201124941 六、發明說明： 【發明所屬之技術領域】 本發明總體上係、關於—種用於使用主動輪廓技術來編輯 數位地圖之方法，且更特定言之，本發明係關於用於在兩 個緊密隔開之主動輪廓表示一塑形、現實世界物體之部分 的情況下防止主動輪廓在操縱期間未對準的方法。 此專利文件之揭示内容之一部分含有受版權保護的材 料。《擁有者並不反對副本複製該專利文件或該專利揭 不内今之任一者，因為在專利及商標局專利槽案或記錄中 有所發表，但除此以外版權擁有者無條件保留所有版權權 益。 【先前技術】 已知在某些應用中使用知識驅動之主動輪廊連同數位地 圖創建及精緻化。主動輪虛在處m a m 勁輪廓係廣闊用於影像及視訊分段以 二用=縱。可在變分方法之架構下公式化該等主動輪 /。土本原理為藉由計算歐拉_拉格朗日方程式卿仏201124941 VI. Description of the Invention: TECHNICAL FIELD OF THE INVENTION The present invention generally relates to a method for editing a digital map using active contouring techniques, and more particularly, the present invention relates to A closely spaced active profile represents a method of preventing misalignment of the active profile during manipulation in the case of a shaped, real-world object. Part of the disclosure of this patent document contains material that is subject to copyright protection. "The Owner does not object to the reproduction of the patent document or the disclosure of any one of the patents, as it is published in the Patent and Trademark Office patent case or record, but otherwise the copyright owner retains all copyrights unconditionally. rights and interests. [Prior Art] Knowledge-driven active corridors are known to be used in some applications along with digital map creation and refinement. The active wheel is in the virtual area and is widely used for image and video segmentation. These drive wheels can be formulated under the framework of the variational method. The principle of the soil is calculated by calculating the Euler_Lagrange equation
Lagrange equation)之草轡公而田 & )系變刀而用—能量準則（通常包含區 域函數及邊界函數兩者）建構一偏微分方 咖根據可被視為該能量準則之—下降方向的某-速度場 而改變流速曲線之形狀。所 密閉-内部區域的㈣，而曲線主動輪廟為 ., , A 放」曲線常常係與常見於 0 床、車道及其他特徵相關聯。通過-最佳 化處理程序，通過使取決於邊緣的特徵之-邊界積分最小 化而堪動主動曲線（閉合或開放）朝向-影像之邊緣。 I457l6.doc 201124941 例如，在於1999年1月19日頒予Ren〇uard等人的美國專 利第5,862,245號揭* -種部分使用—主動 操取-數位地圖影像中之-輪廓的方法。輪_取:= 使用一能量最小化函數來使一主動輪廓（在此情況中為開 放）變形的一反覆處理程序。此等及其他先前技術方法依 賴於在待擷取輪廓附近進入該數位地圖的一初始輪廓（即 一所謂的種子輪廓）之輸入。此初始輪廓可呈可為如一直 線或曲線標記一樣簡單的一多邊形跡線之形式。該反覆處 理程序使該初始輪廓變形直至其找到對應於一能量最小值 的一平衡位置，該平衡位置界定待擷取輪廓之跡線之一部 分。 連同數位地圖編輯活動，該（等）初始或種子輪廓可獲得 自或衍生自探測資料或—空間上可能不準確的現有數㈣ 圖。探測資料輸人係、獲得自低成本定位系統及具有整合 GPS功能的手持裝置及行動電話，用於遞增產生及/或更新 數位地圖之目0。