基于形态学优化处理的标记符分水岭矿石图像分割

Image Segmentation of Watershed Ore Based on Morphological Optimization

针对不同形状、颜色的堆积矿石在图像采集时出现粘连重叠、边缘模糊等问题,提出一种形态学优化处理的标记符分水岭算法对矿石图像进行特征识别与边界分割。首先采用双边滤波进行去噪处理,然后通过Canny边缘检测算子梯度化,得到梯度幅值图像;其次采用形态学优化处理,重构堆积矿石的几何特征信息,分割矿石颗粒边缘;最后将图像进行可视化处理,获得彩色矿石颗粒分割图像。在形态学优化处理过程中,通过阈值分割数来寻找结构元素最佳参数,再通过分割率来确定最优结构元素。结果表明,该方法比传统标记符水岭算法分割更精准,且不同结构元素及其参数的变化对矿石分割效果影响较大,分割差异显著,当矩形结构元素长10、宽5时,其分割率最高,为95.68%,分割效果最佳。

In order to solve the problem of adhesion, overlap and edge blurring in the image collection of piled ore with different shapes and colors, a marker watershed algorithm based on morphological optimization was proposed to carry out feature recognition and boundary segmentation of ore images. Firstly, bilateral filtering was used for denoising, and Canny edge detection operator was gradient to get gradient image. Secondly, the morphological optimization was used to reconstruct the geometric feature information of the accumulated ore and to segment the grain edge of the ore. Finally, the image was visualized to obtain the color ore particle segmentation image. In the process of morphological optimization, the optimal parameters of structural elements were searched by threshold segmentation number, and the optimal structural elements were determined by segmentation rate. The results show that this method is more accurate than the traditional marker water-ridge algorithm. The changes of different structural elements and their parameters have a great impact on the ore segmentation effect with significant segmentation differences. When the rectangular structural elements are of 10 in length and 5 in width, the segmentation rate reaches the highest with the value of 95.68%, and the segmentation effect is the best.