201211941 六、發明說明： 【發明所屬之技術領域】 【先前技術】 始搭載此種行車記^器u因的客觀證據與調查等，而開 奶錄器，可取得：⑴本身車輛的前方動態影像(前方公 f身車輛的加速度(前後或是左右方向）、⑶Ιΐίΐ 2 度正#析在事故或幾近意外 近的=二之===^=^環境(包括例如突然靠 所取得之前方動態影像得知。而在Ϊ故器的攝影機： 組中’藉由操縱方動態影像群 抽取。而操縱者指定所記錄的影像每ί圖框=“外事件事例等 輛：寬=車二；前行車 像，來抽取追撞前 201211941 ' 雜且操作負擔龐大，以上皆為其問 近意外事件事儀_基準，會，追揸事故事例或幾 生上述事例的抽取顏或錯誤，以二2有各種不同，容易發 所抽取的前方動態影像，並非客觀的如 =問題點。由此可知， 件事例，使肖此前方動態影象事故事m幾近意外事 [習知技術文獻]'d述事故預防對料係不正雄的。 [專利文獻] 非專利文獻1 ··北島創及i 錄器所施行之追撞事故發生機制之 a由影像記錄型行車記 與駕駛減速行㈣觀察 ^ 2 =胸餘度的測量 曰本自動車研究所，以險=隹定:」，自動車研究， 牛月第28卷，第6號，P.205-208。 【發明内容】 [發明所欲解決的問題] 器等車用g二二二判斷在從行車記錄 或幾近意外事件事例因之既事 省去^斷’社為其轉職了其酬，或是 [解决問題之技術手段] 由車態行為分析裝置，其_於：判斷 有該其他車輛的全部或—部分之選擇區ί H f準影像圖框以外的影像圖框中，追蹤該選 選擇區域的紋理的旋轉移動以及形態變化，而是 5 201211941 的紋理的平行移細及尺寸数，來追賴轉區域。 = ΐ裝置’容易由其他車輛的最上部(車頂）、最下部(後保 接ί部)或是中央部(後玻璃的下邊或後車箱上部)等產生橫 他車輛。藉此’操縱者利用二 =省：輕操縱者的負擔’更可正確進行操縱者的判斷， 士 ^在以榼向邊緣為基準所產生的選擇區域，可能包含僅人 有天玉或路面的區域。在此選擇區域追】 =紋理的平行移動以及尺寸變更，從而;防:i;=; 算，P，僅含有天空或路面=j他= 持在區域内的獨特性’從而本發明的選擇區域 =像圖框所含有的其他車輛係映出後面或前面而靠】 車輛影像之變化。又，紐雜亦如充分 ^ 觸，撕蝴 排除於其他車輛不得存紅錄產 · ，严緣所產生的選擇區域’以減輕在選擇區以= 基準影侧_的橫崎緣之健，來 生部的橫向雜。 ㈣丨主抑擇&域產 特別是車闕影機係拍攝本身車細 程度確定其他車輛相對於本料輛的位置，從而 201211941 二，J基準影像圖框内’位於將前行或後續的且他車麵不尸 域排除在外卿成之指定區域之樺向邊 ^域具有既定伽上㈣鮮之橫向魏，輸出至選擇區 之 個橫向 在此所謂以 邊 方向產生複數個矩形的選擇區域201211941 VI. Description of the invention: [Technical field to which the invention pertains] [Prior Art] The objective evidence and investigation of the driving device such as the driving device are started, and the milking device can be obtained: (1) The front moving image of the vehicle itself (The acceleration of the vehicle in front of the public (front and rear or left and right direction), (3) Ιΐίΐ 2 degrees positive # 析 in the accident or near accidental = two ===^=^ environment (including, for example, suddenly relying on the previous dynamic The image is known. In the camera of the device: in the group 'by the operator dynamic image group extraction. And the operator specifies the recorded image every frame = "outer event case, etc.: width = car two; front Driving image, to extract the 201211941 before the chaos and miscellaneous operation burden, all of which are related to the accidents of the accident _ benchmark, meeting, the story of the story or the above-mentioned cases of the extraction of the face or error, to 2 A variety of different, easy to send forward dynamic images, not objective such as = problem points. It can be seen that the case, so that the dynamic image of the front of the story is almost unexpected [traditional technical literature] 'd Accident prevention [Patent Document] Non-Patent Document 1 ······························································································· Measuring 曰本Automobile Research Institute, 险=隹定:”,Automobile Research, Niuyue Vol. 28, No. 6, P.205-208. [Summary of the Invention] [Problems to be Solved by the Invention] G22nd judged in the case of driving records or nearly accidents, because of the fact that it was omitted, the agency changed its job, or [the technical means to solve the problem] by the vehicle behavior analysis device, _ Yu: Judging that there are all or part of the other vehicle's selection area ί H f The image frame other than the image frame, tracking the rotation movement and morphological change of the texture of the selected selection area, but the texture of 5 201211941 Parallel shifting and size to chase the turn area. = ΐ device 'easy to be used by the top (top), bottom (back) or center of the other vehicle (below or behind the rear glass) The upper part of the car, etc., produces a transverse vehicle. = Province: The burden of the light manipulator 'is more correct to the operator's judgment. The selection area generated by the shovel on the edge of the sacred edge may contain only the area where the person has the sky jade or the road surface. 】 = parallel movement of the texture and size change, thus; anti: i; =; count, P, only contains the sky or the road surface = j he = hold the uniqueness in the area 'so the selection area of the invention = like the frame Other vehicles included reflect the change of the vehicle image behind or in front. In addition, the new miscellaneous is also fully touched, and the tearing butterfly is excluded from other vehicles and must not be stored in red. It is the horizontal miscellaneous of the genre of the genre of the genus of the genus (4) 丨 抑 & & amp 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 域 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 And his car surface is not excavated. The birch to the side of the designated area of Waiqingcheng has the horizontal direction of the established gamma (four) fresh, and the horizontal direction of the output to the selection area is the selection area that generates a plurality of rectangles in the side direction.