來自此探測f料的待處理輸人由呈一標 準ASCII爭流之形式的已記錄Gps跡線組成，大多數現有 GPS裝置支援該ASCII串流。輸出可為m圖表的一 跡線，該定向圖表具有與行料間f訊相_的節點及邊 緣或連結。依規則間隔產生該等節點或探測器位置的該探 測資料可經由無線（例如’蜂巢式）傳輸、網際網路上傳或 藉由其他手段而被傳輸至—收集服務或其他地圖製作或資 料分析服務。 圖1描繪一例示性數位地圖或其區段，如可屬於本發明 145716.doc 201124941 之該等方法所利用之類型。展示此類蜇之—數位地圖，其 包含可包括分車道高速公路或雙向車道的—級公路連同表 示更窄及較低級別之行車道的二級公路及三級公路的階層 架構。圊2展示一例示性的具備Gps能力之導航裝置， 其包含為呈現導航輔助之目的而料街道之—網路的一顯 示螢幕U或圖形使用者介面。當然，該等街道係由該數位 地圖提供且真實地對應於陸地之一區域。此中所描繪之街 道及公路被含於該數位地圖資料庫作為（當然）現實生活中 對應公路及車道的網路元件。 :常可期望或必須基於某一新資料集評價一數位地圖之 =確性個該新資料集可包括前面所提及的探測器痕跡資 I4，:：個二級數位地圖、-空間上不可靠的數位地圖或其 類型之參考資訊。有時可期望基於實際來 : 地圖的一新資料集而評價探測 :位 4Γ、,舍-^ 早释性。圖4A至圖 種方式，以該方式呈已收集探測器痕跡…的 -新資料集可用作為源幾何形狀，以為使： 法來重新對準-數㈣圖中之 動輪廓方 始或—在此等視圖中，展示供-初 几件Μ，該車道具有沿其長度的現實 ^-網路 度轉彎。展示隨時間推移而收集及根據;中急轉、9〇 聚集的重疊探測器痕跡。使用熟知之路元件Μ所 器痕跡可被衍生成形成如圓4B中所—…该等探測 源幾何形狀，該源幾何形狀表示：主動輪廟16的-習_的該初始或種子輪廊。 主動輪廓操縱練 1457l6.doc 201124941 如以上所提及，如此領域所熟知，—主動輪廓μ取得受 指派之内能量（intent _咖及外能量咖㈣。 -主動輪廓16之㈣㈣僅由該（等）主動㈣本身之形狀 決定且係完全獨立於任何外部源（在先前給出的實例中為 網路段14)。指派給該主動輪扉16的該所謂外能量係由該 外部源決定。該外部源本身未被指派任何能量：其僅充當The Lagrange equation) is used to change the knife—the energy criterion (usually including both the region function and the boundary function) to construct a partial differential based on the direction of the decline that can be considered as the energy criterion. Change the shape of the flow curve for a certain velocity field. The closed-internal area (4), while the curve active wheel temple is ., , A A. The curve is often associated with the common 0 bed, lane and other features. Through the optimization process, the active curve (closed or open) is oriented toward the edge of the image by minimizing the boundary integral of the edge-dependent features. For example, U.S. Patent No. 5,862,245 issued to Ren〇uard et al., issued on Jan. 19, 1999, which is incorporated herein by reference. Wheel_fetch: = A repetitive handler that uses an energy minimization function to deform an active contour (open in this case). These and other prior art methods rely on the input of an initial contour (i.e., a so-called seed contour) entering the digital map near the contour to be captured. This initial profile can be in the form of a polygonal trace that can be as simple as a straight