Accurate segmentation of vessels on computed tomography (CT) images is of great importance for preoperative planning of medical image segmentation tasks. However, manually labeling the mask of vessels is laborious and time-consuming, and the labeling results of different clinicians are prone to inconsistencies. Hence, developing an automatic segmentation algorithm for vessels on CT images has attracted a lot of attention from researchers. This study aims to provide a fully automatic and robust semantic segmentation algorithm for hepatic vessels, guiding subsequent preoperative planning of liver surgery. A structure with a dual-stream encoder combining convolution and transformer block is proposed to extract local features and long-distance spatial information, thereby extracting anatomical information of hepatic vessels, avoiding misdivisions of adjacent peripheral vessels. Besides, an edge-enhanced module based on Sobel filters in both horizontal and vertical directions to compute gradient maps is proposed to extract edge information, and a boundary inject module is introduced for highlighting inter-class edges while retaining intro-class consistency information. In addition, the annotations of the public dataset is revised and the hepatic veins and portal veins are distinguished. After that, the erosion-dilation mechanism is used to split the small vessels with its trunks. Our method is compared with state-of-the-art methods qualitatively and quantitatively and outperforms the comparing methods on the revised dataset.