Research On A Neural Network Image Fusion Algorithm Based On LCLS Mode

Information fusion is based on a certain fusion rule, it can deal with data collected from multiple sensors in many ways. It studies how to utilize multi-sensor information effectively. The information obtained by a single sensor is usually incomplete, the data and information obtained from the sensors are independent and complementary. image information fusion technology is a specific algorithm, The effective information is extracted from the source image data collected by the sensors, The fused image is more comprehensive and clear, and the fused information is more convenient for automatic detection of the machine and identification of the human eye.Multi sensor image fusion algorithm is a key technology to eliminate the image noise, lots of signal level image fusion technology is based on statistical methods, although these data fusion methods are very effective, there are many disadvantages, such as minimum variance estimation method. This method needs a priori covariance information, therefore, the traditional image fusion processing method has been unable to meet the people's needs. The processing function of neural network is embodied by the mutual connection of neurons, which simulates the human brain by simulating the function of a single neuron and the structure of the human brain. Neural network method shows great superiority in simulating the brain processing image, which makes the neural network computing methods in the field of multi-sensor image information fusion has been widely developed.The main research of this paper is a neural network image fusion algorithm based on linearly constrained least square(LCLS) model, which can improves the image quality. The neural network image fusion algorithm based on LCLS model combines the advantages of the classical algorithm and the modern algorithm, which can overcome the shortcomings of the traditional methods such as the singularity of the noise covariance. Based on the traditional neural network image fusion algorithm, this paper proposes a new recurrent neural network. The image fusion algorithm is based on the linear constrained least square model, and the basic idea of this model is to find a set of optimal weight coefficients, the weighted sum of the data obtained by multi-sensor and weight coefficients is the optimal fusion image information, which makes the square sum of the error between obtained data and the actual data minimum.In order to obtain a set of optimal weights, a recurrent neural network algorithm is constructed. In the neural network algorithm, the penalty function is cited, which is more simple and effective to restrain the linear condition. The projection function is referenced by the algorithm, and it is proved that the projection neural network can converge to the optimal solution, so the algorithm proposed in this paper can also converge to the optimal solution. The projection function makes the algorithm easy to the realization of the hardware circuit, and makes the image fusion algorithm structure more simple and fast convergence to the optimal fusion scheme. Using Matlab to deal with the noise color image, the image fusion results show the reliability and validity of the algorithm.At last, the paper summarizes the work of this paper, and looks forward to the next step of the work.