煤矿复杂环境视频拼接技术  

Stitching technology of coal mine video with complex environment

作　　者：厉丹[1,2] 钱建生[1] 刘增宝 杨澎 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008 [2]中国矿业大学徐海学院,江苏徐州221008 [3]兖矿集团东滩煤矿,山东济宁273500

出　　处：《煤炭学报》2011年第5期878-884,共7页Journal of China Coal Society

基　　金：国家高技术研究发展计划(863)重点资助项目(2008AA062200);江苏省产学研联合创新资金资助项目(BY2009114)

摘　　要：针对煤矿井下噪声大、光照不均等复杂环境传统拼接算法并不适用的问题,提出新的拼接算法。算法采用拼接帧抽取方式提取部分关键帧进行视频拼接,根据相位相关冲激函数能量高效检测重叠区域并对乱序帧排序,将统计分析中样本主成分分析引入SIFT特征向量降维中,用互不相关的新样本信息来描述原样本信息,降低了特征向量生成时间,通过结合多频带提升小波,定义融合规则并选择合适的融合策略,实现平滑拼接。实验表明,此算法提高了匹配效率,无需人工排序,解决了复杂环境容易误匹配、接缝明显等现象,拼接效果平滑自然,适合煤矿环境。Proposed a new stitching algorithm to suit for noisy,uneven illumination complex environment of coal mine which traditional stitching algorithm does not apply.Key frames were Abstract to stitch video images.Overlapping regions and frames in confusion were checked and ordered by phase correlation impulse function energy.Statistical sample principal component analysis was brought to SIFT feature vector dimension reduction with new samples of unrelated information to describe the original sample information,so feature vector generation time could be reduced.Multi-band lifting wavelet transform scheme was used to achieve smooth stitching by defining the fusion rules and selecting the right fusion strategy.Experiments show that the new algorithm improves the matching efficiency under no manual sorting,solves the problems of false matching easily and stitching juncture obviously phenomena.Smooth and natural stitching result is suit for coal mine environment.

关 键 词：煤矿 拼接 主成分分析 相位相关 提升小波 

分 类 号：TD672[矿业工程—矿山机电]