Capsule segmentation algorithm based on Wavelet Analysis and Constrained Gradient Search
In industrial production,the application of the online detection and recognition technology of capsule defects is more and more widely.The correct segmentation of the target capsule is essential for the feature extraction and accurate recognition.Because the difference between the oval transparent capsule and background is small and the image exists uneven illumination interference,common methods are difficult to segment target perfectly,a method of combining the single color space projection,wavelet decomposition and reconstruction of one-dimensional signal and constrained gradient search algorithm is proposed to realize the segmentation in the complex background,this method has good segmentation effects on transparent capsules and opaque capsules.The average correct segmentation rate of this method is 99%,the average segmentation accuracy is 99.44%,and the average deviation is only 0.56%.
Capsule segmentation Signal projection Wavelet decomposition and reconstruction Constrained gradient search
Shuang Dong Yu Wang Hao Zhang
College of Information Engineering,Wuhan University of Technology Key Laboratory of Fiber Optic Sensing Technology and Information Processing,Ministry of Education,Wu College of Management,Wuhan University of Technology
国际会议
重庆
英文
1061-1066
2017-03-25（万方平台首次上网日期，不代表论文的发表时间）