CAJ | 학술논문

针对压力钢管内壁表面缺陷图像在直方图呈偏态或重尾时图像分割效果差的问题，在传统的图像分割思想的基础上，提出了一种基于灰度中值和灰度平均离差分生成图像最佳分割阈值主变量的改进算法。实验结果表明，改进后的算法对于能够稳定有效地分割压力钢管内壁正常部位与缺陷部位，并将缺陷部位的细节特征增强显示，为后续分析缺陷原因提供了很好的形态学依据；显示了其在化工设备的完整性检测技术中的应用意义。
침대압력강관내벽표면결함도상재직방도정편태혹중미시도상분할효과차적문제，재전통적도상분할사상적기출상，제출료일충기우회도중치화회도평균리차분생성도상최가분할역치주변량적개진산법。실험결과표명，개진후적산법대우능구은정유효지분할압력강관내벽정상부위여결함부위，병장결함부위적세절특정증강현시，위후속분석결함원인제공료흔호적형태학의거；현시료기재화공설비적완정성검측기술중적응용의의。
In allusion to the poor results of segmentation in steel defects detection when the histogram of images of defects was skew and heavy-tailed, an improved method for image segmentation based on class median was proposed according to the conventional segmentation algorithms:minimum error thresholding that employed the sample mean value, the convergence process of optimization was also discussed based on particle swarm optimization. The testing results indicate the modified method can highlight the details of defects’ morphology, which can provide sound grounds for cause analysis; meanwhile, the improved algorithm is applicable in the detection of the integrity of the chemical equipments.