Abstract: A novel framework of object verification is proposed.Integrating by the prior knowledge of object,object verification is turned into a conditional probability of a set of observable cues available conditioned upon an object hypothesis.Then features of short line segments are extracted,and a new algorithm which verifies objects of building in complex scenes is presented using observation of line segments,.Tests on natural image sets showed that the proposed algorithm could effectively eliminate false objects,and satisfy the request of processing natural images on time performance.