【作者】 邓洪超；

【导师】 马文星；

【作者基本信息】 吉林大学 ， 机械设计及理论， 2007， 博士

【摘要】 本文针对犁式除雪车雪铲避障自动控制问题,在分析、验证积雪路面障碍探地雷达(Ground Penetrating Radar,简称GPR)探测可行性的基础上,对GPR探测回波信号的实时滤噪、障碍的实时成像、障碍信号的实时提取及雪铲避障的自动控制问题进行研究,揭示了积雪路面的GPR回波信号特性及障碍信号特性,解决了犁式除雪车除雪作业的障碍探测与雪铲避障自动控制问题。本文的主要研究及相关结论包括:1、在全面分析路面积雪物理特性、路面材料特性、积雪路面常见障碍的基础上,依据电磁波传播理论,提出以电磁波方法探测积雪路面障碍;经理论分析及实验验证,遵循电磁波传播规律的GPR探测技术应用于除雪车除雪作业的积雪路面探测及雪铲自动控制方面完全可行。2、分析、研究GPR回波信号的一维实时分析、处理技术。应用小波理论的Mallat算法(多辩分析算法)分解和重构GPR回波信号;计算机仿真及相关实验结果验证了Daubechies小波对积雪路面GPR回波信号的一维滤噪处理具有有效性、实时性。3、依据地震波运动学,采用Stolt偏移算法,实现GPR探测积雪路面障碍的实时合成孔径成像;应用小波理论的二元多辩分析算法,对障碍图像进行分解与重构,实现障碍图像的实时增强,提高了除雪车操作人员监控的可视性。4、建立积雪路面障碍GPR探测信息正演模型,采用相关系数法由GPR回波信号中提取路面障碍信号作为雪铲避障的自动控制信号;分析计算GPR信号处理速度、雪铲避障动作速度与除雪车除雪作业速度,对三者的时间匹配特性进行研究;计算机仿真分析及相关实验结果,验证了自动控制信号与时间匹配的正确性。5、对人工神经网络应用于路面积雪厚度探测进行探讨性研究,认为在认知的基础上,通过GPR回波信号,人工神经网络经大量学习过程,能够识别路面积雪厚度。更多还原

