Moving vehicle detection and analysis using low resolution remote sensing imagery

Abstract

A method comprises accessing a plurality of images of a substantially same geographical area, each image of the plurality of images captured at a separate time from each other using a separate imaging sensor, each imaging sensor capturing information corresponding to a different spectral band. The method further comprises identifying, for the images, a set of blobs, each blob comprising a plurality of adjacent pixels, wherein each image of the plurality of images includes a blob of the set of blobs, and wherein each blob in the set of blobs has a location in each image that differs from a location of other blobs in the set of blobs. The method further comprises generating a score indicating a likelihood that the set of blobs correspond to a moving object, storing an indication that the set of blobs correspond to the moving object responsive to the generated score exceeding a threshold.

What is claimed is: 
     
         1 . A method, comprising:
 accessing a plurality of images of a substantially same geographical area, each image of the plurality of images captured at a separate time from each other using a separate imaging sensor, each imaging sensor capturing information corresponding to a different spectral band;   identifying, for the plurality of images, a set of blobs, each blob comprising a plurality of adjacent pixels, wherein each image of the plurality of images includes a blob of the set of blobs, and wherein each blob in the set of blobs has a location in each image that differs from a location of other blobs in the set of blobs;   generating a score indicating a likelihood that the set of blobs correspond to a moving object; and   storing an indication that the set of blobs correspond to the moving object responsive to the generated score exceeding a threshold.   
     
     
         2 . The method of  claim 1 , wherein a blob in each image is identified by:
 determining candidate pixel locations of each image that have pixel attribute values that are at least a threshold difference from pixel attribute values of the same pixel locations in each of the other images in the plurality of images; and   identifying pixels of the candidate pixel locations that are adjacent to each other as the blob.   
     
     
         3 . The method of  claim 1 , wherein a blob in each image is identified by:
 merging adjacent regions of pixels in the image having pixels with pixel attribute values that have a lowest difference amongst adjacent regions of pixels in the image; and   identifying a merged adjacent region of pixels as a blob.   
     
     
         4 . The method of  claim 1 , wherein a blob in each image is identified by:
 applying a sliding window to each image, the sliding window being a rectangular pixel region which is moved across the image; and   identifying a blob at a location of the sliding window responsive to the sliding window being moved to the location of the image which has an average pixel attribute value of pixels bounded by the sliding window that is at least a threshold difference from average pixel attribute values of pixels at the same location in other images of the plurality of images.   
     
     
         5 . The method of  claim 1 , wherein each imaging sensor captures light in at least one of a red visible light spectral band, green visible light spectral band, and blue visible light spectral band. 
     
     
         6 . The method of  claim 5 , wherein the image sensor that captures light in the red visible light spectral captures one of the plurality of images at a first time, wherein the image sensor that captures light in the green visible light spectral captures one of the plurality of images at a second time, wherein the image sensor that captures light in the blue visible light spectral captures one of the plurality of images at a third time, and wherein a first interval of time between the first time and the second time differs from a second interval of time between the second time and the third time. 
     
     
         7 . The method of  claim 1 , further comprising:
 determining a direction of the moving object based on a difference between the location of each blob with other blobs of the set of blobs.   
     
     
         8 . The method of  claim 1 , further comprising:
 detecting paths in the plurality of images using edge detection, the paths being regions within the plurality of images; and   wherein identifying the set of blobs further comprises identifying each blob responsive to the location of each blob being within a region of each image detected as a path.   
     
     
         9 . A system, comprising:
 an aerial imaging device with a plurality of image sensors, each image sensor of the plurality of image sensors capturing an image of a plurality of images at a separate time from each other, each imaging sensor capturing information corresponding to a different spectral band, the plurality of images of a substantially same geographical area;
 an analysis system configured to: 
 access the plurality of images of a same geographical area; 
 identify, for the plurality of images, a set of blobs, each blob comprising a plurality of adjacent pixels, wherein each image of the plurality of images includes a blob of the set of blobs, and wherein each blob in the set of blobs has a location in each image that differs from a location of other blobs in the set of blobs; 
 generate a score indicating a likelihood that the set of blobs correspond to a moving object; and 
 store an indication that the set of blobs correspond to the moving object responsive to the generated score exceeding a threshold. 
   
     
     
         10 . The system of  claim 9 , wherein a blob in each image is identified by:
 determining candidate pixel locations of each image that have pixel attribute values that are at least a threshold difference from pixel attribute values of the same pixel locations in each of the other images in the plurality of images; and   identifying pixels of the candidate pixel locations that are adjacent to each other as the blob.   
     
     
         12 . The system of  claim 9 , wherein a blob in each image is identified by:
 merging adjacent regions of pixels in the image having pixels with pixel attribute values that have a lowest difference amongst adjacent regions of pixels in the image; and   identifying a merged adjacent region of pixels as a blob.   
     
     
         13 . The system of  claim 9 , wherein a blob in each image is identified by:
 applying a sliding window to each image, the sliding window being a rectangular pixel region which is moved across the image; and   identifying a blob at a location of the sliding window responsive to the sliding window being moved to the location of the image which has an average pixel attribute value of pixels bounded by the sliding window that is at least a threshold difference from average pixel attribute values of pixels at the same location in other images of the plurality of images.   
     
     
         14 . The system of  claim 9 , wherein each imaging sensor captures light in at least one of a red visible light spectral band, green visible light spectral band, and blue visible light spectral band. 
     
     
         15 . The system of  claim 9 , wherein the analysis system is further configured to:
 determine a direction of the moving object based on a difference between the location of each blob with other blobs of the set of blobs.   
     
     
         16 . The system of  claim 9 , wherein the analysis system is further configured to:
 detect paths in the plurality of images using edge detection, the paths being regions within the plurality of images; and   identify each blob of the set of blobs responsive to the location of each blob being within a region of each image detected as a path.   
     
     
         17 . A non-transitory computer readable storage medium, comprising computer-readable instructions that when executed by a processor causes the processor to:
 access a plurality of images of a substantially same geographical area, each image of the plurality of images captured at a separate time from each other using a separate imaging sensor, each imaging sensor capturing information corresponding to a different spectral band;   identify, for the plurality of images, a set of blobs, each blob comprising a plurality of adjacent pixels, wherein each image of the plurality of images includes a blob of the set of blobs, and wherein each blob in the set of blobs has a location in each image that differs from a location of other blobs in the set of blobs;   generate a score indicating a likelihood that the set of blobs correspond to a moving object; and   store an indication that the set of blobs correspond to the moving object responsive to the generated score exceeding a threshold.   
     
     
         18 . The non-transitory computer readable storage medium of  claim 17 , wherein a blob in each image is identified by:
 determining candidate pixel locations of each image that have pixel attribute values that are at least a threshold difference from pixel attribute values of the same pixel locations in each of the other images in the plurality of images; and   identifying pixels of the candidate pixel locations that are adjacent to each other as the blob.   
     
     
         19 . The non-transitory computer readable storage medium of  claim 17 , wherein a blob in each image is identified by:
 merging adjacent regions of pixels in the image having pixels with pixel attribute values that have a lowest difference amongst adjacent regions of pixels in the image; and   identifying a merged adjacent region of pixels as a blob.   
     
     
         20 . The non-transitory computer readable storage medium of  claim 17 , wherein a blob in each image is identified by:
 applying a sliding window to each image, the sliding window being a rectangular pixel region which is moved across the image; and   identifying a blob at a location of the sliding window responsive to the sliding window being moved to the location of the image which has an average pixel attribute value of pixels bounded by the sliding window that is at least a threshold difference from average pixel attribute values of pixels at the same location in other images of the plurality of images.