智慧型駕駛輔助暨即時視訊壓縮系統之研究

Abstract

在本論文中，我們以低複雜度與低記憶體資源使用為訴求，發展一套結合離散餘弦轉換 (Discrete Cosine Transform，DCT)與改良型零樹編碼 (Zero-tree Coding，ZTC) 的靜態影像壓縮技術。另外，於差值訊號編碼 (Differential Pulse Code Modulation，DPCM)中，利用每一個DCT區塊的DC (Direct Current) 之間的相關性來建立前文模型 (Context Model)，以增進壓縮率。更在量化器上加上量化階偏移 (Step Size Offset)與視覺感知編碼 (Perceptual Coding for Vision)，以提升壓縮效能與視覺效果。並進行一連串的靜態圖像測試，如Lena、Barbara、Jet與Baboon等。根據測試結果顯示，我們所提出的方法在不同的壓縮倍率 (Bit Rate，bpp) 下比低計算量零樹編碼 (Low-Complexity and Low-Memory Entropy Coder，LLEC) 平均高出0.62 dB、JPEG平均高出1.36 dB以及JPEG-O平均高出0.93 dB。 接著，以我們所提出的靜態影像壓縮技術為基礎，針對即時監控 (Real-time Surveillance)系統之影像特性，設計一套達到高壓縮率、快速編碼/解碼與高影像品質的視訊壓縮編解碼器 (Codec)。並於相同的壓縮率下優於MPEG-4的編碼速度。 最後，針對夜間行車駕駛輔助與車輛自動化駕駛需求，並結合了視訊壓縮技術所發展的智慧型駕駛安全輔助與監控系統。其特點如下： 1) 利用影像切割與圖形分析技術有效地偵測及追蹤前車和來車。 2) 新式的自動切割物件的方法能夠在具有複雜光源環境的夜間市區道路上做車輛偵測。 3) 提供有效的資訊，以輔助駕駛能夠清楚地了解車外周遭的交通狀況。 4) 提供一個有效的機制，以自動操控車上的相關裝置設備，如： a. 智慧型車頭燈之遠光燈與近光燈之切換。 b. 防撞 (Pre-cash) 警示系統以避免與前車追撞。 c. 自動巡航速度控制。 5) 黑盒子機制的設計，監控可能發生的交通事故而觸發即時視訊錄影，以作為日後意外發生原因判讀依據。 從實驗結果的分析，可以證明我們所提出的這套系統於駕駛安全輔助具有的卓越的性能及其高度的實務價值。

In this thesis, we propose a low-complexity and low-memory technique with combination of discrete cosine transform (DCT) and zerotree coding (ZTC) for image compression. Furthermore, Exploiting the correlation of direct current (DC) every DCT block codes the context model in DPCM scheme and step size offset is utilizing in quantization scheme for improvement of compression ratio. Besides, a perceptual quantizer is integrated to enhance the visual quality of image. The PSNR results of our proposed approach at different bit rates for images, i.e. Lena, Barbara, Jet and Baboon etc., are superior to low-complexity and low-memory Entropy Coder (LLEC) by 0.62 dB, JPEG by 1.36 dB and JPEG-O by 0.93 dB on average. Next, we present a video codec for the sequential characteristic of surveillance images to accord with real-time requirement. The codec can achieve high vision quality, high compression speed and high compression ratio based on our proposed image compression approach. It also outperforms MPEG-4 in compression speed at the same bit rate. Last, we present a real-time vision system for assisting driver during nighttime driving with surveillance technique. Our proposed system provides the following features: 1) Effectively detection and tracking of oncoming and preceding vehicles based on image segmentation and pattern analysis techniques. 2) Robust and adaptive vehicle detection under various illuminated conditions at nighttime urban environments benefited by a novel automatic object segmentation scheme. 3) Providing beneficial information for assisting the driver to perceive surrounding traffic conditions outside the car during nighttime driving. 4) Providing a versatile control strategy for in-vehicle facilities of the autonomous vehicles such as: a. Intelligent control of vehicle high-beam and low-beam states of headlights b. Pre-crash safety system for collision avoidance c. Automatic Cruise Speed Control 5) An offering real-time traffic event-driven video surveillance machinery for recording evidences of possible traffic accidents. The analyses of the experimental results demonstrate the effectiveness and feasibility of the proposed system on nighttime driver assistance issues.