or curved mark. The repeating process deforms the initial contour until it finds an equilibrium position corresponding to an energy minimum that defines a portion of the trace of the contour to be captured. In conjunction with digital map editing activities, the (or) initial or seed profile may be obtained or derived from probe data or an existing number (four) map that may be spatially inaccurate. The detection data input system, the low-cost positioning system and the handheld device with integrated GPS function and mobile phone are used to incrementally generate and/or update the digital map of the digital map. The pending input from this probe consists of recorded Gps traces in the form of a standard ASCII contention, and most existing GPS devices support the ASCII stream. The output can be a trace of the m-chart, which has nodes and edges or links to the inter-material. The probe data generating the locations of the nodes or detectors at regular intervals may be transmitted to the collection service or other map making or data analysis service via wireless (eg, 'honeycomb') transmission, internet upload, or by other means. . 1 depicts an exemplary digital map or a section thereof, such as the type utilized by such methods that may be subject to the invention 145716.doc 201124941. This type of —-digit map is displayed, which includes a class road that can include a lane-by-lane highway or a two-way lane, along with a hierarchy of secondary and tertiary roads that represent narrower and lower-level lanes.圊 2 shows an exemplary GPS-enabled navigation device that includes a display screen U or a graphical user interface for the purpose of presenting navigation aids. Of course, the street systems are provided by the digital map and truly correspond to one of the terrestrial areas. The streets and roads depicted here are included in the digital map database as (of course) the real-world network components for roads and lanes. : It is often desirable or necessary to evaluate a digital map based on a new data set. = The new data set may include the detector traces I4 mentioned above, :: a second-level map, - spatially unreliable