車用輔助駕駛之惡劣環境視訊處理技術與實作

Design and implementation of vision processing for advanced driver assistance system in harsh environments

徐盛威(Sheng-Wei Hsu)

指導教授 ： 郭峻因

國立中正大學/工學院/資訊工程學系/碩士(2014年)

摘要

由於智慧汽車設計趨勢的興起，讓有關車用影像主動安全技術之研發蓬勃發展，文獻上有許多防碰撞與物件偵測技術研究發表，如車道偏移警示系統、行人偵測系統、前方車輛防碰撞系統等。許多系統預期在良好的視訊環境下做處理，但是一旦天候狀況改變，例如起霧或者視野昏暗，則會造成擷取影像顏色淡化及對比度降低，使得系統無法正常工作。而現今已經有許多針對單一惡劣天候的影像處理技術的研究成果。絕大部分的惡劣天候影像處理技術只能針對單一惡劣環境做處理，並不能同時有效處理各種惡劣環境的影像劣化之狀況。觀察不管在夜間、起霧、陰天和雨天，都有對比度較低的統一特徵。有鑑於此，本論文以影像增強對比技術之區域直方圖等化為基礎，設計出可以處理大部分受到惡劣環境影響的動態區域對比增強技術。使技術可應用於有霧、夜間之對比降低的影像。最後，本論文設計一套車道偏移警示系統並採用所提之動態區域對比增強方法，同時實作於嵌入式平台以驗證此方法的效能。另外，本論文所提之車道偏移警示系統執行在ARM-Cortex A8 300 MHz平台上，在VGA解析度之影像下，系統操作效能超過每秒30張以上。

關鍵字

惡劣天候處理 ； 夜視 ； 車道偏移警示 ； 去霧 ； 直方圖等化

並列摘要

Due to rapid development of intelligent automobiles, many key functions in advanced driver assistance systems (e.g. Lane Departure Warning System, Pedestrian Detection System, Forward Collision Warning System, etc.) were proposed in the literature. These systems cannot process well at inclement weathers like foggy day or night since these proposed methods were expected to execute with normal weather conditions. Capturing a low contrast image and shooting a color-faded image might cause failure of these systems. Nowadays, the technologies proposed to deal with inclement weather are framed narrowly, which can only enhance one kind of inclement weathers for each technology. There is no solution which can enhance all the inclement weathers well simultaneously. According to our observation, contrast decrease exists at nights, foggy days, cloudy days, and rainy days. Thus, exploiting the idea of adaptive histogram equalization, we propose a dynamic local contrast enhancement (DLCE) technique, which can strengthen the image quality in most inclement weather conditions. The proposed DLCE technique can improve unnatural over-enhancement of the image and reduce noise as well. The DLCE technique we proposed is able to overcome disadvantages encountered in traditional histogram equalization to make images more saturated so that it can be applied on various inclement weather conditions. Finally, a lane departure warning system with the proposed DLCE technique is designed and implemented on an embedded platform to verify its correctness and robustness. Without specific hardware and software optimizations, the proposed LDWS system realized on ARM Cortex-A8 300MHz is able to achieve VGA@30fps real-time processing performance.

並列關鍵字

Night vision ； Lane departure warning system ； De-haze ； Histogram equalization ； Harsh vision processing

參考文獻

[1] R. C. Gonzalez and R. E. Woods, Digital Image Processing, 2nd Edition, Prentice-Hall, Inc., 2002.

[2] Y. Y. Schechner, S. G. Narasimhan, and S. K. Nayar., "Instant dehazing of images using polarization," Proc. of IEEE Conf. on Computer Vision and Pattern Recognition, CVPR 2001, vol. 1, 2001, pp. 325–332.

[3] S. Shwartz, E. Namer, and Y. Schechner., “Blind haze separation,” Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR 2006, vol. 2, 2006, pp. 1984-1991.

[4] S. G. Narasimhan, S. K. Nayar, “Contrast restoration of weather degraded images,” IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 25, June 2003, pp. 713-724.

[5] R. T. Tan, “Visibility in bad weather from a single image,” Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR 2008, June 2008, pp. 1-8.