Research On The Machine Vision And Intelligent Algorithm For Precision Manufacturing And Measurement

Machine vision plays an important role in the field of precision manufacturing andtesting due to its characteristics of high precision, rapidity, efficiency, non-contact and so onwhich is an important approach to enhance the level of equipment manufacturing industry.Real-time measurement system based on machine vision provides the technical support andmaterial preparation for intelligent manufacturing.The research background and significance of machine vision and computationalintelligence are expounded according to requirements of improving the equipmentmanufacturing industry in our country, and the current research foundation, conditions andadvantages. Research is mainly done on machine vision calibration, the flank inspection ofmicro drill, and eye-hand calibration for robot system, which provides theoretical andtechnical support for precision manufacturing and testing. The main research of the paperincludes as follows:1. The analysis and research on the physical model of the machine vision system havebeen carried out, and a neural network embedded rotational matrix and genetic algorithm isadopted to achieve the calibration of camera with non-linear model, where the forwardcalculation of the network is corresponding with the camera’s physical model so that theintrinsic and extrinsic parameters of camera are gotten from the stable weights of the networkwhen the network comes to the global optimal equilibrium position.2. By extracting the eigen-vector of the self-relative matrix corresponding to the smallesteigenvalue, the orthogonal-learning neural network with lateral inhibition is adopted to obtainthe transform relation of3D space information and2D image information for the machinevision system, which results in the binocular vision system calibration, as well as the3Dreconstruction.3. The neural network for testing micro drill is designed which is consistent with thefitting ellipse equation of margin projection, or linear equations of the main cutting edgeswhen the network comes to the global optimization position. The network‘s weights arenormalized to unit weight vector in every training, which is equal to the mutation operation ofgenetic algorithm. And an improved particle swarm algorithm is introduced in order to obtainthe global optimal solution.4. The particles motion trajectories of the particle swarm optimization algorithm areanalyzed from the longitudinal direction and transverse direction by introducing the evolution speed factor and aggregation degree factor. According to the features of the inertia factor ofparticle swarm optimization algorithm, and the fuzzy control rule table obtained from expertexperiences and inference is used to adjust the system s inertia weighting factor dynamically,which makes optimization algorithm to find the global optimal solution more accurately andquickly. In every iteration of particle swarm optimization algorithm, the position of theindividual particle is normalized to unit weight vector. Finally micro-drill feature curves fitting equations are gotten.5. When the robot s hand-eye system is calibrated, a hybrid neural network and particleswarm optimization algorithm with crossover and mutation operation is adopted, where theneural network including the information of hand-eye system rotating component is designed.When the solving system comes to the global optimal equilibrium position, hand-eye relationsbetween the end-effector and the camera installed in robot are obtained from the networkweights. Precision analysis shows that the proposed approach can ensure the orthogonality ofthe vectors in the rotational matrix of the hand-eye relations.On the basis of the above key technologies, machine vision system is calibrated withcomputational intelligence, the PCB micro-drill’s automated detection system is developed,and the relations between camera and robot’s end-effector are gotten by calibrating the robothand-eye system. Finally calibration and testing experiment are carried out to demonstrate theabove research being correct and practicable.The research will help to improve technology level of precision manufacturing andtesting, and provide another dimension for promoting the development and applications ofmachine vision research in our country.