为了对彩色图像进行快速有效的分割,提出了一种用于分割彩色图像的多尺度形态学算法。该算法首先用基于张量梯度的彩色分水岭算法来得到初始分割结果,即局部水平集连通区域,并综合考虑了面积和色彩计算区域间的相似性,构造了区域间的RAG(region ad jacency graph)和NNG(nearest ne ighbor nraph),用于后续形态学处理;接着,基于HSV空间中的色彩全序关系,定义了彩色形态算子;然后采用顶点塌缩算法实现的彩色形态学开闭算子生成了所需的非线性尺度空间;最后,利用图像中的极值点与物体间的对应关系,逐级提取图像中包含的物体来得到分层级的表示,并用区域在不同尺度下熵的变化来决定尺度树的构成,从而完成了彩色图像的分割。试验结果表明,该算法不仅具有出色的形状保持能力,而且可提高计算效率。
It has been a scientific and technological problem in the field of microelectronics for several decades that the electrical method is used to measure the peak junction temperature of power transistors. Based on the excessive thermotaxis effect of low current, a novel electrical measurement method of the peak junction temperature is presented in this paper. The method is called the thermal spectrum analysis method of transistors, simply designated TSA （thermal spectrum analysis method）. Unlike the common method which uses a single measuring current, TSA uses multi-step currents to measure temperature-sensitive parameters. Based on the excessive thermotaxis effect of low current and the sub-transistor parallel model, the peak junction temperature and non-uniform property of junction temperature distribution are analyzed successfully.
The /-V-（T） characteristic curves of p-n junctions with the forward voltage as the independent variable, the logarithm of forward current as the dependent variable, and the junction temperature as the parameter, almost converge at one point in the first quadrant. The voltage corresponding with the convergence point nearly equals the bandgap of the semiconductor material. This convergence point can be used to obtain the I-V characteristic curve at any temperature.
