| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 張鈞為 | zh_TW |
| DC.creator | Jun-wei Zhang | en_US |
| dc.date.accessioned | 2012-7-16T07:39:07Z | |
| dc.date.available | 2012-7-16T07:39:07Z | |
| dc.date.issued | 2012 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=995202069 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 汽車交通事故發生的主要因素是駕駛人在車輛行進中沒有注意到障
礙物而產生的。特別是因為後照鏡死角及車身遮蔽所造成的視線死角，
更是許多駕駛人無法察看的區域；常常會在這個死角範圍內發生擦撞造
成車體損壞、人員受傷。為提高停車時的安全性，並減少停車所需的時
間。我們提出一套改良影像式倒車導引與全周俯瞰監視系統，並將之部
份實現於嵌入式系統中。整個系統共包含兩部份：一是影像式倒車導引
用於協助駕駛調整方向盤停入車位中，二是全周俯瞰監視用於輔助駕駛
觀看車體周遭的狀況。
影像式倒車導引系統則是以車尾影像估計光流，藉由光流濾除及累
積後，計算出車輛前輪轉角，繪製出行車軌跡。而全周俯瞰監視系統在
車輛四邊架設廣角相機以拍攝車輛週遭影像，經過離線處理的相機校正、
扭曲校正、暗角校正，俯瞰轉換後，得到四張俯瞰影像的相對關係。再
以一部相機由上方拍攝車輛周圍的特徵，將俯瞰影像快速對位成一張俯
視車輛週邊的全周俯瞰影像，最後計算色彩混合權重將各項參數建立一
張查找表，在線上處理階段則先以直方圖均勻化調整影像的亮度分佈，
再根據查找表查表內插與暗角消除影像，產生全周俯瞰影像。
後視倒車導引在輸出解析度為 720×480 的解析度下，在 Intel?
Core™2 Duo 2.83GHz 及 3GB RAM 的個人電腦上可達每秒 170 張。在
Texas Instruments? DaVinci™ DM3730 1GHz Digital Media Processor 開
發板上達每秒 12 張的處理速度。全周俯瞰倒車導引在 Intel? Core™2
Duo 2.83GHz 及 3GB RAM 的個人電腦上可達每秒 43 張。
| zh_TW |
| dc.description.abstract | The main reason for car accidents is that drivers couldn’t see the area
around the vehicles. In order to avoid accidents while driving and to reduce
the time for parking. We propose a refining image-based parking guiding
system and a surrounding top-view monitoring system for parking assistance.
Image-based parking guidance system uses the rear view image to
estimate, filter, and accumulated optical flows to calculate the front wheel
angle of vehicle, and then draws the driving trajectory.
The surrounding top-view monitoring system has four wide-angle
cameras mounted on front, rear, and the both sides of the vehicle to capture
images. The system consists of off-line and on-line processes. In the offline
process, we calculate the camera intrinsic and extrinsic parameters and then
estimate parameters of distortion model and vignetting model for distortion
correction and vignetting compensation. Then we estimate the homography
matrices of four cameras by a top-view image. Lastly, we calculate the feature
weights of overlapped regions and bulid a lookup table for the on-line process.
In on-line process, we use histogram equalization to adjust brightness, and
then interpolate the top-view image, compensate for vignetting effect, and
blend the overlapped regions. In our experiment, the frame rate of
Image-based parking guidance system is 170 frames per second.
| en_US |
| DC.subject | 相機參數校正 | zh_TW |
| DC.subject | 影像對位 | zh_TW |
| DC.subject | 鳥瞰轉換 | zh_TW |
| DC.subject | 影像扭曲校正 | zh_TW |
| DC.subject | 光流 | zh_TW |
| DC.subject | 影像暗角補償 | zh_TW |
| DC.subject | distortion corr | en_US |
| DC.subject | camera calibratio1n | en_US |
| DC.subject | optical flow | en_US |
| DC.title | 改良影像式倒車導引與全周俯瞰監視 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Refining Image-based Parking Guiding andSurrounding Top-view Monitoring | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |