An analysis of the effect of gaussian error in object recognition
Karen Sarachik · DSpace@MIT (Massachusetts Institute of Technology) · 1994
In model based recognition the problem is to locate an instance of one or several known objects in an image. The problem is compounded in real images by the presence of clutter (features not arising from the model), occlusion (absence in the image of features belonging to the model), and sensor error (displacement of features from their actual location). Since the locations of image features are used to hypothesize the object's pose in the image, these errors can lead to "false negatives", failures to recognize the presence of an object in the image, and "false positives", in which the algorithm incorrectly identifies an occurrence of the object when in fact there is none. This may happen if a set of features not arising from the object are located such that together they "look like" the object being sought. The probability of either of these events occurring is affected by parameters within the recognition algorithm, which are almost always chosen in an ad-hoc fashion. The implication...