CAJ | 학술논문

计算机图象处理技术可广泛应用各种图象测试、模式识别和计算机视觉等领域，而图象边界检测是计算机图象处理的最基本和最低层的步骤之一。由于噪声的干扰和图象光照不均匀等因素的影响，目前的图象边界检则方法还不能有效地检测出各种不同模式的边界。本文提出了一种基于遗传算法和灰度梯度算子的图象边界检测法，通过对样本图象的训练，设计对训练样本模式最优的灰度梯度算子，增加了对噪声的抗干扰能力，并且使被检测出的边界更准确。
계산궤도상처리기술가엄범응용각충도상측시、모식식별화계산궤시각등영역，이도상변계검측시계산궤도상처리적최기본화최저층적보취지일。유우조성적간우화도상광조불균균등인소적영향，목전적도상변계검칙방법환불능유효지검측출각충불동모식적변계。본문제출료일충기우유전산법화회도제도산자적도상변계검측법，통과대양본도상적훈련，설계대훈련양본모식최우적회도제도산자，증가료대조성적항간우능력，병차사피검측출적변계경준학。
Computer image processing technology has been widely used in various image test,pattern recognition and computer vision,and edge detection is one of the fundamental steps on the lowest level of image processing.Due to the affection of the noise and unbalanced light source,no method can detect the edges of all modals.There fore, an approach based on gray level’s gradient operators and genetic algorithm was proposed in this paper.This approach can optimize the gray level’s gradient operators by genetic algorithm,and the fitness function of the genetic algorithms is the least square error criterion used in the procedure of supervised training.The proposed method can obtain the optimal edge detector for the image modal of training set.The optimized gradient operators present good performance in lower SNR and make the detected edges more precise.