基于DCIWPSO在谷底边缘泡沫图像分割的应用*

矿物浮选过程中，为了预测矿物品位，需要提取大量泡沫图像特征参数,其中泡沫大小是十分重要的图像特征参数。图像分割就是把泡沫图像分割成若干气泡区域的处理技术。谷底边缘分割算法是泡沫分割中一种重要的算法，其中分割阈值是非常重要的量，标准粒子群算法对阈值计算容易陷入局部最优值，难以计算全局最优值，采用改进的粒子群算法，动态改变粒子群中的惯性权重值来得到适合边缘分割的阈值，达到了正确分割泡沫图像的目的。

Based on DCIWPSO in application of valley-edge detection froth image segmentation

In the process of mineral flotation, in order to predict the mineral grade, it needs to extract a large number of bubble image feature parameters, where bubble size is very important image characteristic parameter. Image segmentation is image processing technology that divides a bubble image into several bubble areas. The valley-edge detection segmentation algorithm is an important segmentation algorithm, which segmentation threshold is very important parameter. For standard particle swarm algorithm to calculate the threshold easily trapped into local optimal value, it is difficult to calculate the global optimum value, this paper improved particle swarm optimization, dynamic change value of the inertia weight in particle swarm optimization to achieve to be suitable for the edge split threshold, correctly achieve the purpose of froth image segmentation.