Robust Perceptual Color Identification

Abstract

Systems and methods of robust perceptual color identification are disclosed. The methods include a multilevel analysis for determining the robust perceptual color of an object based on observed colors. This multilevel analysis can include a pixel level, a frame level, and/or a sequence level. The determination may make use of color drift matrices and trained functions such as statistical probability functions. The color drift tables and function training are based on training data generated by observing objects of known robust perceptual color in a variety of circumstances. Embodiments of the invention are applicable to the identification and tracking of objects, for example, in a surveillance video system.

1 . A system comprising: 
 means for acquiring a sequence of images including an object;    means for determining a robust perceptual color of the object using at least two of a pixel level analysis, a frame level analysis and a sequence level analysis.    
   
   
       2 . The system of  claim 1 , further including means for training a statistical classification function to identify a robust perceptual color from an observed color.  
   
   
       3 . The system of  claim 1 , further including means for representing color drift.  
   
   
       4 . The system of  claim 1 , further including means for determining a distance between a set of observed color data generated using an object of unknown robust perceptual color and a set of color data generated using an object of known robust perceptual color.  
   
   
       5 . The system of  claim 1 , further including means for tracking the object from a field of view of a first image sensor to a field of view of a second image sensor.  
   
   
       6 . The system of  claim 1 , further including means for reacquiring the object within the field of view of a first image sensor after the object has been acquired and lost.  
   
   
       7 . A method comprising: 
 (a) obtaining an image including an object of known true color;    (b) generating data by identifying pixels that are representative of the object within the image, the identified pixels being associated with observed colors, resulting from the known true color and being representative of observed colors to which the known true color can drift;    (c) repeating steps (a) and (b) for a variety of true colors; and    (d) aggregating the data generated in each repetition of step (b) to determine a color drift matrix.    
   
   
       8 . The method of  claim 7 , wherein the color drift matrix includes one or more values representative of a probability that a particular true color will drift to a particular observed color.  
   
   
       9 . The method of  claim 7 , wherein step (c) is performed under a variety of observation conditions.  
   
   
       10 . The method of  claim 7 , further including using the color drift matrix to determine the true color of another object.  
   
   
       11 . A method of identifying an object in a digital image, the method comprising: 
 collecting a first image using a first camera;    determining the true color of the object within the first image using a sequence level analysis; and    comparing the determined true color with a reference color in order to identify the object by color.    
   
   
       12 . The method of  claim 11 , further including obtaining the reference color in a query.  
   
   
       13 . The method of  claim 11 , further including obtaining the reference color by collecting a second image.  
   
   
       14 . The method of  claim 13 , wherein collecting the second image is accomplished using a second camera.  
   
   
       15 . The method of  claim 11 , further including tracking the object based on the comparison between the determined true color and the reference color.  
   
   
       16 . The method of  claim 11 , wherein determining the true color of the object within the first image further includes using a frame level analysis.  
   
   
       17 . A color drift matrix stored on a computer readable medium and produced using the method of  claim 7 .  
   
   
       18 . The color drift matrix of  claim 17 , wherein the color drift matrix includes a plurality of possible color drifts for the color red.  
   
   
       19 . The color drift matrix of  claim 17 , wherein the color drift matrix includes at least a color drift for each of the colors blue, yellow, and green.  
   
   
       20 . The color drift matrix of  claim 17 , where the step (b) is repeated using a plurality of separate cameras.