一种动态文稿边缘自适应检测与纠偏方法
|
摘要
摄像头在对文稿进行实时边缘检测与纠偏过程中存在复杂背景干扰,会导致无法准确完成文稿的边缘检测与纠偏的情况。为此,提出了一种动态的文稿边缘自适应检测与纠偏方法。在形态学变换与Canny边缘检测的基础上,根据文稿轮廓特征,结合多边形逼近、最小矩形包裹与面积筛选的方法对复杂背景下的文稿进行边缘检测与纠偏,并通过引入图像压缩比方法,使纠偏前后的文稿布局保持一致。实验结果表明:该方法在单一或复杂背景下均能对文稿边缘进行准确检测与纠偏,解决了上述情况下复杂背景干扰的问题;完成1次准确检测与纠偏的平均耗时为0. 609 2 s与0. 663 8 s,满足实时检测与纠偏的要求。
The camera has complex background interference in the process of real-time edge detection and rectification of the document,which may lead to the inaccurate edge detection and correction of the document. For this reason,a dynamic method for adaptive detection and correction of the document edge is proposed. Based on the morphological transformation and Canny edge detection,this method combines the polygon approximation,the minimum rectangle wrapping and the area screening method to detect and correct the edge of the document in complex background based on the contour feature of the document. In addition,by introducing the image compression ratio,this method can make the layout of the document before and after the correction correct. The experiments show that the method can accurately detect and correct the edge of the document in a single or complex background,and solve the above-mentioned complex background interference problem. In addition,the average time taken to complete an accurate detection and correction is 0. 6092 s. 0. 6638 s,meeting the requirements of real-time detection and correction.
The camera has complex background interference in the process of real-time edge detection and rectification of the document,which may lead to the inaccurate edge detection and correction of the document. For this reason,a dynamic method for adaptive detection and correction of the document edge is proposed. Based on the morphological transformation and Canny edge detection,this method combines the polygon approximation,the minimum rectangle wrapping and the area screening method to detect and correct the edge of the document in complex background based on the contour feature of the document. In addition,by introducing the image compression ratio,this method can make the layout of the document before and after the correction correct. The experiments show that the method can accurately detect and correct the edge of the document in a single or complex background,and solve the above-mentioned complex background interference problem. In addition,the average time taken to complete an accurate detection and correction is 0. 6092 s. 0. 6638 s,meeting the requirements of real-time detection and correction.
引文
[1]王咸锋,黄妙燕.基于Open CV实现文档物体的纠偏方法[J].电脑知识与技术:学术交流,2016,12(8X):87-89.
[2]王恒友,余沾,张长伦,等.基于低秩矩阵分解的批量扫描文档图像纠偏[J].计算机工程与应用,2018,54(17):175-179.
[3] ZHANG Shengnan. Document skew detection inbinary images through pre-estimation of skew angle and Hough transform[J]. ICIC Express Letters,2015:2185-2192.
[4] KAO Chih-hong. Skew detection of document images using line structural information[C]//International Conference on Information Technology and Applications. Australia:IEEE,2005:704-709.
[5]苑擎飏,宋洪波.一种改进的Hough变换图像纠偏算法[J].沈阳理工大学学报,2012,31(2):9-11.
[6]王文豪,姜明新,赵文东.基于Canny算子改进的边缘检测方法[J].中国科技论文,2017,12(8):910-915.
[7]侯涛,张志腾.改进Canny算子在列车车轮踏面损伤检测中的应用[J].铁道科学与工程学报,2018,15(8):2107-2112.
[8]颜丽,谷学静.基于Canny-AAM的人脸特征定位算法[J].湖南科技大学学报(自然科学版),2018,22(2):66-71.
[9]杨先凤,吴媛媛,赵玲.基于Canny改进算子的油管裂纹检测算法[J].计算机工程与设计,2018,39(3):798-803.
[10]徐骏骅.基于边缘检测与模式识别的车脸识别算法[J].控制工程,2018,25(2):357-361.
[11]王慧.基于图像边缘检测的红枣分级方法[J].农机化研究,2018,10:231-234.
[12]王忠华,王超.联合帧间差分法和边缘检测的运动目标检测算法[J].南昌大学学报(理科版),2017,41(1):42-46.
[13]王燕凤,冯海林,杜晓晨,等.一种对年轮图像双边滤波增强的树龄测量方法[J].林业工程学报,2017,2(5):109-114.
[14]高春霞,董宝田,王爱丽.基于人头颜色空间和轮廓信息的行人检测方法研究[J].交通运输系统工程与信息,2015,15(4):70-77,84.
[15]黄山,李众,黄蒙蒙.利用数据场和欧式距离的图像边缘提取[J].计算机工程与科学,2016,38(11):2297-2302.
[2]王恒友,余沾,张长伦,等.基于低秩矩阵分解的批量扫描文档图像纠偏[J].计算机工程与应用,2018,54(17):175-179.
[3] ZHANG Shengnan. Document skew detection inbinary images through pre-estimation of skew angle and Hough transform[J]. ICIC Express Letters,2015:2185-2192.
[4] KAO Chih-hong. Skew detection of document images using line structural information[C]//International Conference on Information Technology and Applications. Australia:IEEE,2005:704-709.
[5]苑擎飏,宋洪波.一种改进的Hough变换图像纠偏算法[J].沈阳理工大学学报,2012,31(2):9-11.
[6]王文豪,姜明新,赵文东.基于Canny算子改进的边缘检测方法[J].中国科技论文,2017,12(8):910-915.
[7]侯涛,张志腾.改进Canny算子在列车车轮踏面损伤检测中的应用[J].铁道科学与工程学报,2018,15(8):2107-2112.
[8]颜丽,谷学静.基于Canny-AAM的人脸特征定位算法[J].湖南科技大学学报(自然科学版),2018,22(2):66-71.
[9]杨先凤,吴媛媛,赵玲.基于Canny改进算子的油管裂纹检测算法[J].计算机工程与设计,2018,39(3):798-803.
[10]徐骏骅.基于边缘检测与模式识别的车脸识别算法[J].控制工程,2018,25(2):357-361.
[11]王慧.基于图像边缘检测的红枣分级方法[J].农机化研究,2018,10:231-234.
[12]王忠华,王超.联合帧间差分法和边缘检测的运动目标检测算法[J].南昌大学学报(理科版),2017,41(1):42-46.
[13]王燕凤,冯海林,杜晓晨,等.一种对年轮图像双边滤波增强的树龄测量方法[J].林业工程学报,2017,2(5):109-114.
[14]高春霞,董宝田,王爱丽.基于人头颜色空间和轮廓信息的行人检测方法研究[J].交通运输系统工程与信息,2015,15(4):70-77,84.
[15]黄山,李众,黄蒙蒙.利用数据场和欧式距离的图像边缘提取[J].计算机工程与科学,2016,38(11):2297-2302.