基于动态轮廓曲线的运动图像特征识别研究  

Research on Moving Image Feature Recognition Based on Dynamic Contour Curve

作　　者：程彭阳子[1] 刘萍[1] CHENG Peng-yang-zi;LIU Ping(Henan Normal University,Xinxiang Henan 453007,China)

机构地区：[1]河南师范大学,河南新乡453007

出　　处：《计算机仿真》2023年第7期163-167,共5页Computer Simulation

摘　　要：针对运动图像帧在识别过程中存在误差大,背景、纹理、边缘对运动目标轮廓边缘影响严重等问题,提出基于动态轮廓曲线的运动图像特征识别方法。首先采用不同大小的抑制强度值对图像轮廓的候选边缘进行处理,将有重叠的像素点加入组合中,通过组合结果,提高图像正确分类的概率,得出图像的轮廓边缘。然后对运动图像中相邻两帧间采用尺度变换,对得出新图像的像素值进行小波分解尺度逼近处理,得出收敛补偿,以Newmark迭代算法为基础,对预处理后的像素点进行提取。最后利用修复优化算法对像素点进行修复,通过拓扑梯度最小重构路径耦合修复方法对运动图像损坏区域进行识别,确定运动图像的拓扑梯度以及轮廓修复的关键点,当峰值信噪比为最大时,运动图像的轮廓像素点已修复完成。实验结果表明,采用所提方法对运动图像帧进行跟踪,帧的匹配度较高,跟踪性能良好,而且能够对图像的细节进行修复,修复效果显著。Aiming at the problems of large error in moving image frame recognition and serious influence of background,texture and edge on moving target contour edge,a moving image feature recognition method based on dynamic contour curve is proposed.Firstly,the candidate edges of the image contour were processed with different suppression intensity values,and the overlapping pixels were added to the combination.Through the combination results,the probability of correct image classification was improved,and the contour edge of the image was obtained.Then,the scale transform was used between two adjacent frames in the moving image,the pixel values of the new image were processed by wavelet decomposition scale approximation,and the convergence compensation was obtained.Based on the iterative algorithm,the preprocessed pixels were extracted.Finally,the pixels were repaired by the repair optimization algorithm.The damaged area of the moving image was identified by the coupling repair method of the minimum Reconstruction Path of the topological gradient,and the topological gradient of the moving image and the key points of contour repair were determined.When the peak signal-to-noise ratio is the maximum,the contour pixels of the moving image have been repaired.The experimental results show that using this method to track the moving image frame,the frame matching degree is high,the tracking performance is good,the details of the image can be repaired,and the repair effect is superior.

关 键 词：抑制强度 候选边缘 收敛补偿 拓扑梯度 峰值信噪比 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]