Study On Photoelectric Detection Based On Artificial Neural Networks

Photoelectric detecting technology is an important method to get information.Research on high precision intelligent photoelectric inspecting technology has become a hot topic of researchers all over the world. Accuracy is one of the most important ability targets for the system of photoelectric detecting. The nonlinear error of photoelectric detecting system is a major factor affect of accuracy. Because semiconductor manufacture technics and its dynamic characteristics led photodetectors are sensitive to changing of working temperature,which leds to creating serious non-linear error when the photoelectric detecting system is working at the atmosphere that the temperature is continually changing.So the chief question of increasing the accuracy of photoelectric detecting system is to overcome the nonlinear error. The traditional method of nonlinear distortion correction,such as hardware compensation or checking table,curve fitting etc,is complex and needed too much sample with low accuracy. Artificial Neural Network (ANN) is based on people's Cerebral Neural Network,which can reprsent any non-linear relationship and has the ability to self-learning,consequently,ANN provides a new idea and method to resolve non-linear problems.The thesis analizes kinds of factors which can creat non-linear earror of photoelectronic measuring system, gives a brief introduction of traditional methods for the compesation of non-linear error of photoelectronic measuring system,proves the feasibility of applying ANN in the system of photoelectronic measuring,using BP and RBF artifical neural network in the system of measuring illuminance based on photodiode.As traditional back propagation network gets local minimization too easily, converges slowly and weak generalization ability, using genetic algorithm's ability of optimizing the non-linear references to optimize BP ANN,realizes the compensation of non-linear error of illuminace measuring system which caused by changing of temperature. At the last, using the designed BP and RBF ANN in the illuminance measuring system based on photoelectric cell. The simulation results show that using ANN in photoelectronic measuring system is practical and efficient.