机械零件加工过程中的位置识别模型

Position Recognition Model During Machining of Mechanical parts

首先本文基于未经轮廓提取的原始零件图像数据的数据,通过origin8软件做出工件扫描信号强度分布图(见图2)及零件信号分布图(见图3)。从图中可以看出各区域信号的强弱分布情况,以及工件扫描图中存在着外部噪音和各零件啮合处干扰因子。为了更好的提取零件轮廓,需对这些干扰项进行剔除,以减少对零件识别过程中存在的影响,从而达到精准识别零件位置的目的。然后本文通过统计像素点信号强度值及其个数(见图4),确定零件和外部噪音与背景两者的信号强度分界值为150,从而除去了背景中大部分干扰项。由于零件①与其他零件相连且图像灰度值相对弱于相连的零件,因此为了对零件尽可能的无损分割,本文接着对图像进行四次凹凸增强处理(见图7),拉大了各零件之间像素点灰度值的差距,接着通过fspecial滤波算子函数提取出了②②④零件与①号零件啮合处的齿轮形貌,然后将其覆盖于初次增强图中,通过matlab编程分割出了各零件的完整形貌图(见图11-12),最后运用DBSCAN聚类法对四个零件区域进行分割识别,并通过计算得到各零件的重心位置,运用Harris轮廓角点识别法对图像各角点进行识别,确定了轮廓较为突出的区域。对于不平整的表面上的零件,本文通过COMSOL软件做出三维图像示意图(见图17),随后对其进行灰度化处理。由于该模型为理想化模型,且零件之间不相连,因此可省去凹凸增强、啮合处识别、分割等图像处理步骤,仅通过角点识别、轮廓识别等方法对其定位,得到的效果较好。

Firstly,based on the data of the original part image without contour extraction,we obtain the workpiece scanning signal intensity distribution map(Fig.2)and the part signal distribution map(Fig.3)through Origin8 software.From those figures,we can acquire the distribution of the strength of the signals in each area,and the external noise in the workpiece scan and the interference factor at the meshing of the parts.In order to better extract the contour of the part,these interference items need to be eliminated to reduce the influence on the part identification process,so as to accurately identify the position of the part.Then,by counting the intensity values of the pixel signals and their number(Fig.4),it is determined that the signal strength boundary value between the part and the external noise and the background is 150,so that most of the interference items in the background can be removed.Since the part①is connected to other parts,and the parts whose image gray value is relatively connected are weak,in order to achieve the lossless segmentation of the part as much as possible,the image is then subjected to four times of b ump enhancement processing(Fig.7),which enlarges each the difference in the gray value of the pixel between the parts.Then,the gear shape of the mesh between the②②④part and the①part is extracted by the Fspecial filter operator function,and then it is covered in the initial enhancement map.The complete topography of each part is obtained by MATLAB programming(Fig.11-12),finally using the DBSCAN clustering method to segment and identify the four parts.Later,we calculated the position of the center of gravity of each part by calculation,and used Harris contour point recognition method to identify the corner points of the image,and determined the area with prominent outline.For parts on uneven surfaces,this paper uses the COMSOL software to make a three-dimensional image schematic(Fig.17),and then we grayscale the schematic.Since the model is an idealized model and the parts are not connected,the processing steps such as bump enhancement,meshing recognition,and segmentation can be omitted.We can only use corner recognition,contour recognition and other methods to locate it can also get excellent results.