I am doing a project on possible methods to mitigate train-animal collisions. The idea is for a small buggy to survey the track in front of the train and report if it detects a big animal. I am using ultrasonic range finders to detect the preseence of a large animal; but want to confirm the type of animal using OpenCV. Relaying the video or an image back to the train driver for an ID is not an option. I cannot use colour as these need to work with night vision. Also because the animals are of all ages, size is not constant either. However, I only need to categorize animals into Elephant or Cattle, and they have very different profiles; except when looking from the rear. Any ideas as to how I should proceed?