【作者】 王廷军；

【导师】 贾瑞清；

【作者基本信息】 中国矿业大学（北京） ， 机械设计及理论， 2011， 博士

【摘要】 在分析煤矿井下环境条件的基础上,设计了一种四履带移动矿山搜寻机器人及其视觉系统,研究了机器人视觉煤矿井下识别的关键技术,基于形状特征、纹理特征、颜色特征的识别技术,模式识别系统以及井下矿工图像的决策判别方法。在详细分析井下矿工特征的基础上,研究矿工目标的不变矩特征,在模糊和仿射变换下的矩不变性,实现矿工目标在变形和模糊情况下的特征提取。对矿工的一般不变矩、仿射不变矩和模糊仿射不变矩,以及图像发生平移、旋转、尺度变换、仿射变换、模糊等进行实验研究,并对不变矩误差进行了分析,结果表明不变矩可以作为矿工目标识别依据。对未遮挡矿工提出不变矩特征识别方法,对部分遮挡矿工提出基于工装特征识别方法,给出矿工帽局部特征识别方法、工靴局部特征识别方法。探讨了基于人手部特征的识别、井下灯光特征及其识别问题。研究了井下矿工识别的决策判别方法,通过实验研究均值近似法构造不变矩原始分类器,提出了综合判别分类器。通过对比实验,研究了井下其他物体对矿工识别的影响。更多还原

【Abstract】 This paper studies a mobile miner searching robot with four tracks and designs its vision system. It researches key technologies of robot vision recognition; including recognition technologies according as shape feature, texture feature and color feature, pattern recognition system and decision method of miner’s image. This paper studies the moment invariants characteristic of miner person’s shape target, and pays more attention to invariance of moment on affine transformation and blur transformation based on analysis of miner’s features. This paper realizes feature extract of miner person’s shape target on distortion and blur image. Experimental researches of miner’s general moment invariants, affine moment invariants, blur moment invariants, blur-affine moment invariants and condition of image change in translation, revolving, scale anisotropy, affine transformation and blur transformation are done. And it also discusses errors of moment invariants. The results indicate that moment invariants may use as recognizing miner’s target. It points out moment invariants characteristic for no sheltering miners. And it points out recognition method based on frock characteristic for partly sheltering miners. Miner’s cap and boot part characteristic recognition are put forward. It also discusses recognition method based on a person hand characteristic and lighting. It studies design-making method for recognition of miner. It constructs moment invariant originality classfy by experiment mean. And it put forwards synthetize distinguish classifier. It does a contrast experiment about other objects in coal mine underground.更多还原