Moving Object Detection on a Vehicle Mounted Back-Up Camera

• Published in: Sensors (Basel, Switzerland) Vol. 16; no. 1; p. 23
• Main Authors: Kim, Dong-Sun, Kwon, Jinsan
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 01-01-2016; MDPI
• Subjects: Algorithms; Backing; backup collision intervention; Cameras; Field programmable gate arrays; Methods; mixture of Gaussians; moving object detection; Platforms; pyramidal Lucas-Kanade optical flow; Roads; Sensors; Telematics; Vehicles; Vision; backup collision intervention; moving object detection; pyramidal Lucas-Kanade optical flow; mixture of Gaussians
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s16010023

In the detection of moving objects from vision sources one usually assumes that the scene has been captured by stationary cameras. In case of backing up a vehicle, however, the camera mounted on the vehicle moves according to the vehicle's movement, resulting in ego-motions on the background. This results in mixed motion in the scene, and makes it difficult to distinguish between the target objects and background motions. Without further treatments on the mixed motion, traditional fixed-viewpoint object detection methods will lead to many false-positive detection results. In this paper, we suggest a procedure to be used with the traditional moving object detection methods relaxing the stationary cameras restriction, by introducing additional steps before and after the detection. We also decribe the implementation as a FPGA platform along with the algorithm. The target application of this suggestion is use with a road vehicle's rear-view camera systems.