一种基于改进Sobel算子的苹果图像边缘提取算法的研究  

Research on an Apple Image Edge Extraction Algorithm Based on Improved Sobel Operator

作　　者：陈浩[1] 黄勋[1] 赵志明[1] CHEN Hao;HUANG Xun;ZHAO Zhiming(Shaanxi University of Science and Technology of Shaanxi Province,Xian,Shaanxi 710021,China)

机构地区：[1]陕西科技大学,陕西西安710021

出　　处：《天津农业科学》2020年第9期51-55,共5页Tianjin Agricultural Sciences

基　　金：陕西省重点研发项目(2020GY-174)。

摘　　要：苹果图像边缘提取是苹果分级过程中的重要一步,正确地提取出苹果边缘,可有效提高分级的正确率。针对传统边缘提取算法边缘定位精度低、抗噪能力弱、边缘点模糊等缺点,提出一种基于多方向改进Sobel算子和自适应阈值的苹果边缘提取算法。将传统Sobel算子方向模板增加至8个来提高边缘定位精度;通过改进小波系数邻域进行方差估计,来获得自适应最佳阈值;结合阈值和改进Sobel算法获得苹果边缘图像。通过苹果边缘检测实验表明:该自适应算法处理时间相比于传统Sobel边缘检测降低了30.4%,分级正确率从93%提高至97.5%,表明该算法在去噪的同时能够较好保留边缘信息,有利于提高后续苹果分级检测的精度。Apple image edge extraction is an important step in the apple grading process,correctly extracting apple edges can effectively improve the accuracy of grading.Aiming at the shortcomings of traditional edge extraction algorithms such as low edge positioning accuracy,weak anti-noise ability,and fuzzy edge points,an apple edge extraction algorithm based on multi-directional improved Sobel operator and adaptive threshold was proposed.The traditional Sobel operator direction template was increased to 8 to improve the edge positioning accuracy;the adaptive optimal threshold was obtained by improving the wavelet coefficient neighborhood to estimate the variance;combining the threshold and the improved Sobel algorithm to obtain the apple edge image.Apple edge detection experiments showed that the processing time of this adaptive algorithm was reduced by 30.4%compared with traditional Sobel edge detection,and the classification accuracy rate was increased from 93%to 97.5%,indicating that the algorithm could better retain edge information while denoising,which helped to improve the accuracy of subsequent apple grading detection.

关 键 词：改进Sobel 边缘提取 最佳阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]