摘要
基于传统的Harris算法,提出了一种改进的Harris算法用来进行角点检测。此算法通过改变原来的梯度算子,提高了算子的检测敏感度。采用一个新的角点响应函数,避免了人为设定参数。在图像拼接过程中,通过卷积神经网络(Convolution neural network,NCC)对图像进行配准,利用稳健的RANSAC算法剔除误匹配,最终图像融合。实验结果表明,该算法提高了图像拼接的准确性,减少了图像拼接所耗费的时间,提升了图像拼接效率,具有良好的实用性。
An improved Harris algorithm is proposed for comer detection based on traditional Harris algo- rithm. The original gradient operator is changed firstly so as to improve the detection sensitivity of opera- tor. A new angular point response function is adopted to avoid the artificial parameter setting. The image is registered by NCC during process of mosaic. The false matching will be gotten grid off by the robust RANSAC method so as to realize image fusion. Experimental results show that the algorithm will improve the accuracy of image stitching, reduce the time consumed by image stitching, and improve the efficiency of image stitching, so as to possess good practicability.
作者
霍东旭
朱朦
任洪娥
HUO Dongxu;ZHU Meng;REN Honge(School of Information and Computer Science, Northeast Forestry University, Harbin 150040, China;Forestry Intelligent Equipment Engineering Research Center of Heilongjiang Province, Harbin 150040, China)
出处
《黑龙江大学自然科学学报》
CAS
2018年第2期212-217,共6页
Journal of Natural Science of Heilongjiang University
基金
中央高校基本科研业务费专项基金资助项目(2572017PZ10)