Reference information for a digital map or its type. Sometimes it can be expected to evaluate the probe based on a new dataset of the map: bit 4Γ, 舍-^ early release. Figure 4A to the mode in which the new data set of the collected detector traces can be used as the source geometry, so that: the method is to realign the number of the moving contours in the (four) graph or - here In the view, the display is for the first few pieces, and the lane has a realistic ^-network turn along its length. Shows the traces of overlapping detectors that are collected over time and based on; Using known path elements, the traces can be derived to form the source geometry as in circle 4B, which represents the initial or seed rim of the drive wheel temple. Active contour manipulation practice 1457l6.doc 201124941 As mentioned above, as is well known in the art, the active contour μ obtains the assigned internal energy (intent _ coffee and external energy coffee (4). - active contour 16 (four) (four) only by this (etc. The shape of the active (four) itself is determined and is completely independent of any external source (network segment 14 in the previously given example). The so-called external energy system assigned to the active rim 16 is determined by the external source. The source itself is not assigned any energy: it only acts as
產生該速度場之源。該主動輪㈣之外能量接著係由此I 動輪扉16如何被安置在由該外部源所界定的該速度場令決 定(在大多數實施方案中此為主動輪廊控制點之若干位置 處的所有速㈣值之總和）。因此，在此實財該網路元 件14決定該速度場，該速度場轉而決定該主動輪廊16之外 能量。 一主動輪廓操縱策略企圖藉由通過設法降低該主動輪廓 16之内月b里及降低其外能量（其係（經由該速度場）由網路元 件η決；而使該主動輪廓16反覆地調整至更期望之位 置，從而相對於該網路元件14重新定位該主動輪靡16。已 隨時間推移被改善且㈣有效率的此等技術可導致開放型 主動輪廓16擬合於該網路元件14，如圖財所示。因此， 通k應用熟知之主動輪廓處理程序，該初始主動輪廊16可 很好地擬合於該網路元件14。自此初肖，可繼續進一步分 析步驟及編輯步驟，包含該網路元件14之潛在重新對準或 新為路TL件之增加等等。在先前實例中，—探測線（以某 方式或另一方式衍生自探測資料）為用以產生初始輪廓 的忒源幾何形狀，該初始輪廓在由一網路元件1 6所決定之 145716.doc 201124941 一速度場中被重新定位。 叱位反過來，其亦可為：一網路元件 16可為用作為一初如< 於内 始輪廓的该源幾何形狀，該初始輪廓 ㈣測資料(某另一外能量源)所決定之一速度場中被重新 疋位自然地’兩項技術均表示此概念之有效及可行的實 施方案。 幻耳 在該新資料集產生彼此相對緊密隔開且形成一特定形狀 的一或多個主動輪廓時出現_特定問I此問題可出現 (例如）在兩個靠近的開放主動輪廓形成現實中之某—固定 =體之各部分的時候或出現在一閉合輪廊單獨形成— .疋4何形狀的時候。例如’如圖5A中所示，第一主動輪 廓18及第二主動輪廓2〇可對應於包括分車道或雙向車^ 型之-總體上南北走向高速公路的一網路元件22。該第一 主動輪磨18及該第二主動輪廒 _ 土 動輪廓20表不各自車道内之交通 〜，5亥4主動輪廓1 8、20可P夕-4 ώ # 了已何生自探測資料。