Research On Wheel Hub Character Recognition Based On Linear CCD Scanning Image

Wheel hubs are mostly castings,which adopt mixed flow production mode,so they need to be classified and stored for later processing.With the improvement of technology level and production efficiency of modern manufacturing industry,it has become an urgent need to design a visual real-time recognition system for automatic recognition and classification of wheel hub models.There is a series of embossed characters on the side of the hub,composed of letters and Numbers,which is the model code of the hub.The characters are printed directly on the side of the hub through the metal mold.There is no color difference between the characters and the hub.Therefore,this paper has done the following research on the hub imprint character recognition technology.In this paper,aiming at the characteristics of hub embossing characters as curved convex characters,FPGA(Field-Programmable Gate Array,FPGA)driven linear array CCD(Charge-coupled Device,CCD)is adopted to collect hub character images,and appropriate hardware is selected according to the requirements of this topic.In order to improve the contrast between the hub embossed character and the background,after analyzing the effect of illumination on the embossed character image,a low Angle illumination scheme using a small linear array LED lamp was designed.The collected character image of hub embossing has many noises,poor contrast between characters and background,and slanting character.In this paper,aiming at the problem of image noise,four commonly used filtering methods are tried,and bilateral filtering with edge preserving effect is adopted through comparative analysis.In order to improve the contrast between the characters and the background,a Retinex algorithm is applied to enhance the filtered image.The multi-scale Retinex algorithm is modified for the phenomenon that it is easy to produce halo artifacts when the light is uneven.A character image with a clear edge and a good contrast is obtained.The character skew will affect the positioning effect and reduce the character recognition rate.In this paper,Radon transform is used to calculate the character image tilt Angle,and bilinear interpolation method is used to assign the gray level of the rotated image.The current character recognition algorithm is mainly for binary image,in the character image threshold segmentation using the Otsu algorithm and Bernsen algorithm,and the characters of fracture in view of the Otsu algorithm and Bernsen algorithm more noise disturbance,the Bernsen algorithm is improved,the Bernsen algorithm combined with a gaussian filter,remove most noise at the same time,the continuous and smooth the binarization of image characters.In order to improve the speed of character recognition,it is necessary to accurately locate the character region.In this paper,a hub character region location algorithm based on mathematical morphology and projection method is adopted.In character recognition,a character recognition method based on template matching and character structure features is proposed to reduce character recognition errors.C#.NET was used to design the interface of hub character recognition system.Experiments show that the system has a good recognition effect on hub characters with different tilting angles.