Siiiif該橫向邊緣下 = ίί，:為此該選擇區域產生部，宜以含有！： 含有橫向邊緣的方式產在、^；；：二 ^史伴匕场。在此所謂 勒含有橫向 顯示;;:;=;===來判斷是否含有 靠 擇區域内的特徵區域追蹤部的追縱結果，從該選 身車她影像是否含有對於本 資料:儲‘部儲影機所拍攝的複數個動態影像 例，該靠近車輛判斷部，幾近意外事件事 刖 定 由車用攝影式’其特徵在於:判斷 否含有對於本身車二 ==之動態影像，是 仃為分析程式，使電腦 ς :、、=車輛，該車輛動態 緣;選擇區域產生部，、二 圖框内之橫向邊 $基準，於該基準影像圖框内產：有^取得的橫向邊緣 分之選擇區域； 生3有該其他車輛的全部或-部 以及選擇區域追縱部，在該鱗影侧框以外的 連度*近之其他車輛之動態影像。有對於本身車輛以既 201211941 影像圖框中，追蹤該選擇區域；而 算，不偵測該選擇區域的紋理的旋鐘域追蹤部使用概度計 測該選擇區域敝理的平行移動 及職變化，而是積 域。 尺寸變更，來追蹤該選擇區 [發明之效果] •根據如此淑財舯，觸 所取得之動態影像内，是否含有顯二丁 車用攝影機 減輕操縱者的負擔，更可正確進行操&者 斷 的判斷，或是省去其判 【實施方式】 施形Ρ參關式朗依本發日㈣車她態行為分㈣統的-實 <1.系統構成> 依本實施形態的車輛動態八 行車記錄器1〇〇，拍攝汽車(本身如圖1所示，具有： 為分析裝置200，從該行車呓鲦哭）的刖方，以及車輛動態行 料，來判斷在該動態影像資料戶^;，所拍攝的動態影像資 身車輛V顯示既定動態行為之態影像内，是否含有對本 <2.行車記錄器> ° 行車記錄器100，係黏接名‘ 近意外事件時⑽後一定期間二玻=記，在事故發生時或幾 之車用型；係在單一機殼2内，^ 的動態行為或周圍狀況等 所示的_機構3、倾處理機構广了，本構成要素，亦即圖2 通信機構6、可拆卸式記錄機構4、輪人機構5、 狀態，來輸出顯示該狀態動態行為或狀況等之 影機構31、加速度感測器32罢，，至少使用3種：辑 拍攝本身車輛前方的車外狀況來出攝影機構3卜係 出,，、'員不其衫像的狀態資料(動態 201211941 機。加速度感測态32，係利用例如壓電電 用於車輛的卜3維(例如為3維，前後、左右 輸出顯示該加速度的狀態資料(加速度資料）。m ’來 榻取來自例如複數個衛星的紐以侧汽車v ^亥位置的狀態貧料(位置資料)之GPS接收機。另外，狀賴:粗、、貝 .、他有從汽車v的車速感測n送來的車速㈣ 卩ϋ ;車門開閉資料、顯示煞車的〇卿之“資料等 連接器™，亦可_‘二 4通知麟4，由在m的表面所露出的發光 ^内建在賊2鱗鳴n或揚聲料聲音輸_(棚示)等所構 輸入機構5，在此係設於機殼2表面的按_關。 的側纽’細W卩方式絲在於機殼2 開之溝槽2b ’例如CF記憶卡或SD記憶卡。 中央ίίί :構8i在構造上具有CPU(ce_ P職—_’ f 隐體82(例如非揮發性記憶體)、⑷緩衝 含AD轉換器等之情形)83等亦即電腦迴路，係内建 程t 5述CPU81依照儲存在記憶體82的既定區域之 \1Ιησ纟’藉此進彳7上述各機構的控制與資訊處理。 亦’ ^81在記憶體82内所設㈣暫時區域(以下 貢料儲存部)中，時常依序更新並暫時儲存行車中的 且告’亦即力i速度雜或位置資料、動態影像資料等， 時幾近意外事件或事故、異常等發生之現象產生 前後―錢關上述狀纽缝82内的疋 -（下亦稱之為正規圮錄資料儲存部)並記錄。 上述現象，相當於在加速度資料所顯示的加速度(減速度)超過 201211941 既定基準值之情形、或其時間持浐一一 情形、車輛的電源切斷之情形等上之情形、車門開閉之 在車速在上限速度以上之情升$、、 ^，根據所產生的現象，僅 與煞車的有無等幾個其他條‘同，在下限速度以上之情形、 來進^料記錄，而盡可能不進行無益其做為觸發點 亦即，在進行資料記錄之而言’亦備有學習功能。 人是否發生了幾近意外事件或事二，上述通知機構4通知駕駛 與否的輸人(例如上麵來自冑驶人之正確 能以-定程度掌握駕駛人的駕敬 =)。藉由重複此動作， 基準值之變更等，來學習間接 j ，進打加速度的既定 象。 ”，、貞轉轉之該_人特有的現 再者，所記錄的狀態資料，係士 因應所應記錄的重要度來分類之、。己錄時的狀況進行加權， 形，從重要度低者開始刪除而記錄新的量到達極限等情 如此正規地記錄的狀態資料，二、4。 送到分析_心(未圖示；)，或者是傳可二’以無線方式傳 =，送至分射心可胁使 機構7而將其 事後分析。 勖慼仃為分析裝置200的 <3.車輛動態行為分析裝置> 車輛動態行為分析裝置2〇〇，從 器1〇〇所取得的動態影像群組，抽取載的行車記錄 度犬然罪近的其他車輛(以下，亦古ς身車辅以既定速 像’來支援事後分析。該車輛動態行為又象車。)之動態影 備構成，係具有CPU、記憶體、輪^刀^置200的具體性設 用乃至於專用的電腦，依照於上述記慎’二丨、AD轉換器等通 輛動態行為分析程式，使CPU、周邊既疋區域所記憶的車 3所示’發揮做為資料接收部20卜動^= 起運作，藉此如圖 向邊緣抽取部2〇2、選擇區域產生部:像貝料儲存部D1、橫 部的功能之麵車_斷部崩，資^^==^蹤 10 201211941 以下’用圖4〜圖11决士董、+.々μ ^ 資料接收部201，接D2'201〜204與動作。 儲存部之動態影像資料⑼錄《1GG的正規記錄資料 像資料儲存部m(圖4，步====資^^到動態影 無 的可S5S:卩例如CF記：=^器1〇。 有系統地儲存由複數‘ 各動ίίίίΓί f2，取得儲存在動態影像資料儲存部以的 而橫向邊緣抽取部2G2，如圖5所示，將 =⑦度影像’例如用以下的卿階段(3侧以= (1)由Caony邊緣偵測器所施行之邊緣偵測 ；ίϊ ί；；；!3^^; (3) 另外，吾人思及抽取橫向邊緣Χ的方法係有各種手法。Siiiif under the horizontal edge = ίί,: For this purpose, select the area generation part, which should contain! : The method of containing the lateral edge is produced in the ^;;: two history with the market. Here, the so-called Le includes a horizontal display;;:;=;=== to determine whether or not the tracking result of the feature area tracking unit in the selected area is included, and whether the image contains the information from the selection body: In the case of a plurality of motion pictures captured by the shovel, the vehicle is close to the vehicle determination unit, and the near-accident event is determined by the vehicle photographic type, which is characterized in that it is determined whether or not the motion picture for the own vehicle 2 == is 仃For the analysis program, the computer ς:,, = vehicle, the dynamic edge of the vehicle; the selection area generation unit, the horizontal side of the two frames, the reference, in the reference image frame: the lateral edge obtained by ^ The selected area; the raw 3 has all or part of the other vehicle and the selected area tracking