计算机应用与软件
計算機應用與軟件
계산궤응용여연건
COMPUTER APPLICATIONS AND SOFTWARE
2013年
10期
116-119
,共4页
徐琛%温振市%毛亚文%白瑞林
徐琛%溫振市%毛亞文%白瑞林
서침%온진시%모아문%백서림
轴承侧面缺陷检测%轴承定位%灰度变换%正态分布
軸承側麵缺陷檢測%軸承定位%灰度變換%正態分佈
축승측면결함검측%축승정위%회도변환%정태분포
Defect detection of the bearing side%Bearing positioning%Gray transform%Normal distribution
提出一种利用离线样本学习实现轴承外侧表面缺陷在线快速检测的方法。采用间接照射面光作为光源,通过CCD摄像头采集轴承外侧面图像。在离线情况下:定位轴承侧面的检测区域;分别拟合横纵向的灰度分布规律曲线;并以灰度变换后的样本图像作为检测时的依据。在线检测时:根据学习知识提取出待检测区域、将轴承图像变换成灰度分布均匀的图像;然后对图像进行动态阈值分割并作出决策判断。实验表明,提出的方法能有效地将被测图像变换成灰度均匀的图像;判别一张轴承图像平均时间为20 ms,准确率达98.2%以上,具有较高的实时性和准确性。
提齣一種利用離線樣本學習實現軸承外側錶麵缺陷在線快速檢測的方法。採用間接照射麵光作為光源,通過CCD攝像頭採集軸承外側麵圖像。在離線情況下:定位軸承側麵的檢測區域;分彆擬閤橫縱嚮的灰度分佈規律麯線;併以灰度變換後的樣本圖像作為檢測時的依據。在線檢測時:根據學習知識提取齣待檢測區域、將軸承圖像變換成灰度分佈均勻的圖像;然後對圖像進行動態閾值分割併作齣決策判斷。實驗錶明,提齣的方法能有效地將被測圖像變換成灰度均勻的圖像;判彆一張軸承圖像平均時間為20 ms,準確率達98.2%以上,具有較高的實時性和準確性。
제출일충이용리선양본학습실현축승외측표면결함재선쾌속검측적방법。채용간접조사면광작위광원,통과CCD섭상두채집축승외측면도상。재리선정황하:정위축승측면적검측구역;분별의합횡종향적회도분포규률곡선;병이회도변환후적양본도상작위검측시적의거。재선검측시:근거학습지식제취출대검측구역、장축승도상변환성회도분포균균적도상;연후대도상진행동태역치분할병작출결책판단。실험표명,제출적방법능유효지장피측도상변환성회도균균적도상;판별일장축승도상평균시간위20 ms,준학솔체98.2%이상,구유교고적실시성화준학성。
A method is proposed that utilises the offline sample learning to achieve online fast detection of the defects on outside surface of the bearings.The method uses indirect irradiation surface light as the light source,through CCD camera to capture the outside surface images of the bearing.In off-line situation,we locate the detection zone on side of the bearing,and fit the horizontal and vertical gray distribution rule curves respectively;then we use the gray-scale transformed sample image as the basis of detection.In on-line situation,we extract the region to be detected according to the knowledge learned,transform the image of bearing to the image with uniform gray distribution,then conduct dynamic threshold segmentation on the image and make the decision judgment.Experiments show that the proposed method can effectively transform the image to be detected to the image with uniform gray distribution,the average time for distinguishing a bearing image is about 20 ms,and the accuracy can be 98.2% and higher,so the method has high real-time performance and accuracy.