Research Of Image Mosaic And Remote Panoramic Monitoring Of Fisheye Camera

Security monitoring technology plays an important role in various industries,and low-cost and high-quality security monitoring technology has always been one of the goals pursued by consumers.In order to enhance reliability,modern security monitoring systems usually add cameras to reduce the existence of monitoring dead space,but the increase of equipment will inevitably lead to the increase of system cost.However,the monitoring system with wide view fisheye lens can achieve a wider monitoring range with fewer cameras,which means that consumers who buy such a monitoring system can obtain high-quality services at a low price.The angle of view of a single fisheye camera can reach or even exceed 180 degrees.In this paper,two fixed assembly and fixed installation positions of the network fisheye cameras are used to take pictures,and the captured pictures are preprocessed.Then,using GPU hardware acceleration,video image rendering is done through OpenGL,the effective area of fisheye image is rendered to a special spherical panoramic model,and In order to achieve the goal of panoramic monitoring,image mosaic and fusion are completed on the spherical panoramic model.In the preprocessing of fisheye image,because of the uneven installation or assembly error of the camera base,the image captured may have a certain oblique angle,hence an error in the subsequent image.An effective way to eliminate the oblique angle of the picture captured by the fisheye camera is put forward in the paper.And experimental results show that Camera calibration image rotation can effectively resolve the oblique angle.After the fisheye image is corrected,feature points are matched and the average distance of feature points is calculated for subsequent steps.The average distance registration algorithm of image feature points is improved.The acquired data make the image registration better.The coordinate of the vertex of the sphere is directly corresponding to the pixel coordinate of the effective area of the fisheye image by the mapping relation of longitude and latitude.The video image is rendered to the spherical panoramic model by using OpenGL hardware acceleration.In order to acquire some overlapping area of the two fisheye images on the sphere,the algorithm of image pixel arrangement needs to be improved and the OpenGL fragment shader for image fusion should be applied.Because the overlapping area is a cambered surface,there will be gaps when the overlapping area is fused according to the conventional rectangular overlapping area.The system linear fusion algorithm is improved in this paper and the improved algorithm effectively eliminate the gaps and is validated by experiments.The experimental results show that the proposed algorithm has high efficiency,good stitching effect and strong authenticity.It can meet the needs of 360-degree panoramic real-time monitoring and achieve the desired results.