unit, and the moving image of the other vehicles in the vicinity of the scale side frame. For the vehicle itself, the selected area is tracked in the 201211941 image frame; and the clock field tracking unit that does not detect the texture of the selected area uses the probabilistic measure to measure the parallel movement and the job change of the selected area. It is the accumulation domain. Change the size to track the selection area [Effects of the invention] • According to such a good fortune, whether or not the camera is included in the dynamic image obtained by the touch, can reduce the burden on the operator, and can operate correctly. Judgment of judgment, or omission of its judgment [Implementation] Shiyan Ρ 式 朗 朗 本 本 ( ( 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四 四The dynamic eight-vehicle recorder 1〇〇, photographing the car (as shown in FIG. 1 itself, having: an analysis device 200, crying from the driving vehicle), and the vehicle dynamic material to determine the dynamic image data User ^;, the captured dynamic image of the vehicle V shows the state of the dynamic behavior of the image, whether it contains the current <2. Driving Recorder > ° Driving Recorder 100, the name of the glue 'near accident event (10) After a certain period of time, the second glass = record, in the case of an accident or a few types of vehicles; in a single casing 2, the dynamic behavior of the ^ or the surrounding conditions, etc. _ mechanism 3, the tilting mechanism is wide, this composition Element, that is, Figure 2 communication mechanism 6, detachable The recording mechanism 4, the wheelman mechanism 5, and the state, output the shadow mechanism 31 and the acceleration sensor 32 that display the dynamic behavior or condition of the state, and at least use three types: the camera is photographed in front of the vehicle. The photography agency 3 is out, and the status data of the member's shirt is not dynamic (201211941 machine. The acceleration sensing state 32 is used for the 3D dimension of the vehicle using, for example, piezoelectric electricity (for example, 3D, before and after, The state data (acceleration data) showing the acceleration is outputted left and right. m 'coming to the GPS receiver of the state poor material (position data) from the position of the vehicle on the side of the vehicle, for example, a plurality of satellites. Rough, Bei. He has the speed of the car from the speed of the car v (4) 卩ϋ; door opening and closing information, display the brakes of the "Qian Qingzhi" information connector TM, can also _ '2 4 notice Lin 4 The light-emitting surface exposed on the surface of the m is built in the input mechanism 5 of the thief 2 scales n or the sound sound input _ (shelf display), and is mounted on the surface of the casing 2 by _ off The side button 'fine W卩 way wire lies in the groove 2b of the case 2 opening 'such as CF memory card Or SD memory card. Center ίίί : The structure 8i has a CPU in the structure (ce_ P----f hidden 82 (for example, non-volatile memory), (4) buffered with AD converter, etc.) 83, that is, a computer The loop, the internal build process t5 describes the control and information processing of the above-mentioned various institutions according to the \1Ιησ纟' stored in the predetermined area of the memory 82. Also, '^81 is set in the memory 82 (4) In the temporary area (the following tributary storage department), it is often updated in order to temporarily store and report in the traffic, that is, the speed i or the position