双目立体视觉技术在天车机器人吊运系统中的应用解析.doc

摘　　要 point spread function，PSF）进行恢复的，恢复精度有限，而在曝光时间内运动形式不变造成的模糊，才能用同一PSF进行的恢复，振动导致成像过程中产生像移是空间移变的，因此，传统的恢复方法受到很大限制，本文提出了应用逐行法恢复空间移变降质图像；同时又根据曝光时间和振动周期的关系，分别论述了低频和高频振动模糊图像的恢复方法。再次，以振动模糊复原图像为基础，应用不变矩神经网络的模糊辨识方法识别吊钩，实验结果表明对复原图像直接计算不变矩，不同目标图像的不变矩数值分离性较差，识别精度低；经Sobel边缘特征提取后再计算图像的不变矩，不变矩在数值上分离性较好，能有效的提高目标识别精度。最后，以轴承作为目标干扰样本，证明了该识别方法的有效性。 在实验设计和研究的过程中，主要从整体上研究了振动模糊图像的复原方法。同时，模糊图像复原后应用不变矩和神经网络识别方法，有效的提高了吊钩的识别精度，结果表明，不变矩神经网络的识别方法，能满足天车机器人视觉控制要求，有效提高了天车作业的安全可靠性。 图 幅；表 个；参 篇。 关键词： 分类号：TP391.4 Abstract Binocular stereo vision technology?application?in?crane?transport system?in crane?robot,?which greatly improves the?reliability of the overhead traveling crane?hoisting?system.?Even through the?structure optimization?design is reasonable,?but?in the process of?lifting goods,?mechanical vibration?to the image acquisition?of fuzzy?phenomenon?still exists,?the vibration?usually?limitsthe resolution of the image,?brought great security risks?to the productionprocess.?In order to effectively reduce?the occurrence of safety accidents,?puts forward?recovery method,?complicated?space variant?vibration?image quality?at the same time,?and the?invariant moments and?neural network?identification ofeffective combination of?hook,?improves the reliability?of crane. In order to?better recover the?vibration?blurred image,?minimize?or suppressingvibration fuzzy,?improve image quality,?specially?designed and built a?single mass vibration?model with known?fuzzy?image acquisition?experiment platform.First,?analyzed the process of?degraded?image?vibration,?established the?3Dstructure?model of the vibration?of the experimental platform.?Secondly,traditional?restore?space variant blurring?image?is the use of?the image?of the point spread function?(point?spread,?function,?PSF)?recovery,?recovery Precision?Co.,?and?movement in?the exposure time in?form of?change due to the?fuzzy?and?recovery?can be carried out?with the same?PSF,?vibration causedthe imaging process?of image motion?is?space?shift,?therefore