| At present,the measurement way of bulk material through dynamic weighing in the process of conveying is widely used in many fields,such as coal,mining,power,metallurgy,ports and food.Much attention has been paid to the dynamic measurement of bulk material for being related to various interests of people.It is an important research area to improve the measurement accuracy and intelligent,in order to meet the various requirements of users.Electronic belt scale and nuclear belt scale are the two main types of belt scale of the traditional dynamic measuring modes.The electronic belt scale belongs to the contact measuring method,its measurement accuracy depends not only on structure of carrier,mathematical weighing model and accuracy of sensors,but also on the inherent non removalble factors of the belt transportation system,such as installation location,vibration and operating conditions.The system measuring accuracy can only reach 0.5% at most when the sensors accuracy is 0.01%,therefore,in order to improve the accuracy of measurement,system calibration is often required in practical use of system.Now the existing calibration methods of measurement system mainly include haning weight calibration,chain weight calibration and material calibration.Material calibration is the most effective,it is also an electronic belt scale calibration method recommended by the national standard.But it is not suitable for any site,troublesome operation and waste manpower and material resources.One of the key problems lies in how to grasp the calibration time.At present calibration is made when the incorrect value is found,then the effect on measurement accuracy has been produced.In addition,some circumstances,such as the dynamic measurement of the field of stope of coal mines cannot use the electronic belt scale to measure due to the belt installation site and belt length changed frequently,so the nuclear belt scale is often used in this condition.Nuclear belt scale belongs to non-contact measurement method,and it needs to use radioactive source.The measuring principle determines that its actual measuring accuracy is lower than the electronic belt scale.Because of the needs of environment protection and safety,the application of nuclear belt scale is more and more limited.Laser belt scale and vision belt scale using new technology principle to realize the dynamic measurement of belt scale.These systems can run stably for a long time in a suitable working condition after calibration,but it needs a higher requirements for operating environment,so its practical application is restricted greatly too.In order to solve the problem of judging calibration time and improve the applicability of system,basing on the advantages of machine vision measurement,a new design scheme of dynamic measurement system for bulk materials based on machine vision is proposed.The system integrates volume measurement and weight measurement on the same system platform,which using the vision image to measure material volume and measure mass by sensors.Taking advantage of volume value and mass value,the bulk material density is computed.The run state of system and whether needs to calibrate or not can be determined by judging the density of bulk material.For some bulk materials,such as the dynamic measurement of coal which materials include coal and gangue,according to the bulk material density,the type of material can be distinguished,so the measurement accuracy of system can be further improved.Based on the above target of research,this paper mainly completed the following work:1.Select specific research method of machine visionAccording to the analysis of principle and application of machine vision methods,contrasting research difficulty and feasibility,this paper select monocular vision method based on line structured light to study dynamic weighing of bulk materials.2.Establish vision measuring platform(1)On the basis of detailed analysis of volume measurement model and system design constraints,this article compares the system structure with direct type of laser and oblique type of laser,designs a set of visual measurement device and develops the prototype.This measuring device choose laser as light source.It puts the image processor and laser projection device in the same closed metal cavity.In the cavity,the special relative position between camera and laser projection device is fixed.In order to measure targets with different scope,only needs to adjust the fixed position of metal reflector in regulating lever and relative angle between metal reflector and regulating lever.Changing the laser projection position through using metal reflector.This structure design decreases the number of devices,reduces the cost of manufacturing and inspection.Through angle adjusting the position of laser projection is simple and accurate.Equipment structure is compact which make it easy to install and maintain.Visual acquisition device with the protection of metal cavity,it is suitable for industrial site conditions.(2)The influencing factors on the quality of the image visual measuring system are discussed in detail,which mainly include the visual sensor and lens parameters selection,structure light power and wavelength of the light source,the features of measured material,such as types,shape,particle size,color,etc.This analysis for these influencing factors through theoretical analysis and practical experience.The analysis provides the basis and direction for the construction and research of measurement system.3.Research on image processing algorithms(1)Image preprocessingBased on the image pretreatment process,each step of pretreatment is discussed in detail.Image gray level test was carried out by using three methods: weighted mean method,mean value method and maximum value method.The test result indicates that:1)The gray value computed by weighted mean method is higher than the gray value computed by mean value method when the image includes green laser stripe,that is,image brightness which is processed by weighted mean method is bigger than it is processed by mean value method.2)The image brightness difference is very small between the image dealed by weighted mean method and the image dealed by mean value method when the laser stripe of image is red.3)Maximum method select max gray to realize gray level,so it's brightness degree is the biggest.Through the analysis of the characteristics of the target image,Otsu method is adopted to obtain the threshold,and the mean filter of 3x3 window template is used.Experiments prove that these dealing can achieve better effect for image pretreatment under advisibale light conditions.For the image with environment light too strong or too weak,in order to reduce the effect of ambient light,this paper put forward to a new image preprocessing method based on adaptive brightness image preprocessing method according to human visual sensitive principle and deduced the image preprocessing model of adaptive brightness.