data, motion picture data, etc., when accidents or accidents, abnormalities occur, etc. Before and after the phenomenon occurs, Qian Guan is in the above-mentioned shape of the slit 82 (also referred to as the regular record data storage unit) and recorded. The above phenomenon is equivalent to the case where the acceleration (deceleration) indicated by the acceleration data exceeds the predetermined reference value of 201211941, or the case where the time is constant, the power supply of the vehicle is cut off, and the vehicle speed of the door opening and closing. In the case of the upper limit speed or higher, the price rises by $, , ^, depending on the phenomenon that occurs, only with the presence or absence of the brakes, etc., in the case of the lower limit speed or more, the feed record is recorded, and as far as possible, no benefit is obtained. As a trigger point, that is, in the case of data recording, 'there is also a learning function. Whether the person has experienced a near-accident or a second incident, the above-mentioned notification agency 4 notifies the driver of the driving or not (for example, the above-mentioned correctness from the driver can grasp the driver's respect with a certain degree =). By repeating this action, changing the reference value, etc., the indirect j and the predetermined acceleration are learned. ",, the transfer of the _ person's unique present, the recorded state data, the class is classified according to the importance of the record should be recorded. The situation at the time of recording is weighted, shape, low importance The person starts to delete and records the new quantity to reach the limit and the status data recorded so regularly, 2, 4. Send to the analysis _ heart (not shown;), or pass the two 'wirelessly pass=, send to The splitting heart can threaten the mechanism 7 and analyze it afterwards. & is the <3. vehicle dynamic behavior analyzing device> of the analyzing device 200> the vehicle dynamic behavior analyzing device 2〇〇, obtained from the device 1〇〇 The dynamic image group extracts the other vehicles with the driving record of the dog's sin (hereinafter, the old car is also equipped with the established speed image to support the post-mortem analysis. The dynamic behavior of the vehicle is like a car.) The configuration includes a CPU, a memory, and a specific setting of a computer, and a dedicated computer, and the CPU and the periphery are provided in accordance with the dynamic behavior analysis program of the above-mentioned Kee's two-way, AD converter, and the like. The car 3 that is remembered in the area is played as 'played as The data receiving unit 20 operates to generate an image, as shown in the figure, to the edge extracting unit 2〇2, and to select an area generating unit: a function like the bedding storage unit D1 and the horizontal part of the vehicle. ==^踪10 201211941 The following 'With Figure 4~11, the singer Dong, +.々μ ^ data receiving unit 201, connected to D2 '201~204 and action. The dynamic image data of the storage unit (9) recorded the regular record of 1GG The data storage unit m (Fig. 4, step ==== ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The horizontal edge extracting unit 2G2 stored in the moving image data storage unit, as shown in FIG. 5, uses the following 7-stage image (for example, the following side is used (the 3 side is = (1) by the Caony edge detector). Edge detection; ίϊ ί;;;! 3^^; (3) In addition, I have a variety of methods for thinking about the extraction of lateral edges.
