煤尘特性参数的去噪机理与图像识别研究

The coal dust measurement technologies which had the poor stabilization, low accuracy and being difficult to save data don’t consider the coal dust physical characteristics in the coal mine safety environment. The application of the coal dust image recognition method was constrained. This topic is proposed to apply image processing techniques for studying coal particle size distribution and concentration classification measurement problem with a large number of trials.The change regularity of coal dust characteristic parameters is thoroughly studied.And the impacts of coal dust particle size, particle size distribution information characteristic are analyzed. The mathematical models of fractional order variational image denoising and coal dust image recognition are developed.The effects of regularization parameter and fractional order derivative on the coal dust image recognition parameters are discussed. The relationship between image parameters and coal dust characteristic parameters are determined. And the inner mechanisms of coal dust imagery analysis characteristic parameters are revealed. Based on the coal dust physical properties imagery feature analysis models are set up.The relational expressions of image parameters and coal dust characteristics are deduced. The parameters of coal dust particle size, size distribution and coal dust concentration are calculated. According to the parameters model the basic morphological characteristics of particles are determined. The experiment results validate that the proposed model and theoretical analysis are valid and feasible.The obtained research achievements will provide theoretical and experimental bases for the design of explosion of coal dust concentration measurement.

现有煤尘测量技术存在测量稳定性差，精度低，数据难以保存,且未考虑煤尘物理特性等问题，这些制约着煤尘图像识别方法在煤矿安全环境中的应用。本课题拟运用图像处理技术对煤尘粒度分布及浓度分类测量问题进行大量试验，深入研究煤尘特性参数的变化规律，并分析煤尘粒度、煤尘粒度分布信息特性；建立具有分数阶变分图像去噪和煤尘图像识别的数学模型，探讨正则化参数，分数阶导数等算法参数对煤尘图像识别的影响，并确定该图像参数与煤尘特性参数之间的关系，揭示煤尘特性参数图像分析的内在机理；基于煤尘物理特性，建立图像特性分析模型，推导出其图像参数和煤尘特性的关系表达式，获取煤尘粒度、粒度分布和煤尘浓度等参数，根据参数模型确定颗粒群的基本形态特征的判定依据，实验验证模型及理论分析的正确性和可行性。研究成果可为煤尘爆炸浓度检测设计提供理论基础和实验依据.

煤矿井下的煤尘爆炸危险具有一定的普遍性，故采用煤尘特性参数量化指标确保煤矿安全的问题至关重要。本课题运用图像处理技术对煤尘粒度分布及浓度分类测量问题进行大量试验，深入研究煤尘特性参数的变化规律，构建深度学习神经网络模型分析煤尘特性参数图像内在机理，进一步明确煤尘颗粒群基本形态特征。探讨煤尘特征参数对煤尘图像识别的影响，并确定该图像参数与煤尘特性参数之间的关系，获取煤尘粒度、粒度分布和煤尘浓度等参数.研究结果表明，所提出的改进网络模型对不同形状及粒径的颗粒类别具有较优的学习效果，其学习性能指标表现良好。可有效增强煤尘特征的学习能力，大幅缩短训练时间，并精确获取颗粒特征更多细节信息。实验验证模型及理论分析的正确性和可行性。研究成果可为煤尘爆炸浓度检测设计提供理论基础和实验依据.