Scene Depth Detection Technology Via Multi-line Scanning

The depth information of the scene is usually obtained by a specific machine vision device.Typical systems are TOF depth cameras,structural light triangulation based devices or camera array systems.The multi-view image is obtained mainly by the camera array.The camera array refers to a certain number of cameras arranged according to certain rules.Each camera can acquire images at different viewing angles at the same time point,thereby obtaining a multi-view image sequence.Aiming at the problem of parts inspection in industrial environment,this paper proposes a multi-line scanning light field image acquisition and processing system by adopting a camera array method combining single camera and high precision displacement platform.The image acquisition system consists of an area scanning camera that extracts a small number of sensor lines at a high frame rate,a high-precision displacement platform that transmits the detected object at a constant speed and direction,and a dome light source.The acquisition system scans the target scene from different angles by means of multiple baselines,greatly increases the number of perspectives of the target scene,and obtains image information of the same target scene at different viewing angles.The image processing system splits the acquired multi-view image,first extracts the central pixel column of each image,and sequentially arranges the central pixel columns extracted in each image to form a central perspective image of the three-dimensional light field data,and then sequentially extracts a pixel column having a parallax with a central pixel column in each image is combined into a plurality of light field images at different viewing angles.The EPI is generated from the synthesized multi-view three-dimensional light field image,and the initial range of the assumed slope of the three-dimensional light field image is obtained according to the EPI,and the assumed initial slope range is uniformly sampled to obtain a hypothetical slope set.Dense slope estimation is performed on the assumed slope set,that is,the cost maps under different baselines are calculated by the adaptive weight stereo matching algorithm under each hypothesis slope,and the cost graphs obtained under all hypothetical slopes are compared to determine the minimum cost value.,based on the minimum cost value to determine the best slope of the match.On the basis of obtaining the optimal slope,the quadratic fitting method is used to optimize the parallax,and the accuracy of the system depth estimation is further improved.Finally,experiments are carried out on the targets in some scenarios.The experimental results show that the proposed multi-line scanning light field image acquisition and processing system can accurately estimate the depth information of the target scene.