Estimating color of vehicles on a roadway

Abstract

The present invention extends to methods, systems, and computer program products for estimating color of vehicles on a roadway. The color can be at least estimated from captured imagery, such as, frames of a video stream. The captured imagery can be captured by variably configured traffic cameras, other public cameras, private cameras (e.g., CCTV), or virtually any other camera, etc., which may have different operating conditions.

What is claimed: 
     
         1 . A method comprising:
 accessing a camera image of an unknown quality that was captured by a camera of an unknown configuration;   determining one or more image characteristics and image content of the camera image;   detecting a vehicle in the camera image considering the one or more image characteristics;   using one or more vehicle type specific detection models formulating one or more scored bounding boxes around the detected vehicle;   selecting a bounding box indicative of a specific vehicle type from the one or more scored bounding boxes based on calculated confidence scores and taking into account the image content and the image characteristics;   accessing a cluster configuration setting defining that bounding boxes are to be divided into a specified number of color clusters;   dividing the contents of the selected bounding box into a plurality of variably sized color clusters within the bounding box based on color differentiation between pixels within the bounding box and in accordance with the cluster configuration setting;   for each variably sized color cluster included in the plurality of variably sized color clusters:
 matching a color to the variably sized color cluster based on color values of pixels contained in the color cluster; and 
 determining a corresponding percentage of the bounding box area represented by the variably sized color cluster; and 
   deriving a vehicle color from the matched colors and corresponding percentages.   
     
     
         2 . The method of  claim 1 , wherein matching a color to the variably sized color cluster based on color values of pixels contained in the color cluster comprises calculating a mean color value of pixels contained in color cluster. 
     
     
         3 . The method of  claim 1 , further comprising:
 receiving a request from an entity to detect vehicles of a specified color traveling on roadways within a geographic area;   computing a color confidence score that the derived vehicle color is the specified color based on: detecting the vehicle and the image characteristics; and   sending the vehicle detection, the derived vehicle color, and the computed confidence score to the entity in response to the request.   
     
     
         4 . The method of  claim 1 , wherein accessing a camera image comprises:
 receiving a video stream from a camera; and   accessing the camera image from the video stream.   
     
     
         5 . The method of  claim 1 , wherein accessing a camera image comprises:
 receiving a cached video stream from a storage device; and   accessing the camera image from the cached video stream.   
     
     
         6 . The method of  claim 1 , wherein determining one or more image characteristics comprises:
 detecting a darkness level in the camera image;   computing a contrast from a plurality of different locations in the camera image; and   calculating an entropy of the camera image.   
     
     
         7 . The method of  claim 6 , wherein detecting a vehicle in the camera image comprises detecting the vehicle based on at least two of: the darkness level, the contrast, or the entropy. 
     
     
         8 . The method of  claim 6 , wherein calculating an entropy comprises:
 computing a greyscale image histogram; and   computing pixel entropy values.   
     
     
         9 . The method of  claim 1 , wherein dividing the contents of the selected bounding box into a plurality of variably sized color clusters comprises using an unsupervised learning technique to derive the plurality of variably sized clusters. 
     
     
         10 . The method of  claim 1 , wherein dividing the contents of the selected bounding box into a plurality of variably sized color clusters within the bounding box comprises representing each pixel in the bounding box in a five dimensional space including a Red dimension, a Green dimension, a Blue dimension, an X dimension, and a Y dimension. 
     
     
         11 . A system comprising:
 a processor;   system memory coupled to the processor and storing instructions configured to cause the processor to:
 access a camera image of an unknown quality that was captured by a camera of an unknown configuration; 
 determine one or more image characteristics and image content of the camera image; 
 detect a vehicle in the camera image considering the one or more image characteristics; 
 use one or more vehicle type specific detection models formulating one or more scored bounding boxes around the detected vehicle; 
 select a bounding box indicative of a specific vehicle type from the one or more scored bounding boxes based on calculated confidence scores and taking into account the image content and the image characteristics; 
 access a cluster configuration setting defining that bounding boxes are to be divided into a specified number of color clusters; 
 divide the contents of the selected bounding box into a plurality of variably sized color clusters within the bounding box based on color differentiation between pixels within the bounding box and in accordance with the cluster configuration setting; 
 for each variably sized color cluster included in the plurality of variably sized color clusters:
 match a color to the variably sized color cluster based on color values of pixels contained in the color cluster; and 
 determine a corresponding percentage of the bounding box area represented by the variably sized color cluster; and 
 
 derive a vehicle color from the matched colors and corresponding percentages. 
   
     
     
         12 . The system of  claim 11 , wherein instructions configured to match a color to the variably sized color cluster based on color values of pixels contained in the color cluster comprise instructions configured to calculate a mean color value of pixels contained in color cluster. 
     
     
         13 . The system of  claim 11 , further comprising instructions configured to:
 receive a request from an entity to detect vehicles of a specified color traveling on roadways within a geographic area;   compute a color confidence score that the derived vehicle color is the specified color based on: detecting the vehicle and the image characteristics; and   send the vehicle detection, the derived vehicle color, and the computed confidence score to the entity in response to the request.   
     
     
         14 . The system of  claim 1 , wherein instructions configured to access a camera image comprise instructions configured to:
 receive a video stream from a camera; and   access the camera image from the video stream.   
     
     
         15 . The system of  claim 11 , wherein instructions configured to access a camera image comprise instructions configured to:
 receive a cached video stream from a storage device; and   access the camera image from the cached video stream.   
     
     
         16 . The system of  claim 11 , wherein instructions configured to determining one or more image characteristics comprises:
 detect a darkness level in the camera image;   compute a contrast from a plurality of different locations in the camera image; and   calculate an entropy of the camera image.   
     
     
         17 . The system of  claim 16 , wherein instructions configured to detect a vehicle in the camera image comprise instructions configured to detect the vehicle based on at least two of: the darkness level, the contrast, or the entropy. 
     
     
         18 . The system of  claim 16 , wherein instructions configured to calculate an entropy comprise instructions configured to:
 compute a greyscale image histogram; and   compute pixel entropy values.   
     
     
         19 . The system of  claim 1 , wherein instructions configured to divide the contents of the selected bounding box into a plurality of variably sized color clusters comprise instructions configured to derive the plurality of variably sized clusters using an unsupervised learning technique. 
     
     
         20 . The system of  claim 1 , wherein instructions configured to divide the contents of the selected bounding box into a plurality of variably sized color clusters within the bounding box comprise instructions configured to represent each pixel in the bounding box in a five dimensional space including a Red dimension, a Green dimension, a Blue dimension, an X dimension, and a Y dimension.