claim 1 , wherein the traffic flow patterns indicated by the set of training features comprise at least one of:
 a stopping pattern along at least one road segment of the first set of one or more road segments,   an acceleration pattern along the at least one road segment of the first set of one or more road segments,   a speed pattern along the at least one road segment of the first set of one or more road segments, or   a length of a queued line of vehicles along the at least one road segment of the first set of one or more road segments.   
     
     
         13 . The method of  claim 1 , the training the traffic obstruction identification model comprising:
 training the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create the trained traffic obstruction identification model to identify a first traffic obstruction as being present on a first road segment of the second set of one or more road segments when a traffic flow pattern of the first road segment matches a traffic flow pattern of a second road segment of the first set of one or more road segments having the first traffic obstruction; and   training the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create the trained traffic obstruction identification model to identify a second traffic obstruction as being present on the first road segment of the second set of one or more road segments when the traffic flow pattern of the first road segment matches a traffic flow pattern of a third road segment of the first set of one or more road segments having the second traffic obstruction.   
     
     
         14 . A computer readable medium comprising instructions which when executed perform a method for determining a category of traffic obstruction present on a first road segment, comprising:
 obtaining a dataset comprising vehicle location points;   extracting a set of features from the dataset based upon the vehicle location points, the set of features indicative of traffic flow patterns of road segments; and   evaluating the set of features using a trained traffic obstruction identification model created from traffic obstruction identification labels corresponding to traffic obstructions configured to control a flow of traffic and disposed at fixed locations along the road segments to determine whether the first road segment, that is not part of the road segments, has a first category of traffic obstruction or a second category of traffic obstruction, wherein the evaluating comprises, for the first road segment:
 comparing a traffic flow pattern of the first road segment to the traffic flow patterns of the road segments; 
 classifying the first road segment has having the first category of traffic obstruction when the traffic flow pattern of the first road segment matches the traffic flow pattern of a second road segment of the road segments having the first category of traffic obstruction; and 
 classifying the first road segment has having the second category of traffic obstruction when the traffic flow pattern of the first road segment matches the traffic flow pattern of a third road segment of the road segments having the second category of traffic obstruction. 
   
     
     
         15 . The computer readable medium of  claim 14 , wherein the extracting a set of features comprises:
 evaluating the vehicle location points to identify a count of vehicles having speeds below a speed threshold; and   comparing the count of vehicles to a total count of vehicles to determine a vehicle speed feature for inclusion within the set of features.   
     
     
         16 . The computer readable medium of  claim 14 , wherein the extracting a set of features comprises:
 evaluating the vehicle location points to determine a median speed; and   identifying a standard deviation from the median speed to determine a median average deviation feature for inclusion within the set of features.   
     
     
         17 . The computer readable medium of  claim 14 , wherein the extracting a set of features comprises:
 identifying a first count of vehicle location points within a fourth road segment of the road segments; and   comparing the first count of vehicle location points to counts of vehicle location points within one or more neighboring road segments to determine a relative point density feature for inclusion within the set of features.   
     
     
         18 . A system for training a traffic obstruction identification model, comprising:
 a model training component configured to:
 obtain a training dataset comprising sample vehicle location points and traffic obstruction identification labels, wherein the traffic obstruction identification labels correspond to traffic obstructions configured to control a flow of traffic and disposed at fixed locations along road segments; 
 extract a set of training features from the training dataset based upon the sample vehicle location points, the set of training features indicative of traffic flow patterns, the traffic flow patterns indicative of a flow of traffic along a first set of one or more road segments; and 
 train the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create a trained traffic obstruction identification model for classifying uncategorized traffic obstructions along a second set of one or more road segments different than the first set of one or more road segments into one or more categories of traffic obstructions based upon traffic flow patterns of vehicles encountering the uncategorized traffic obstructions along the second set of one or more road segments. 
   
     
     
         19 . The system of  claim 18 , wherein the traffic flow patterns indicated by the set of training features comprise at least one of:
 a stopping pattern along at least one road segment of the first set of one or more road segments,   an acceleration pattern along the at least one road segment of the first set of one or more road segments,   a speed pattern along the at least one road segment of the first set of one or more road segments, or   a length of a queued line of vehicles along the at least one road segment of the first set of one or more road segments.   
     
     
         20 . The system of  claim 18 , wherein training the traffic obstruction identification model comprises:
 training the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create the trained traffic obstruction identification model to identify a first traffic obstruction as being present on a first road segment of the second set of one or more road segments when a traffic flow pattern of the first road segment matches a traffic flow pattern of a second road segment of the first set of one or more road segments having the first traffic obstruction; and   training the traffic obstruction identification model using the set of training features and the traffic obstruction identification labels to create the trained traffic obstruction identification model to identify a second traffic obstruction as being present on the first road segment of the second set of one or more road segments when the traffic flow pattern of the first road segment matches a traffic flow pattern of a third road segment of the first set of one or more road segments having the second traffic obstruction.