令以此種手法抽取的橫向邊緣的長度為^，該橫向邊 所指^_，域W1重疊的面積為v t Iin/Ii〇>th4，視為^ 棱向邊緣X存在於前方區域。另—方面，當WIiD _時，視為= 橫^邊緣X林細翁區域W1，⑽其㈣生麵區域為^ 的橫向邊緣X排除在外。另外，％係狀的常數。在此斤★田义 方區域(指定區域)W1，係將事故對象車不得行驶的區域排除U 11 201211941 =成之區域’如圖6所示，錢影像圖框的巾心、軸係對稱，從 ^既定亨置具有一定寬度’從該既定位置，下方係逐漸 、革’略王逆漏斗形狀。另外，前方區域Wl亦可不限於 延漏斗形狀，亦可為其他形狀。 產生部203 ’如圖7所示，取得橫向邊緣抽取部2〇2 ^ 件所抽取的橫向邊緣龍，以該橫向邊緣X為基 ㈣產生含有其他車輛的全部或—部分之選 ，步驟S4)。具體而言選擇區域產生部203，以含 W2。其1°，的?式於上下方向產生複數個矩形的選擇區域 標示處理所/配的H:包含構成橫向邊緣X的晝素’以及由 所抽從由橫向邊緣抽取釋 [數學式1] ram > 1) 在此OShSl。又p係線段上的點 [數學式2]Let the length of the lateral edge extracted by this method be ^, and the area of the lateral edge indicated by ^_, the area W1 overlaps is v t Iin / Ii 〇 > th4, and it is considered that the edge edge X exists in the front region. On the other hand, when WIiD _, it is regarded as = horizontal ^ edge X Lin Weng area W1, (10) its (four) green area X is excluded from the lateral edge X. In addition, the constant of the % system. In this area, the Tianyifang area (designated area) W1 excludes the area where the accident vehicle is not allowed to travel. U 11 201211941 = the area of the area is as shown in Fig. 6. The center of the money image frame is symmetric. From ^ predetermined set to have a certain width 'from the established position, the lower part is gradually, the leather 'slightly reversed the funnel shape. Further, the front region W1 is not limited to the shape of the funnel, and may have other shapes. As shown in FIG. 7, the generating unit 203' obtains the lateral edge dragon extracted by the lateral edge extracting unit 2, and generates all or part of the selection of other vehicles based on the lateral edge X (step S4). . Specifically, the area generating unit 203 is selected to include W2. The 1°, the pattern of the selection of the plurality of rectangles in the up and down direction indicates that the processing/matching H: contains the pixels forming the lateral edge X and the extraction from the lateral edge is extracted [Math 1] ram > 1) In this OShSl. And the point on the p-line segment [Math 2]
CmiD ~ ^ 少 min + G Ο^π + G O^rr Ληώ) )’min •y ^min ). (2) (3) 在此的OSaSl係用以考慮以適 之效果而橫向邊緣線段的兩端縮小 向邊^ X的原始據鏡 {>^in，ymax}之間取得水平線段的y座標之 =t2$ 1係用以在 而選擇區域產生部203,暫定。 上部_)、最下職鎌桿.接地ff事故縣車的最 邊·後車紅部)來產生選籠域W2。^巾央^(後玻璃的下 203 ’如圖7所示，於基準影像圖框“ =5收選擇區或產生部 (横向邊緣X)保持為上邊的正方形、保持^邊 12 201211941 - 持為連結右邊中點與左邊中點的線的正 線段產生3個正方形的區域(選擇 1個水平 S影像，㈣之情形， 接著’選擇區域產生部2〇3，如圖8所- 除：理:選 2 為接地點，則在接地^存在於右將選擇區域W2的下it視 =身 遠離本身車輛ν時;的上方時(亦即’十分 ‘；B，_ 域上二 [數學式3]CmiD ~ ^ Less min + G Ο^π + GO^rr Ληώ) )'min •y ^min ). (2) (3) The OSaSl is used here to consider the appropriate effect and the ends of the lateral edge segments The y coordinate of the horizontal line segment = t2$1 obtained between the original esthetic mirror {>^in, ymax} of the edge ^X is used to select the region generating portion 203, tentatively. The upper part _), the lowest post mast. The ground ff accident county car, the rear side of the rear car red part) to generate the selected cage field W2. ^巾央^(the lower 203' of the rear glass is as shown in Fig. 7, in the reference image frame "=5, the selection area or the generation part (lateral edge X) is kept as the upper square, and the ^ edge 12 201211941 - The positive line segment connecting the line on the right midpoint and the left midpoint produces three square regions (select one horizontal S image, (4), then 'select region generation portion 2〇3, as shown in Fig. 8>: If 2 is selected as the grounding point, then the grounding ^ exists in the right side of the selection area W2 and the body is away from the vehicle ν; that is, 'very'; B, _ domain 2 [Math 3]
