A Rapid Survey Method for Calibrating Outdoor Surveillance Cameras Using a Vehicle‐Mounted Dual‐Camera Stereo Measurement System

ABSTRACT Video surveillance systems play an important role in many fields such as public safety, transportation management and city operations. As the demand for video surveillance continues to grow, the number of outdoor surveillance cameras is increasing rapidly, but there are also problems such as the lack of positional information during installation and angular shift during use, which seriously affects the layout management and analysis applications of video surveillance systems. Camera calibration is a common method to obtain the position and orientation information of the surveillance camera, but for the millions of monitoring cameras in the city in a large area, the existing camera calibration method is highly accurate, but the efficiency is too low and vulnerable to complex environmental restrictions and can not be carried out in practice. To address this problem, this paper proposes a rapid survey method for outdoor surveillance cameras based on a vehicle‐mounted dual‐camera stereo measurement device, which utilizes the street images automatically collected by dual cameras, combines the deep learning keypoints recognition algorithm and triangulation algorithm to quickly calculate the 3D spatial coordinates of the feature points in the vicinity of the surveillance camera, and then obtains the intrinsics of the surveillance camera through the deep learning network, The spatial position and orientation information of the surveillance camera is finally calculated by the Perspective‐n‐Point (PnP) algorithm, realizing the automated and rapid survey of large‐scale outdoor surveillance cameras. The experiment proves that this method can realize efficient and automated survey of outdoor surveillance cameras, and the calibration results obtained meet the accuracy requirements of layout management and optimization of outdoor surveillance cameras.