基于煤位检测的多尺度特征匹配算法

针对煤矿企业煤仓内煤位的检测，运用计算机视觉检测识别技术，提出了一种基于双目视觉W-SIFT图像特征匹配算法的非接触式煤仓煤位深度测量方法，将小波压缩编码和尺度不变特征变换算法相结合，去除双目图像中的冗余信息后，通过基于尺度因子变化的模版进行自适应调节，并对双目视图配准的煤位图片进行处理，构建多尺度空间从而提取煤位图片的特征点，然后对比左右煤位视图，运用双目视图双向配准的策略，进行深度匹配。最后依据W-SIFT匹配算法预先标定的左右图像得出视觉传感器的内外参数，代入坐标转换公式，准确确定煤仓煤位的真实深度值。实验结果表明将这一新算法应用到煤位检测中，与传统SIFT算法相比具有较明显的优势，实时性、稳定性好，取得了满意的效果。

Based on the multi-scale feature matching algorithm of the coal level detection method

Using computer vision technique to detect coal level of coal bin，a kind of non-contact depth measurement method of coal bunker was put forward based on binocular vision and W-SIFT image feature matching algorithm.Through the wavelet compression coding and scale invariant feature transform algorithm，remove all kinds of redundancy in image and keep important information.By adaptive regulating based on template of the changing scale factors and dealing with the coal bit image waiting for the binocular view registration，multi-scale spatial could be created for extracting feature points in coal level image.The left and right view is compared，and extraction feature points for in-depth matching based on the bidirectional registration strategy of binocular view so that we can more accurate restore the 3D information.According to the W-SIFT matching algorithm，prior calibration left and right image so that we can get internal and external parameters of visual sensor，then ，determine the coal bunker coal bit depth value based on the coordinate transformation formula.Using the new algorithm to detect coal level，The result is satisfy compared with the traditional SIFT algorithm.