Segmentation And Recognition Of 3D Objects In Automatic Navigation

This paper presents a segmentation and recognition method of 3D objects in automatic navigation. In the first part of this paper, a segmentation method based on nautical scene is proposed, which is composed of image preprocess, ROI detection, subarea process and object detection. In addition, an applied visual resolution calculation method is presented to control the simplification of original 3D model. Then a novel clustering method is introduced to merge the aspect graphs ulteriorly after viewpoint space partition to the simplified model. Finally, the segmented object will be sent to compare with the aspect graphs of the 3D model, using an improved method of Fourier descriptor. A set of experiments based on ship models are designed and implemented, and the results demonstrate the effectiveness of the method proposed in this paper.