【Abstract】 The snowfull in winter always brings direct serious influence to the road traffic in the north areas. Technology for cleaning out RSSI (road-surface’s snow cover and ice, abbr. RSSI) is developed to clean out RSSI in time and to reduce the influence of snowfull. Because the mechanical method has the marked virtues to clean out RSSI, Japan has investigated the plough snow-shovel for cleaning out RSSI since 1950s. Up to now, the various equipments to clean out snow have been studied in Japan, Canada, Ameica, Russia, and so on, which are obtained widespread application. The technique of the mechanical method has been studied since 1970s in our country and made mighty advances up to the present. Meanwhile it is limited by the conditions such as design means and the correlative leave of technology; there are a great difference between China and other developed countries. And the effectual detecting for SCRS(the snow cover road-surface, abbr. SCRS ) has not been actualized when the sowplough cleaning out snow cover. So the snow-shovel has not automatically avoided the road-surface’s obstacles. Thus the snowplough or the road-surface is harmed easily. For this reason, it is important to develop the special snowplough which is suit for our country roads, can effectively detect road-surface’s obstacles,can keep a lookout over road-surface’s obstacles and automatically avoid road-surface’s obstacles. This paper studys on GPR(Ground Penetrating Radar, abbr GPR) detecting technology for SCRS and the signal processing and the automatic controlling technology of the snowplough’s snow-shovel from the practical requirement.Analyse and Choice of Detecting SCRS MethodThe snow-shovel configuration of the snowplough and the detecting SCRS method and the automatic control method of the snow-shovel and the capabilities of the snowplough to clean out snow cover depend on at the first hand the physical characteristics of the road-surface’s snow cover and the physical characteristics of the the road-surface’s materials and the road-surface’s obstacles. The physical characteristics are different with the circumstances. For this reason, they are the base of the studying on snowplough configuration and its control system to systemically analyse the physical characteristics, and to master the properties of the road-surface’s materials, and to investigate the familiar road-surface’s obstacles. In the non-contact detecting methods, the feasibility of sound-wave detecting SCRS and the electromagnetic wave detecting method was discussed in this paper, and the electromagnetic wave detecting method was deemed fitting for work of the snoeplough cleaning out snow cover. In the electromagnetic wave detecting methods, GPR detecting equipment is small in the size, easy for carrying about and so on. Form the theory of the electromagnetic wave radiating and the experiments of GPR detecting the snow cover, GPR is used to detect SCRS. This paper analysed and processed GPR signals to show the synchronous images of SCRS’s obstacles and extract the control sings for avoiding the obstacles when the snowplough cleaning out snow cover.Linear Analysing and Processing of GPR SignalsThe reflected signals of GPR detecting SCRS contains various interferential signals from the working environment. The interferential signals are filtrated by analysing and processing GPR reflected signals to retain the usable signals. The analysing and processing technology of the signals must be real-time and reliably to satisfy detecting SCRS during the snowplough working. The analysing and processing methods of the numerical signal contain the Fourier Transform (FT) and the Discrete Fourier Transform (DFT) and the Fast Fourier Transform (FFT) and the Short-Time Fourier Transform (STFT) and the Wigner-Ville Distribution (WVD) and the Wavelet Transform and so on. The characteristics of these methods were analysed in this paper. The analyse shows that the Wavelet Transform is more suitable too analyse and process GPR reflected signals than other methods. The Mallat algorithm of the Wavelet Transform from the multiresolution analysis is used to decompose and reconstruct GPR reflected signals, so the interferential signals was effectively filtrated at the same time. The Daubechies wavelet of the Mallat algorithm was used to analyse and manage GPR reflected signals, and its validity and realtime are analysed by the computer simulation analysis. The validity and realtime for analysing and processing GPR reflected signals was tested by the experiments of GPR detecting SCRS.Planar Imaging of GPR Reflected SignalsFrom the reflected signals of GPR detecting, the planar image of SCRS’s obstacles is constructed for the visual during the snowplough operator to enhance the security of the snowplough. From the theory of the seismic wave radiating, the finite-differences time-domain algorithm and the Stolt migration algorithm and the Phase-Shift algorithm are the representative algorithm of the wave equation migration imaging algorithm. Their capabilitys of the operating speed and the wave speed adaptability and the full eliminating boundary influence and etcetera were discussed in the paper. The discussions shows that the Stolt migration imaging algorithm is suitable for processing the real-time imaging information. GPR synthetic aperture imaging algorithm from the Stolt migration imaging algorithm is able to process the real-time imaging of SCRS’s obstacles detected by GPR. But the image visualization is not satisfying. Based on the duality multiresolution analysis algorithm of the Wavelet Transform, GPR image from the Stolt migration imaging algorithm was decomposed and reconstructed. So the real-time image of SCRS’s obstacles detected by GPR was enhanced, and the visualization was enhanced also at the same time.Extracting Automatic Control Signal of Snow-shovel Obstacle-Avoidance and Analysing of Speed MatchFrom the linear analysing and processing of the reflected GPR signals, the obstacles signals was extracted in time by establishing the information model and forward modeling and using the correlation coefficient method. The extracted obstacles signals was used to control the snow-shovel automatic obstacle-avoidance when the snowplough cleaning out snow cover. The validity of the extracted obstacles signal was tested by the computer simulation analysis and the correlative experiments. The speed of the snowplough cleaning out snow cover and the analysing speed of GPR reflected signals and the response times of the snow-shovel automatic obstacle-avoidance form the match characteristic,and the match characteristic was analysed in the paper. From the analysis of the match characteristic, the paper discussed the connection between the setting distance of GPR antennas and the speed of the snowplough cleaning out snow cover.Study of Using Artificial Neural Network to Calculate Snow Cover ThicknessAlthough the snow-shovel automatic obstacle-avoidance was realized and the planar real-time image of SCRS’s obstacles was showed by GPR detecting technology of SCRS when the snowplough cleaning out snow cover, it is difficult to differentiate the stratum signal from GPR reflected signals because of the randomicity of physical characters of the snow cover and the material of the roads. In this paper, the artificial neural network was explored to differentiate the stratum signal from GPR reflected signals and to calculate the road-surface’s snow cover thickness from the plentiful learning and cognizing of the artificial neural network. So the accurate information for the control and watch would been provided to the snowplough when it was cleaning out the road-surface’s snow cover. The time delay point algorithm was adopted to calculate the thickness of the road-surface’snow cover. The capability of the artificial neural network to calculate the thickness of the road-surface’s snow cover was simulated by the computer analysis. The actual detection data of GPR detecting the road-surface’s snow cover was provided to the artificial neural network to calculate the snow cover thickness. Compared values of the snow cover thickness calculated by the artificial neural network with values of the snow cover thickness measured actually, it was found that the artificial neural network calculated values of the snow cover thickness is near the actually measured values of the snow cover thickness when the snow cover was enough thick, and the thickness error is great when the snow cover was thin. On the condition of the plentiful learning and cognizing, the artificial neural network can calculate the snow cover thickness by using GPR reflected signals from the academic analysis.更多还原