摘要
为了能部分解决基于视觉的石块识别问题,首先利用SUSAN算子对图像进行分割后利用边界跟踪算法进行滤波,确定出原始图像中可能存在石块的区域,对原始图像中该区域利用矩不变法进行分割以缩小可能的石块区域,然后利用Sobel算子对原图像中该区域进行处理,得到该区域梯度图像并对此进行分割和边界跟踪滤波,最后利用投影法确定出原始图像中的石块位置。试验结果表明该方法具有一定的环境适应性和较好的实时性。
In order to realize a stone detection based on machine vision, SUSAN operator was used to segment an original image. The areas where stones may exist could be obtained by using edge track filter in the segmented image. The area in the original image could be dwindled by using a segmentation method based on invariable moments. A grads image could be obtained by using Sobel operator to enhance the edge of this area. The grads image was segmented and filtered by using edge track. The stone position in the original image could be acquired by coordinating projection for the grads image. Experiment results show that the proposed method is characterized by a definite adaptability to environment and real-timeness.
出处
《计算机应用》
CSCD
北大核心
2006年第1期114-116,119,共4页
journal of Computer Applications
