论文部分内容阅读
全面分析并建立了复杂物体透视投影的视觉事件的确定方法和计算公式,视觉事件分为两类七种.文中提出用变半径高斯球表示三维透视投影空间,通过测试点对透视投影视觉事件的计算,可大大降低计算的复杂性;并且形态图的表示采用分子结构图表示,从而解决了透视投影空间的形态图的求解和表示.论文用实例??棱柱与圆锥相贯验证了该算法的有效性.“,”Current research on aspect graph has been limited, to our best knowledge, to simple objects. We now extend it to complex objects encountered in applications. We present an algorithm for computing aspect graph of complex objects under perspective projection. We give a detailed analysis of the visual events for complex objects and the formulas to compute for parceling planes. Fig.1 shows three kinds of local singularities and Fig.2, four kinds of multi local singularities. The computation of aspect graph, discussed in detail in section 4, is our main contribution. We disperse the view space into a set of points, and record the value of each visual event by 0 and 1 because parceling plane always divides the view space into two parts. By merging points whose values of visual events are equal, we can obtain an exact parcellation of view space into regions. Our solution reduces significantly computing complexity, and allows aspect graph of complex objects under perspective projection to be represented as a molecular structure graph as shown in Fig.3. In Fig.3, each node represents a region. Neighboring nodes, such as nodes 1 and 17, differ by one in value of visual events. Fig.4 shows aspects obtained for a complex object containing a prism and a cone with our algorithm on a PC386 in Turbo C ++ . Results show that our algorithm works with satisfactory stability.