The matching method of high resolution palmprint is one of the research focus in Biometrics Recognition area.lt includes key problems as palmprint alignment, similarity computing ,optimization of matching process and so on. This paper proposes a novel single-ring-four-sectors local structure and a new method of datum mark election, as well as a minutiae-based palmprint matching algorithm. The paper also presents an effective similarity computing approach and a multi-level palmprint matching approach to improve matching efficiency and accuracy. We use 44 latent palmprints and 10,400 archive palmprints which were from criminal investigation field to do experimental evaluation. As the experiments shows, the proportion for the correct results ranked first in the indentification of candidate reach to 78%, and the proportion for top 20 reach to 86%. The matching time for single palmprints is only 86ms.Compared with other existing, our approach has been a mark improvement both in identification accuracy and recognition speed.