Visual Recognition Of Semiconductor Particles In Thermoelectric Refrigeration Chips And Trajectory Planning Of Mount

Thermoelectric refrigerating chips uses the pelter effect of semiconductor material to achieve refrigeration effect which is composed of ceramic substrate,conductive copper chips and many semiconductor particles.It is widely used because of its characteristics of fast refrigeration,small cold heat inertia and environmental protection.At present,the arrangement and installation of semiconductor particles in the thermoelectric refrigeration chips manufacturing process is mainly completed by manual or semi-automatic production.The basic process is to pour the semiconductor particles into a mold for arrangement,and then remove and correct the side up or unqualified particles through manual inspection.With the development of technology,the shape and size of semiconductor particles are required to be smaller and smaller,and semiconductor particles have front and side,using the current production process will be prone to jam and more side up particles and other problems.For manual or semi-automatic production technology cannot satisfy the smaller semiconductor particle size and blanking under the demand of the need to distinguish between front and side,this thesis studies a method for semiconductor particle mounting of thermoelectric refrigeration chips by visual recognition.First of all,through the traditional threshold segmentation method to identify location of semiconductor particles,and then the neural network algorithm to obtain front semiconductor particles fot further recognize.Finally,the optimal feeding trajectory planning is carried out based on the coordinate information of semiconductor particles that met the requirements.The main research contents of this thesis are as follows:(1)Study a process method of semiconductor particle mounting technology which is based on the principle of SMT machine.Design a semiconductor particle mounting system for thermoelectric refrigeration chips,and build an experimental platform for image acquisition.The collected image is processed by image processing technology,and then the semiconductor particles are identified and located by multi-threshold segmentation algorithm,edge detection algorithm and minimum peripheral rectangle algorithm which's target is to obtain the front semiconductor particle image information.(2)Study a method of classification and recognition for semiconductor particle feature surface which is based on deep convolutional neural network.According to the different color features of semiconductor particles on the front side,a neural network recognition and judgment model of semiconductor particle feature surface was established,and the neural network was trained.Neural network is used to identify and judge the image of front semiconductor particles obtained in the last step to prevent the failure of the product caused by side semiconductor particles being mount.(3)Conduct experiments and studies using improved ant colony algorithm for thermoelectric refrigeration chips mounting trajectory planning.The front semiconductor particle information that met the requirements through the traditional identification method and neural network identification was imported into the trajectory planning algorithm to find the optimal feeding path.To shorten the semiconductor particle taking path to improve the speed of mounting and production efficiency.(4)Develop a set of software for semiconductor particle mounting system of thermoelectric refrigeration chips.The software integrates image processing module,image location identification module,neural network identification module and trajectory planning module,which has compactness and high efficiency.