Hose = height)7I 2σ,2 C4) 令左高度隔：絲]，在影像平面中， =時:控制二r基 度ί使用W將各選擇區域W2的下 [數學式4] (5) ^nira) — Base ^，β@ γ係接收的選擇區域W2的尺寸 其他，選擇區域產生部203，亦檢杳在卜、喜沾—:Ρ==γ 刻t的時侯過於偏向下方之情形(下述式—$時 13 201211941 盾 [數學式5] (6) (7) y, > (tkh x height) yb < (^, X height) 框:内以上產可生t婁 近車輛判斷部204。、°° 5 選擇區域資料輸出到靠 的各部2〇4’針對由上述選擇區域產生部2〇3所取r 框回= 度(圖4,步驟V)。产内疋否含有事故對象車之可靠 準影像圖框Ft為起始點的的影像圖框，係回溯到以基 的40圖框。又，可靠;;始點的過去之時間順序連續 性之條絲算出。持有的特 t (1)能叫細的物^= (2)在選擇區域W2内含有充足的角點。 ίϊίτΐ縱的選擇區域W2每在回溯時刻時變得更小。 叫二胸蚊：因轉故縣車财後背面 動僅藉由; 内：，，雜，所 呆持在附近内的獨特性，靠微弱的追蹤功能，ΐ蹤 而_域W2是否具有十分複雜的紋理 弈2所、他f車的後背面觀測到充足的角點，但另一方面，於 先則所述的路面或天空的區域大多未能觀測戦足的角點。 14 201211941 寸變更履歷而追縱結果所取得的選擇區域W2的尺 2〇4，^由粒檢，’本實施形態的靠近車柄判斷部 運算子所進行的ϋ斤f订的追鞭(用於條件⑴、⑺)、藉由Harris 之麵内以及角點移動量 的_。 U _其結果來异出可靠度，進行事故對象車 9所^有作為選擇區域追縱部的功能，如圖 W9 ^ t皮進订追縱的結果，得到任意的時刻t-i的選擇 W2的左上頂點的座標(χ 二的于域 度綱(圖4 ’步祕υ。然而，在圖9中，長 前2像圖框&以及2圖框前的影另外 Ϊίί波 使用概度計算，不偵測選擇區域W2的钕採二 轉私動以及形態變化，*是侧選擇區域’ j 以及尺寸變更，來追縱選擇區域W2。 々'、、文理的平行移動 域w==；r2G4,贿滿树2贿件的選擇區 [數學式6] (8) (9) 条件（a) : ..Hose = height) 7I 2σ, 2 C4) Let the left height be separated: silk], in the image plane, = when: control two r base degrees ί use W to select each region W2 [Math 4] (5) ^ Nira) - Base ^, β@ γ is the size of the selected area W2 received, and the selection area generating unit 203 also checks the case where the smear, the smear, and the smear are too biased downward ( The following formula - $13 131111941 Shield [Math 5] (6) (7) y, > (tkh x height) yb < (^, X height) Box: The above production can be found near the vehicle The part 204., °° 5 is selected to output the area data to the respective parts 2〇4', and the r-frame is returned by the selected area generating unit 2〇3 (Fig. 4, step V). The reliable image frame Ft of the target vehicle is the image frame of the starting point, which is traced back to the 40 frame of the base. Moreover, it is reliable; the starting point of the past time sequence continuity is calculated. The special t (1) can be called fine object ^= (2) There are sufficient corner points in the selection area W2. The selection area W2 of the ϊ ΐ ΐ 变得 变得 每 每 每 选择 选择 选择 选择 选择 选择 选择 选择 选择 选择 选择The back of the county Only by means of; inside:,, miscellaneous, the uniqueness of staying in the vicinity, relying on weak tracking function, _ _ domain W2 has a very complicated texture game 2, his rear view of the car In the case of a sufficient corner point, on the other hand, most of the road surface or the sky area described above failed to observe the corner point of the foot. 14 201211941 Inch change history and the result area of the selected area W2 obtained by the tracking result 2 〇4,^ is checked by the grain, and the amount of the whip (for conditions (1), (7)), the amount of movement in the face of the Harris, and the corner point by the operator of the handle judgment unit of the present embodiment _. U _ The result is different from the reliability, and the vehicle 9 is used as the function of selecting the area tracking unit. As shown in Fig. W9, the result of the tracking is obtained, and the selection of any time ti is obtained. The coordinates of the upper left vertex of W2 (χ2's domain degree (Figure 4 'step secret. However, in Figure 9, the long front 2 image frame & and the 2 front frame shadow Ϊίί wave use the probation Calculate, do not detect the selection of the region W2, the second rotation of the private movement and the morphological change, * is the side selection area 'j And the size change, to trace the selection area W2. 々 ',, the parallel movement domain of the arts and sciences w ==; r2G4, bribe full tree 2 bribes selection area [Math 6] (8) (9) Conditions (a) : ..