同樣地’ 二主動輪廓24及第四主動輪_對應於_總體上東西走 向的網路元件28。該等主動輪廒 ’廓18' 20、24、26以某種程 =未經加工或未經處理的格式呈現，即在相對於該數位 地圖破最佳化前。如圖5Α中清 . 〃吓兄忍第一主動輪廓18 ^第二主動輪廊2〇之間建立一特別顯著的平行關係，且 =三主動輪廓24與該第四主動輪廓26之間建立一特別顯 者的平行關係。在此實例中之該等平行關係表示現實中之 一固定形狀物體（即雙向車道公路）之各部分。 使用習知之最佳化技術，各主動輪靡an、％係 經個別及單獨最佳化，此（在此特定領域中）常常導致不良 145716.doc 201124941 對準。例如，圖5B展示由於習知最佳化步驟而在該網路元 件22上的該第—主動輪㈣及該第二主動輪廓2G之無意及 不合意合併或萎縮。同樣地，該第三主動輪㈣及該第四 主動輪廟26在東西網路元件28之—區段上已萎縮或合併在 一起。此等結果總體上並不纟意且導致細節<一丟失、增 力爲輯難度及通*非常需要對主動輪靡處理程序進行手動 干預’以便幫助维持緊密隔開之主動輪紅間之某些相對 形狀。類似之未對準情形亦可出現在其他情況中。例如， 發生在—圓環(即，-駕駛圓環)之區域内的-或多個主動 輪癖亦可成為在最佳化處理程序期間之不良對準的犧牲品 且需要手動干預以重新調節無意之結果。 ,主動輪廓之最佳化處理程序中之此等不足同樣發生在被 田作源幾何形狀（-主動輪廟係擬合於該源幾何形狀）的新 資料集為探測資料或一現有數位地圖的時候。相應地，在 此項技術中需要一種用於使用主動輪廓操縱技術來重新對 準-數位地圖中之網路元件使得可在最佳化處理程序期間 維持某些合意輪廓及/或形狀的改良方法。 【發明内容】 本發明精由提供一種用·^你m ;使用主動輪廓操縱技術以用 f新對準—數位地圖中之特徵的方法而克Μ前技術之 等不足及缺點。提供對應於現實中之陸地之—區域的— 位地圖。該數位地圖且右 /、有至少一對應於現實中之一特徵 網路元件。亦提供對應於盥 、/、邊網路件之特徵相同的現 中之特徵的一新資料集。兮雄1 °玄..同路或該新資料集被當作具 145716.doc 201124941 相互隔開之至少第— 主動輪廓及第二主動輪廟見 形成-相對形狀的H , 動輪廓且於被此間 輪廓及該第二主動於 之省第—主動 兩了依據特定應用而對庳於兮亥網敗一 件或對應於該新資料隹π+ 了應於及網路π L才4集。同時該網路及該新資料集之另— 者被當作一外能晉、'β γ 力 r .丨 "、（eXternal energy s〇uixe)。内能量值 (mtemal energy val 董值 祜浐浪仏兮笛 ）及外此里值（灿咖energy value) 该第—主動輪廟及該第二主動輪廊之每-者，且 一速度場被指派給外能量源。 改良方案包括：將士女贫 將5亥第一主動輪廓耦合至該第二主 廊，而同時使用— 祛m 一士 動輪廓操縱策略來相對於外能量源最 動輪廟及該第：主動輪廓之位置，其中在最 佳化處理程序期間不改變該第一主動輪廓與該第二主動輪 廓之間之相對形狀的情况下，藉由設法降低與外能量源有 關的該第-主動輪廓及該第二主動輪廓之各別内能量，而 使相對於該外能量源的該第一主動輪廊及該第二主動輪廊 之位置一起反覆地調整至更期望之位置。 相應地’ 5亥第一主動輪磨及該第二主動輪廟（及如果該 第三主動輪廟及該第四主動輪廓等等適用）之各者均具有 所指派的-内能量及一外能量。在外部速度場中最佳化該 等主動輪廓之位置。此速度場可由該等網路元件界定，使 得該等主動輪廊（來自探測資料或某一其他類型之資料集） 根據該等網路元件之位置而被最佳化。或者，反過來，其 亦可為：言亥等主純冑可為沿由該探測資料所界定之一外 部速度場移動的網路段（即，該第一主動輪廓及該第二主 145716.doc -10· 201124941 動輪廓之初始位置可與該網路元件之位置相同）。 本發明藉由使兩個主動輪廓聚集成一 實現網路幾何形狀之-特定部分之模型化，：中在最 期間维持正被模型化的該網路幾何形狀之該特定部= - 口疋屬/±。此與其巾主純廓均被個料佳化且導致不' 合意對準情況的先前技術之技術形成鮮明對比。 在本發明之—實施财，模龍具有兩條平行公一 雙向車道或分車道公路。萨由 #由使兩條車道彼此察覺且將兩 者一起處置成為其位置被最佳化的一個主動輪廓，而使令 兩條f道(各車道均係用-習知之開放主動輪廓表示)之間 呆持固定。以此方式’在最佳化之所有階段期間維 夺對應於現貫世界之情況的兩條平行車道之間之固定距 離。在另-實施例中’本發明之該等方法用一已輕合主動 輪磨之鄰近道路來模型化一圆環或迴車圓環環，其中在最 佳化期間被維持中心部分之環狀。當然，許多其他應用將 為熟習此項技術者明白，且可以此方式操縱多於或少於兩 個的主動輪廓’以便在最佳化處理程序期間維持現實中之 某些物體之形狀。 【實施方式】 在連同以下實施方式及附圖一起考慮時本發明之此等及 其他特徵及優點將變得更容易理解。 立參考圖丨’其中在所有視圖中相同數字指示相同或對應 4分’在圖3中示意性展示一種藉由使用主動輪靡而最佳 化源幾何形狀之方法。