Abstract:

A local feature description algorithm for sensitive target recognition-oriented in high-resolution  remote images, which is called the normalized pixel distribution histogram local descriptor (NPDHLD), is proposed. First object edges are extracted. The “log-polar” coordinate is established by using every edge points as the coordinate origin. Normalize every pixel value, the local descriptor is constructed by capturing the distribution of the object edge pixel points which are situated beyond the current origin point. The sensitive objects are described with the proposed local  description algorithm to build a sensitive objects feature database. The local features extracting from the objects to be recognized are matched with the ones in the sensitive object database under a “twostep matching” strategy. Object recognition is completed after matching.