，卜I G G = Σ siSn(s(t -j + \)~s(t- j)) ·/·=】 (10) 条件⑹··“祕.. 條件⑻，判斷選擇區域W2的面積是否持 判斷選擇區域W2是否在影像圖框的中心附近移動。]，條件(b) ’ 接著，靠近車輛判斷部204，在所有時刻藉ς h° 進行之角點抽取’從選擇區域W2内的特徵 子： 點的移動量或是角點的移動量之總和等，來ϋ^、、息和、角 4’步叫在此，所謂「可靠度」，係 15 201211941 是否存在車輛，亦即選擇區域 是該其他車翻是否靠近本身車m其他車麵進行追縱，或 於每個選擇區域 ’軚。 輸出，在最大可靠度^臨界值m靠，輪觸部2〇4 象車區域醇_象輕輕域，將_該事故對 f驟况）。又，靠近車輛判斷器等輸出機構(圖4， 析貧料儲存到分析資料儲存部〇2。將样員不上述判斷結果的分 <本實施形態的效果> ，據依如此構成的本實郷態 易由其他車輛的最上部(車 動L仃為分析系統，容 央部(後破璃的下邊或後車J最部^二險^戈接地部}或是令 f黃向邊緣X設定選擇區域W2，X’從而藉由 自動追蹤其他車輕。藉此 擇區域W2，藉此可 的功能之車輛靠近判斷部204、之追縱^有=選擇區域追蹤部 顯示既絲ϋ行為㈣他柄。;觸是否含有 更可省去操縱者的判斷。 可減輕刼縱者的負擔， 不積 斷部 204， 概度計算來谓測選擇區诚、轉移動以及形態變化’而是使用 〈其他的選擇區域…. 另外，本發明並不限於上述實施形態。 例如’亦能以可推定方式來構成前車 使用在攝影機_中_ 的車間距離。此時 前提停件係五肋子_純饤例如以下_序。但是 二气d)(令影像平面的原點為左上’ y軸下方為I方 平ί)θ(ϋί^ίί角方向為正’仰角方向為負，G。係與路‘ 的下邊所持有之H標中的Υ座標—Μ)⑽選擇區域W2 16 201211941 此時，首先從y〇ff=fxtan( I θ丨）求出無窮遠線與影像中心 的偏移M。接著為了判定接地點是硫於無窮遠線以 c 的式子求出消失點所持有的y座標y.f。 田以下 [數學式7] 1 (11) =3^. -sign(e)xy(!ff. dist 下，係符號函數’在k &之情形，接地點位於無 窮退、，、下在yinf >yin之情形，接地點位於無窮遠線之上。 而若接地點在絲遠線以下，則由以下的式子求出車間距離 [數學式8] dist: 0 + sign(yin - vjxtan' (12) 生3 i選形態中選擇區域產生部雖根據橫向邊緣X產 不限於此，亦可於上下方向產生2個選 間距i歲太直二/Λ個的選擇區域W2。從如此求出的車 身車Ϊ的加速度等算出追撞危險度，二 馬駛的駕敬评價或駕駿技術提升之資料。 、二 域W^t尺ΐίίϊί”車辅靠近判斷部，亦可僅藉由選擇區 的尺寸fe更履歷，來判斷事故對象車的突然靠近。 區域!i部實施形態中，車輛靠近判斷部雖具有做為選擇 追縱部，；輛近判斷部與選擇區域 事故對象車輛制携區域賴㈣追果來判斷 判斷事故縣車是否突然靠近。果』諸旦面上’由操縱者 更力上，在上述貫施形態t ’雖將行車纪錄哭士凡认义A 士 、丁車.己錄為或其他車用攝影機設於後破璃等本身車輛的 17 201211941 後部來拍攝後方’以_有無顯魏定動 其他，本發明並不限於上述實施能，二、後方車幸兩。 圍内可實施各種變更，此乃無須待言/心，在未脫離其要旨的範 [產業上利用性] 、° 根據本發明，判斷在從行車記錄器等 是否含有顯示例如追撞事故事:以得之動 寺原因之既定動態行為之其他車輛，在次忍外事件事例 的負擔，更可正確地進行判斷，或是省去其判g僅可減輕操縱者 【圖式簡單說明】 統之圖。 圖 圖 圖 圖 3係顯示同-形態的車輛動態行為分析裝置的 4係顯不同-形態的車姉態行為分㈣置_作^呈 意圖 圖5係顯示同-形態的橫向邊緣抽取部的橫向邊緣彳貞測之示 抽取同—織的指定區域及考_魏域之橫向邊緣 圖7係顯示同一形態的選擇區域產生部的選擇區 例之示意圖。. / 一圖8係顯示同一形感的選擇區域產生部的選擇區域之限縮之 示意圖。 、'’ 圖9係顯示同一形態的選擇區域追蹤部的追蹤態樣之示意圖。 【主要元件符號說明】 2 :機殼 2b .溝槽 3:偵測機構 4:通知機構 18 201211941 5:輸入機構 6:通信機構 7:可拆卸式記錄機構 8:資訊處理機構 31 :攝影機構 32 :加速度感測器, IGG = Σ siSn(s(t -j + \)~s(t- j)) ·/·=] (10) Condition (6)··“Secret.. Condition (8), determine whether the area of the selected area W2 is It is judged whether or not the selected area W2 is moved near the center of the image frame.], condition (b) ' Next, near the vehicle determination unit 204, the corner point extracted by h° at all times is extracted from the feature in the selection area W2. : The amount of movement of the point or the sum of the movements of the corner points, etc., ϋ^,, and the angle of the 4' step is called here, the so-called "reliability", is 15 201211941 whether there is a vehicle, that is, the selected area is Whether the other car is turned around is close to the other car surface of the car, or in each selected area '軚. Output, at the maximum reliability ^ critical value m depends, the wheel contact 2 〇 4 like the car area alcohol _ like a light field, will _ the accident on the f situation). Further, it is close to an output mechanism such as a vehicle determiner (Fig. 4, the analysis of the lean material is stored in the analysis data storage unit 〇2. The result of the above-described determination result is not the result of the present embodiment> The real state is easy to be the uppermost part of other vehicles (the vehicle is L仃 for the analysis system, the central part of the body (the lower part of the rear glass or the rear part of the car, the second part of the second part of the ^^^^^^) or the yellow to the edge X By setting the selection areas W2, X', the other vehicles are automatically tracked. By selecting the area W2, the vehicle approaching determination unit 204 of the available functions can be displayed by the selection area tracking unit (4). He handles; whether the touch contains more judgments that can save the operator. It can reduce the burden of the escapement, without accumulating the part 204, and calculating the degree of selection, the change of the area, the movement, and the morphological change. Other selection areas. The present invention is not limited to the above-described embodiments. For example, it is also possible to construct the inter-vehicle distance of the front vehicle used in the camera_in the _ _ _ pure For example, the following _ order. But the second gas d) (to make the image flat The origin is the upper left 'the lower side of the y axis is the I square ί) θ (ϋί^ίί the direction of the angle is positive 'the elevation angle is negative, G. is the Υ coordinate of the H mark held by the lower side of the road' - Μ (10) Selecting the region W2 16 201211941 At this