Dropped Object Detection System Design Based on Binocular Stereo Vision Sensor

Wenchao Wang,Ting Wang,Ruonan Sun,Shiming Xu,Hong Huang
DOI: https://doi.org/10.1145/3579109.3579119
2022-01-01
Abstract:Along with urbanization and the growth of the number of high-rise buildings, incidents of damage to personal property caused by objects thrown from high-rise buildings have increased in recent years. This paper designs a sensing method for the detection of objects thrown from high-rise buildings based on a binocular stereo vision sensor. The sensor is based on a binocular camera that acquires the 3D coordinates of a falling object after Camera Calibration, Rectification of Pixel Pairs and Extraction of Feature Point and Feature Points Searching & Matching. Feature Points Searching & Matching is one of the steps that implements Harris corner detection and ORB algorithm to accurately identify object feature points. This improves the recognition accuracy of thrown objects and also helps to determine the source of falling objects to get their precise floor height and the responsible party, thereby reducing falling object incidents and providing a reference for future development and application of binocular stereo vision.
What problem does this paper attempt to address?