Research On Registration And Fusion Method Of Infrared And Visible Images On The UAV

With the rapid development of imaging sensor technology and the increasing complexity of the application environment, the imaging sensor of single-source is difficult to meet the current needs of practical applications. But multi-source imaging sensor is able to treat with information obtained by various image sensor and increase the usefulness and richness of the information. Installing the imaging sensor on UAV makes the UAV widely used in military and civilian fields. Recent research is focus on the registration and fusion of infrared-visible images. This paper studies the registration and fusion algorithm of the infrared and visible images, based on the UAV embedded infrared target detection system. The main contents are as follows:(1) Based on the registration algorithm of mutual information, the paper studies the parameter optimization algorithm to maximize the mutual information value. According to the Powell algorithm, several experiments are conducted to analysis the impact from the choice of initial point, the number of iterations and the given error, on algorithm results; According to the Particle Swarm algorithm, the experiments analysis the impact on algorithm results when choosing different particle number and different iterations. Besides, the paper puts forward Powell-PSO algorithm. The experiments show that this algorithm can not only take advantage of the Particle Swarm algorithm to find the global optimal solution, but also avoid falling into local extremum of Powell algorithm. Thus, this algorithm can greatly reduce computing costs and enhance the overall optimization performance.(2) Based on the specific real-time registration problem of the images captured by UAV imaging sensor, the adaptive mutual information registration algorithm is proposed. The experiments verify that the adaptive mutual information algorithm is useful and ideal for infrared and visible images, which are of different resolution and captured by long-range imaging sensor. And it proves that the algorithm is effect on UAV platform.(3) The paper studies several common fusion algorithms of infrared and visible images: principal component analysis PCA fusion algorithm, Laplace pyramid fusion algorithm and the wavelet transform fusion algorithm. Based on the features of infrared and visible images captured by UAV imaging sensor, a fusion algorithm of the IHS-upgrade wavelet transform is put forward. The experimental results show that the algorithm can not only obtain the rich details and color information of the visible image, but also obtain the brightness information of the infrared target object.(4) The paper builds a video capturing and transmission system on the infrared target detection of UAV embedded platform, and it realizes the real-time two channel images, which are infrared image and visible image, acquisition and transmission. Using the real-time capturing infrared and visible images on UAV platform, the experiments verify the effectiveness and feasibility of adaptive mutual information registration algorithm and algorithm of IHS-upgrade wavelet transform fusion.