time, first, the offset M of the infinity line from the image center is obtained from y〇ff=fxtan(I θ丨). Then, in order to determine that the ground point is sulfur at the infinity line, Find the y coordinate yf held by the vanishing point. Field below [Math 7] 1 (11) = 3^. -sign(e)xy(!ff. dist, the symbolic function 'in k & In the case where the grounding point is infinity, and in the case of yinf >yin, the grounding point is above the infinity line. If the grounding point is below the far line, the inter-vehicle distance is determined by the following equation [mathematics] Desc: 0 + sign(yin - vjxtan' (12) The selected area generating unit in the selected form is not limited to this depending on the lateral edge X, and two selection pitches may be generated in the up and down direction. The second selection area W2 is calculated from the acceleration of the vehicle body ram obtained in this way, and the information on the risk of collision, the evaluation of the driving of the second horse, or the improvement of the driving technique. In the second embodiment, the vehicle approaching determination unit may determine the sudden approach of the accident target vehicle only by the size of the selection area, and the vehicle approaching determination unit has the vehicle approach determination unit. As the selection of the chasing department, the vehicle near the judgment department and the selected area accident object vehicle manufacturing area (4) to trace the fruit to judge whether the accident county car suddenly approached. Fruit "the surface of the face" is more powerful by the operator The above-mentioned form of application t', although the driving record, the crying singer, recognizes A, Ding, has been recorded as or other vehicle camera, and is located in the rear of the vehicle, such as the rear of the glass, etc. Others, the present invention is not limited to the above implementation, and second, the rear car is two. It is possible to carry out various changes in the circumference, and it is not necessary to say that it is a heart, and it is not deviated from the gist of the industry [industrial useability], and according to the present invention, it is judged whether or not the display of the driving recorder or the like contains, for example, a chasing story: Other vehicles with the established dynamic behavior of the reason of the movement of the temple, in the case of the case of the external event, can be judged correctly, or the judgment can be omitted, only the operator can be lightened [simple description of the schema] Figure. Figure 3 is a diagram showing the 4-series of the same-form vehicle dynamic behavior analysis device. The morphological behavior of the vehicle is divided into four. The fourth embodiment shows the lateral direction of the same-edge lateral edge extraction. The marginal speculative display extracts the designated region of the same weave and the lateral edge of the test_wei domain. FIG. 7 shows a schematic diagram of a selection region of the selection region generating portion of the same form. / Fig. 8 is a schematic view showing the limitation of the selection region of the selection region generating portion of the same shape. '' Figure 9 is a schematic diagram showing the tracking pattern of the selection area tracking unit of the same form. [Main component symbol description] 2: Case 2b. Groove 3: Detection mechanism 4: Notification mechanism 18 201211941 5: Input mechanism 6: Communication mechanism 7: Detachable recording mechanism 8: Information processing mechanism 31: Photographing mechanism 32 : Acceleration sensor
33 :位置感測器 41 ： LED33 : Position sensor 41 : LED
81 ： CPU 82 :内部記憶體 83 : I/O缓衝迴路 1〇〇 :行車記錄器 200 :車輛動態行為分析裝置 201 :資料接收部 202 :橫向邊緣抽取部 203 :選擇區域產生部 204 :靠近車輛判斷部(選擇區域追蹤部) CN :連接器 D1 :動態影像資料儲存部 D2 :分析資料儲存部 Ft :基準影像圖框 Fw、Ft_2 :影像圖框 V :本身車輛(汽車） W1 :指定區域 W2 :選擇區域 X :橫向邊緣 S1〜S5-3 :步驟. 1981 : CPU 82 : Internal memory 83 : I/O buffer circuit 1 : Driving recorder 200 : Vehicle dynamic behavior analysis device 201 : Data receiving unit 202 : Lateral edge extraction unit 203 : Selection area generation unit 204 : Close Vehicle determination unit (selection area tracking unit) CN: Connector D1: Motion picture data storage unit D2: Analysis data storage unit Ft: Reference image frame Fw, Ft_2: Image frame V: Vehicle (automobile) W1: Designated area W2: Select area X: Horizontal edge S1~S5-3: Step. 19