(2)Research on extraction algorithms of light stripe centerClassification and summary of the light stripe center extraction algorithms,analysis of the principles,advantages and disadvantages of these methods.Skeleton extraction is one of methods to extract the center of laser light stripe,so the skeleton extraction algorithm is analyzed in depth.Two kinds of main methods for the skeleton extraction of binary image are described,which are topological thinning,distance transform and the improvement research with the two methods.In general,the processing speed based on distance transform is faster than topological thinning.Therefore,this paper proposes a fast skeleton extraction algorithm base on gradient optimization of distance field.In order to improve dealing speed of method,checkboard distance is chosen as distance measure,and only computed the gradient of direction of 0°,45°,90°,135°,then selected the maximum gray value in the four orientations as the final gray value of current pixel.Through the gradient optimization,the dealing speed of algorithm can further improve and meet the real-time requirement.In this paper,the concept of distance threshod is defined.Based on the distance threshold,a new method is proposed to patch the break laser stripe.This method combines the linear interpolation with three B spline interpolation.Define distance threshold as T and the distance value as L between the two adjacent break sub-curves of light stripe,when L<T,we use linear interpolation method directly.In the contrast,when L?T,three B spline curve interpolation method is used.In the method of three B spline curve interpolation,the key is to obtain a reliable control points.The research object of this paper runs at a certain speed,therefore,we can get the control points through using adjacent two images and belt speed to compute.If the light stripe of the adjacent images is also broken at the same position,we use the linear interpolation for this broken curve directly.4.Research on system calibration methodThis paper presents a method named “Separate Calibration Method” for structure light measurement system with distortion through the analysis of measurement model and the summarization the characteristics of general calibration methods.The method of nonmetric calibration is adopted for system distortion,and the planar template method is used to solve camera interior and exterior parameters,the calibration of light plane is realized by the method basing on free moving plane checkboard.The method reduces the coupling between parameters,simplifies the calibration process and improves the calibration efficiency.(1)In this paper,the area between the straight line and the distortion curve of the straight line is chosen as the distortion measure to realize non-metric calibration.Based on the analysis of nonlinear optimization methods,a method that combined the genetic algorithm with LM algorithm is proposed to optimize the nonlinear distortion measure function.Firstly,using genetic algorithm to find a global solution,and putting the solution as the initial value of LM algorithm to realize local optimization,then getting the final optimal solution.This kind of optimization method has the advantages of global optimization of GA,and avoid GA losting in local iteration,so the efficiency of the algorithm is improved.Using RMSE to measure the effect of optimization algorithms,by comparison of test correction effect and time consuming with several methods proposed,effectiveness of the algorithm proposed by this paper is verified.(2)Applying the “Separate Calibration Method”,based on free moving checkboard,only need to collect two images with higher location and lower location with target,the light plane equation can be solved with the two images.Furthermore,no need to calibrate all parameters again when changed the positions of camera or light source.Compare the effect with the proposed calibration method and the method based on the invariance of cross ratio,use the two methods to calculate the distance with feature points.Regarding the calculating result of basing on the invariance of cross ratio as reference,and taking the difference measurement between the two methods as the error evaluation target,the max absolute error with distance computing result using the proposed method by this paper is 0.12 mm.The average relative error is 0.290%.The measurement results are satisfactory and meet the precision requirements.5.Design and implementation of coal dynamic measurement system based on machine vision.For the coal belt conveyor system,utilizing the key technology and methods,this paper has realized the design of a set of coal dynamic measurement system based on machine vision.The core component of system-Output Monitoring Unit is designed and the prototype of output monitoring unit and monitoring software platform is developed.Bulk material mass is obtained by weighing sensors and corresponding bulk material volume is received by vision measuring system.According to the formula of mass calculation,the bulk material density can be computed.So we can judge the state of the system and the opportunity of calibration.At the same time,the system realized the judging of coal and gangue on-line.So the measurement accuracy of the system is greatly improved.The monitoring software platform further enhances the interaction between human and computer,it can better meet the needs with diverse display of dynamic measuring. |
PDF Full Text Request