IVVS：智慧車用視訊雷達系統研發計畫---總計畫暨子計畫一：540 度半球狀環場鏡頭陣列視訊合成技術與物件感知之高動態夜視處理技術

Abstract

自從 Google 自動駕駛車輛成功挑戰歐亞長距離行駛以來，世界各國車廠對於自動駕駛汽車技術更具信心，也帶動各式輔助汽車駕駛人的先進自動駕駛技術的蓬勃發展。EE Times 在2013/08/23曾報導自動駕駛車輛(包含部分功能自動化駕駛輔助、全自動駕駛輔助等)於2025年到2030年間的市占率將達25%-30%，全球各大車廠及汽車零組件廠如BENZ、BMW、AUDI、VOLVO、Toyota、Honda、NISSAN、Ford、Continental AG、Bosch、Mobileye、Renesas、Denso等，對於智慧型車輛技術的研發投入相當積極。而為了搶攻全球高達數百億的車用安全產品商機，國內車用電子相關業者無不磨拳擦掌與工研院、財團法人車輛研究測試中心(ARTC)、學術界研究團隊密切合作，多方持續投入研發資源。未來具多媒體影像輔助之各式車用電子系統產品將持續蓬勃發展，這些創新技術可以有效提高行車安全性與便利性，降低車禍發生率，保障用路人之生命安全。此技術發展趨勢將從主動式安全駕駛輔助技術，朝向部分自動駕駛功能技術，甚至未來的全自動駕駛功能技術。 為因應此技術發展趨勢，本計畫將基於現階段執行第三年的國科會微電子學門整合型計畫：「智慧型行車安全紀錄系統(ISDRS)晶片之關鍵性技術研發」之研發成果(含360度環景視訊、雙鏡頭HDR夜視技術、前車防碰撞技術、行人偵測技術等)為基礎進行持續研發，提出全新之智慧視覺系統應用整合計畫：「IVVS: 智慧車用視訊雷達系統研發計畫(IVVS: Intelligent Vehicle ViDAR System)」，希望研發智慧車用視訊雷達關鍵技術與其系統雛型，讓現階段以視訊為基礎之車用駕駛輔助技術能夠更進一步提升至具有雷達功能(如環景、定位、追蹤、測距、夜視、辨識、行為分析等)之先進視訊駕駛輔助功能，以朝向部分自動駕駛功能技術，甚至未來的全自動駕駛功能技術的目標來邁進。我們也希望藉由此計畫，逐年完成數項具前瞻性之創新智慧視覺系統應用技術，以提昇國內學術界創新技術之研發能量。本計畫內容包含總計畫：「IVVS: 智慧車用視訊雷達系統研發計畫」與子計畫一: 「540度半球狀環場鏡頭陣列視訊合成技術與物件感知之高動態夜視處理技術」。本整合型計畫及其子計畫之規劃如下： 計畫項目 主持人/共同主持人 計畫名稱 總計畫 郭峻因/楊家驤 IVVS: 智慧車用視訊雷達系統研發計畫 子計畫一 郭峻因/葉經緯 540度半球狀環場鏡頭陣列視訊合成技術與物件感知之高動態夜視處理技術 子計畫二 蘇慶龍 多重移動物件方向定位、測距與行為分析技術 子計畫三 陳冠宏 多重移動物件辨識與追蹤之演算法開發與晶片設計 子計畫四 張添烜 即時超高畫質車用視訊編碼 子計畫五 鄭經華 效能功耗優化之多視訊智財晶片系統設計策略與晶片驗證 搭配總計畫之設計目標，本計畫在三年規劃中預計將可產出 2 (1) 高資料存取頻寬能力之視訊雷達系統SoC架構設計; (2) 540度半球狀環場鏡頭陣列視訊合成技術; (3) 物件感知之高動態夜視處理技術; (4) 多重移動物件方向定位、測距與行為分析技術; (5) 多重移動物件辨識與追蹤技術; (6) 智慧車用視訊雷達系統之視訊資料壓縮技術; (7) 效能功耗優化之多視訊智財晶片系統設計策略與晶片。

Since the Google automatic vehicle successfully challenges the automatic driving from Europe to Asia in 2012, the vehicle manufacturers over the world have expressed much more confidence on the development of the automatic vehicle technology, which also enables a rapid and emerging trend in developing the advanced automatic driver assistance technology recently. In 2013/8/23, EE Times reported that the market share of the automatic vehicles (including vehicles with partial automation functions and fully automation functions) will be about 25%~30% during the years of 2025~2030. According to this technology development trend, the famous leading companies in car manufacturing like BENZ, BMW, AUDI, VOLVO, TOYOTA, HONDA, NISSAN, FORD, Continental AG, Bosch, Mobileye, Renesas, and Denso, all deeply involved in developing the novel intelligent technologies for car driving. For pursuing the market share of this new technology, the domestic autotronics companies work closely with ITRI, ARTC, and Academia to invest the research in this area. With this trend, it would be seen that there will be prosperous evolution in the various video/image supplementary technology for safety car driving in the near future. Using these technologies will further improve the safety in driving and reduce the car accident events as well. This technology evolution will start from the active safety nowadays to the technology for partial automatic vehicles or even full automatic vehicles. Following this technology trend and the research results (including 360 degree panoramic view, HDR night vision processing, Forward Collision Warning System (FCWS) technology, and pedestrian detection technology, and so on) of our previous NSC grand project, “ISDRS: Intelligent Safety Driving Recording System and its key technology development”, we propose a brand new project entitled “IVVS: Intelligent Vehicle ViDAR System,” with the design goal to develop intelligent vision radadr technology and the associated system prototype to further enhance the current advanced driver assistant system (ADAS) technology to incorporate the functions of radar, like surround view, identification, location, ranging, night vision, and behavior analysis, for the next generation ADAS or even automatic driving assistance systems. The design concept of the proposed IVVS system is shown in Fig. 11-1. In this project, we would like to investigate the interesting research topics about the vision radar technology including 540 degree surround view technology, object aware high dynamic range night vision processing based on camera arrays, multiple object detection and tracking technology, multiple object identification, ranging, and behavior analysis, ultra high definition video coding technology for vision radar, and high performance/low power circuit design for vision radar SoC. Combining the above mentioned technologies could effectively induce various novel applications in the area of autotronics in the near future. Moreover, we will develop some advanced 2 video/image processing technology in this project to enhance the research capability of Taiwan in this area. The plan of the main project and its associated sub-projects are shown in the following: Projects PI/Co-PI Title Main project J-I Guo/C-H Yang IVVS: Intelligent Vehicle ViDAR System Sub-project 1 J-I Guo/C-W Yeh Vision processing technology for 540 degree half-sphere camera array and object aware HDR night vision Sub-project 2 C-L Su Multiple object identification, ranging, and analysis Sub-project 3 K-H Chen Algorithm Development and IC Design for Multiple Moving Objects Recognition and Tracking Sub-project 4 T-S Chang Real time ultra HD video encoder for automobiles Sub-project 5 C-H Cheng The Performance-Power Optimized Multiple Video IP System Development Methodology and Chip Validation In the three-year period of this project, we plan to deliver the following items: (1) SoC architecture design for ViDAR SoC with high memory accessing bandwidth; (2) Vision processing technology for 540 degree half-sphere camera array; (3) Object aware HDR night vision processing technology; (4) Multiple object identification, ranging, and analysis based on camera array; (5) Multiple moving objects recognition and tracking based on camera array; (6) Ultra high definition video coding technology for ViDAR SoC; (7) High performance/low power circuit design methodology and test chip validation for ViDAR SoC.