Study On Computer Vision Technology And Its Applications In Industry Fields

1. PrefaceAs a branch of computer science, the computer vision developed rapidly in past 30 years, ithas wide application in the fields of industry, agriculture, military affairs, national defense andso on. But from the viewpoint of progressing, many computer vision theories and algorithmshave need further to improve, its application fields are far from popularization yet. Thepotential ability of computer vision should be developed and utilized as soon as possible.2. Applications of Computer Vision Technology in Piston AssemblyDuring the process of piston package mounting to cylinder, the accidence of piston missedassembling,reversed assembling and mixed assembling often occur due to the operation fault .In order to ensure the quality of piston assembling and reduce the worker labor intensity, thepiston assembling recognition system based on computer vision technology has been developed,Figure 1 shows the schematic diagram of recognition system. At piston assembling station, anengine cylinder is placed on a pallet. After the four piston packages have respectively been putinto four bore holes by manipulator, the pallet is transported to the inspection station bytransmit apparatus. At first the bar code reader reads the type of the engine and the sortinformation of corresponding piston package, and pick up the arriving signal of the pallet.When the pallet reaches to the position where the four engine bore holes and the four industrialCCD exactly aligned, the hydraulic stop is raised up automatically, the pallet stops itsmovement, then the control system operates the industrial CCD to pick up the images of thepiston installed in cylinder bore hole and save the images in computer memory. Subsequently,the hydraulic stop falls down, the transmit apparatus transports the pallet and cylinder to thebuffer station. Simultaneously, according to the images acquired, computer inspects the faultsof the missed assembling, reversed assembling and mixed assembling of the pistons, on thesake of speaker's warning and LED displaying, the operator corrects the accidents bycorresponding assemble measure.In this paper, the histogram analysis method is used to recognize the accident of pistonmissed assembling, by comparing the grey scale peak values in histogram of four cylinder borehole images. The template matching method is used to recognize the accident of pistonreversed assembling by comparing the coordinates of the maximum similar matching oncylinder hole images. Following two different methods are provided for recognizing theaccident of piston mixed assembling: the first method is to analyze the characteristic imagesof the valve pits which are located on the top surfaces of the pistons to determine whether fourpistons assembled in one cylinder are of the same type;the second method is to recognize thecharacters image printed on the top surfaces of the pistons to determine whether the fourpistons assembled in one cylinder are of the same type, to compared with the first one, thesecond method possesses following advantages : more rapid recognition speed, higherefficiency, better universality and more wide application fields.The piston assembly recognition system based on image recognition developed in thispaper has been applied to engine assembly in an automobile factory since Oct. 2002. Theapplication effects show that in the case of ensuring productive tempo, the system canaccurately recognize 5 different type pistons simultaneously, it can remove the accident hiddentrouble, ensure the quality of piston assembling and reduce the operator labor intensity.3. Applications of Computer Vision Technology in Auto Positioning System of EccentricBungholeEccentric bunghole barrels are most often used for liquor storage and transportation. Onliquor perfusion line, the empty eccentric bunghole barrels move by the transfer belt to theperfusion station, perfusion equipments can not be aligned with center of eccentric bungholebarrel because of the bunghole random distribution along its circular direction. At present,many plants of domestic and abroad often adopt hand-operated alignment method to aim thegun at the bunghole, which leads to low production rate,especially during perfusing toxicityliquor, such as sodium cyanide, hydrofluoric acid and so on, the perfusing operation becomesvery dangerous operation. At the base of researching the perfusing equipments ready-made, aneccentric bunghole auto positioning system based on computer vision is developed in this paper,which can make the workers prevent from poisons danger, improve automatization andmechanization level of perfusion operation.Figure 2 shows the schematic diagram of eccentric bunghole auto positioning system basedon computer vision. When the eccentric bunghole barrel moves with the transfer belt to theperfusion station, the computer system catches the coming signal of the empty eccentricbunghole barrel by a photoelectric switch, and controls the CCD camera hung over the barrel tocollect an image of the barrel top surface,which will be transferred to A/D and imageacquisition card where the image is decoded, then store in computer memory through PCI bus.According to the barrel top surface image, the center coordinates of the bunghole can becalculated, then the computer transfer the signal of the center coordinates of the bunghole to thestep motor control system which drives the perfusion gun align with the center of the eccentricbunghole.In order to meet the requirement of automatic liquor perfusing, the calibration method ofCCD, which developed in this paper will be performed automatically, the whole process of thecamera calibration involves following six steps: target image segmentation, label of thereference circle array, to extract the center of the reference circle array, calculation of cameraparameters by least-squares method, etc. The world coordinates of center of eccentric bungholecan be deduced from its corresponding image coordinates and cameral calibration coefficients.to be detectedFig2 eccentric bunghole computer vision auto positioning system based on computer vision4. Study of Some Fundamental Image Processing AlgorithmsFor the sake of improving the detect precision and calculation efficiency of computervision recognition (auto positioning) system, some relative fundamental algorithms werediscussed. An improved rapid median filter is developed utilizing the relationship between thefront and back filter windows;to compare with the traditional median filter, the improved rapidalgorithm has a higher computing efficiency for it can prevent all the data in current window toparticipate in sorting calculation. This paper presents a new improved Hough transform, whichutilize the edge contour characteristic of the circle image to delete many straight line and noiseelements, and refuses invalid accumulation of image information, to compare with thetraditional Hough transform, the parameter field is of two dimensions while that of traditionalHough transform is three, it can lead to decrease memory consumption, reduce the computationcapacity, increase the algorithm efficiency. A new CCD calibration method based onleast-squares has been developed in this paper, the target for calibration is constructed by a setof reference circle array, at first to measure the world coordinates of the target referencecircle centers, then to determine their image coordinates and calculate the CCD cameraparameters by least-square method. The world coordinates of eccentric bunghole can bededuced from its corresponding image coordinates and camera calibration coefficients. Theexperimental results show that the method has a high calibration precision and convenient forimplementation.5. Development of Device Driver Program for Computer Vision Recognition(Positioning) systemBased on the analysis of Windows interruption management system, the Vireo Software'sVtools D programming tool and VC++ language are used for developing the device driverprogram for computer vision recognition (positioning) system. When an accident occurs indevice driver,, the system will generate an interrupt requisition automatically and call theinterrupt service routine to handle the interrupt event immediately.In order to fully utilize the computer resources, the data output function of the parallel portcontrolled by timer interrupt service routine is used for generating numerical signals which candrive the step motor to make the perfusing gun align with the center of the eccentric bunghole.6. Summary(1) In order to improve the inspect precision and calculation efficiency of computervision recognition (auto positioning) system, some relative fundamental algorithms arediscussed in detail.(2) A set of engine piston assembling recognition system is developed by imagerecognition and character recognition technology. The accidents of piston missed assembling,reversed assembling and mixed assembling are successfully avoided, and the system has beenapplied to an automobile factory on an engine assembly line since Oct. 2002.(3) An auto positioning system of the eccentric bunghole is developed based oncomputer vision technology, which can accomplish the predetermined aim of autopositioning of the eccentric bunghole. Now the prototype equipment of above auto positionsystem has successfully trial-production.