Manual pre-operative planning is time consuming and depends mainly on the clinical experience of surgeons. It is helpful for automatic pre-operative planning and robotic surgery to explore the systematic methodology for operative environment analysis and understanding. Since shape segmentation, structure extraction and representation are fundamental to shape understanding, this paper aims to explore the semantic segmentation algorithm and the structure representation of the generalized body cavity. Our general strategy for abstract shape segmentation is the explicit boundary method. The cutting contours combining critical loops and local feature contours are constructed to slice the input mesh. To achieve reasonable critical loops, we first defined a real function that is invariant to translation, rotation and scaling transformation. Then, an efficient segmentation algorithm was proposed to obtain approving semantic segmentation. In addition, we presented a structure representation that was expected to be applied to the path planning of surgery. Finally, the articular cavity of human knee joint was taken as an example to verify the proposed approach.