Method and system for processing a sequence of images to identify, track, and/or target an object on a body of water

摘要

Airborne tracking systems use sensors to track objects of interest. In order to track the objects of interests, the sensors need to be steered such that the object is kept, ideally, in the center of the sensors field of view. Automatic steering of optical sensors requires the generation of a track on an object of interest. When tracking boats on the water, current approaches to image processing may generate multiple detections on the object of interest. Embodiments of the present disclosure solve the track multiplicity problem by grouping tracks associated with the object of interest into a cluster and by estimating a most likely location of the object within the cluster of tracks. Based on the estimated location, embodiments of the present disclosure outputs a single track for the object. The single track is used by an automatic steering system to maintain a sensor aimed at the object of interest.

主权项

1. A method, executed by one or more processors, for processing a sequence of images to identify, track, and/or target an object on a body of water, the method comprising:
identifying at least one detection in an image of a series of images, wherein each detection is at least one of: an object on a body of water or at least a portion of a wake generated by the object on the body of water; associating the at least one detection with a single track of a set of tracks, the set of tracks being stored in a memory; creating at least one new track for each of the at least one detection that does not associate with any track of the set of tracks; classifying each pixel in the image of the series of images as either an object class or a water class based on a range of pixel values; for each pixel classified as an object class, grouping each pixel in the image of the series of images into at least one boat-wake cluster, wherein each boat-wake cluster includes pixels belonging to the object class that are adjacent to each other; associating each of the at least one boat-wake cluster with a subject track of the set of tracks; clustering tracks of the set of tracks that are associated with a common boat-wake cluster; assigning each cluster of tracks a unique global track identifier; and selecting a representative track of each cluster of tracks having a same unique global track identifier; the selected representative track being a first track in a ranked list of tracks, wherein a rank of a track is based on features that are indicative of a real object.