Design and Implementation of High Speed Real-Time Stereo Vision Tracking Object System

• Main Authors: Syu, Jhen-Siang, 徐振祥
• Other Authors: 許毅然
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/97601390342672007701

碩士 === 南台科技大學 === 電機工程系 === 98 === ABSTRACT A common phenomenon has been found to occur during the usage of consumer electronic products such as camera phones, digital cameras, surveillance systems, and webcams. When the movement speed of the object across the camera screen is too quick, image distortion, blurriness and transience object shift will occur. The main factor that leads to this phenomenon is the insufficient frame rate of the screen. To resolve this problem, we attempted to increase the frame rate to over 60 fps , which is beyond the image visual frequency of the human eye. At this rate, high-speed real-time images can be achieved, effectively resolving the difficulty of tracking and identifying target objects in motion video. This will also allow for the widespread application of machine vision/computer vision technologies. Furthermore, in the field of artificial intelligence, it can be utilized to aid machine platforms complete tasks beyond the reach of humans and further enhance system efficiency, functionality and application. The purpose of this research is to construct a real-time 3D visually-controlled self-propelled vehicle for the tracking of target objects. The system employs a binocular stereo vision to detect the actual distance between the vehicle and the target object, and features real-time lock-in with the target object to maintain distance and direction. We implemented FPGA (Field Programmable Gate Array) to realize the design and verification of real-time 3D visual target object-tracking systems. In addition, the resolution of 320X256 was utilized to increase the frame refresh rate to 166 fps . This improved the slow image processing rates of previous systems, and further overcame the inability of the real-time processing of images. This grants self-propelled vehicles and the system platform with visual perception and high-speed real-time 3D functionality for the visual tracking of target objects.