源幾何形狀功能塊3〇表示一新資 145716.doc 201124941 料集。此新資料集可（例如）源 區域的一數位地圖類型。或者勺〜於現實中之陸地之一 資料集可衍生自探測資料或某_^括言亥源幾何形狀的該新 本文中之許多相關内容將該探二:::乂應再次注意雖然 將網路特徵歸為外能量源’ "為該源幾何形狀及 表達。實際上，該源幾何形狀為、*為此等一般概念之一種 反覆最佳化&新继 ."、破變成主動輪廓且被直接 汉復敢佳化成一新幾何形狀的 恢 資料被視為比一現有數位地 / °因此’在該探測 π数位地圖中之網路 下，網路特徵將被當作該源幾 :更了靠的隋况 外於互％ AL Ab θ 戍仃形狀且該探測資料被當作 。外…原功能塊32表示可源自一數位地圖、r n 戈來自相關源之某_其他類型的更可靠資料 集。數位地圖含有對應於 貝τ之特徵（如公路、道路戎 其他此類特徵）的至少一個網 一 百萬計的㈣元件。 ^件且通u數以千計或 源幾何形狀提供該-或多個主動輪廓，根據主動輪廟模 型化領域之技術者所熟知之技術而使該等主動輪廊均指派 有一内能量值及-外能量值。内能量係僅衍生自該源幾何 形狀’且外能量係衍生自外能量源。與本發明有關的該等 主動輪廓係相互隔開且彼此間形成一相對形狀。該等主動 輪廟對應於-特定網路元件，且可（例如）模型化公路之分 車道運轉情況或通過一迴車圓環的車輛之移動。 使用已知之技術，在功能塊34中進行該等主動輪廓之最 佳化。在此步驟中，係根據與一外能量源之已指派速度場 有關的該等主動輪廓之能量值最佳化該等主動輪廓之位 145716.doc -12· 201124941 1疋塊3 6表示一主動輪廓操縱處理程序之反覆性質， =主動㈣操縱處理程序在運行至其自然結論時產生該源 歲何形狀之最佳化38。如果探測器痕跡被當作外能量源， 則3亥等主動輪廓源自該等網路元件，反之亦然。 匕等It況下，一種用於將總是衍生自外能量源之外部 速又易的外此置指派給該等主動輪廓之方法可包括使用 具有梯度值的光柵影卜在此等情況τ，光柵影像内之像 素之間之梯度值可被視為外力°此等光柵影像可由-數位 地圖或探㈣料組成°例如’該等網路元件可以對應於導 出之車輛位置概率密度的光柵影像之格式呈現。 現轉至圖6 ’顯示第一主動輪廓4〇及第二主動輪廓42， 該等主動輪廓包含沿該等主動輪磨遞增關的複數個各別 控制點44、46。該等控制點或形狀點表示彼此間具有干預 段的節.點，該等干預段形成與熟知之技術一致的邊緣及連 結。根據本發明之該等方法，在最佳化處理程序期間該第 一主動輪廓40係耦合至該第二主動輪廓42。因此，根據本 發明之最佳化與該等輪廓40、42係耦合或固定在一起大體 上成為-單-主動㉟磨同時發±。用ffi6中 <延伸於相對 控制點44、46之間的點線式箭頭表示該等主動輪廓4〇、42 之耦合。因此，使用一主動輪廓操縱策略，其中相對於一 網路元件最佳化該第—主動輪㈣及該第二主動輪廊U之 位置’以一反覆調整方式一起移動此等已耦合輪廓仂、 42，以實現相對於外能量源的一更期望之位置或若干位 置。如先前所陳述’一主動控制操縱策略包含設法相對於 I457I6.doc -13· 201124941 外旎量源之速度場降低該等主動輪廓40、42之各自内能量 的步驟。藉由使該第-主動輪廓40與該第二主動輪廊_ 合在-起，可在最佳化處理程序期間維持此等輪廊之間之 相對形狀。換而言之，以圖6中之實例為例，該等主動輪 廊4〇、42之間之形狀係由其等之總體上平行性質界定^ 被保存，因為各輪廓40、42均非被單獨或獨立最佳化。在 該等主動輪廓40、42係耦合或聯繫在一起後，其等將相互 一致或協調-致地經歷最佳化處理程序，但沿相同方向及 相同距離移動使得該等主動輪廊4〇、42之間之特定相對形 狀在最佳化處理程序期間不改變。 最佳地，輕合步驟包含使該第一主動輪靡4〇中之一個控 制點44幾乎直接連結至該第 矛一王勁輪屏中之一對應控制點 。可期望耦合在控制點之各自 分㈢王動輪屝中相互靠近的該 44 ’⑼此m因此’藉由維持已連結控制點 广間之一固定間隔，可相對於外能量源操縱此等主 動輪廓40、42且不改變由哕忐彻 殳由6亥兩個主動輪廓4〇、42所形成之 重要特定形狀。 ^芬如圖7中所示’如果使用本發明之新賴搞合處理 1來最佳化主動輪廓18，、2〇’及24，、26，，則用各自主動 化:：已Γ’、24.、26’所表示之平行車道未萎縮且在最佳 4匕期間已维持隼j 之 s固疋距離，該等車道與現實世界 更緊卜致。因此’各輪廓係报好地對準於其等之 各自外能量源，在此實例曰 、 22、28。 …專外…包括網路元件 145716.doc -14- 201124941 因此，在可用兩個單獨、平行主動輪廓來模型化—雙向 車道的情況下，本發明之原理係尤其有利。使用先前^ 之技巧，各車道獨立地最佳化其位置。例如在圖5B中展示 不合意之結果，其中通過力吸引一條車道，另一車道亦為 如此。當雙向車道的未對準大於兩車道間的固定距離時： 很可能其中一車道會自行重新定位於另—車道之預定位置 上。使用習知之主動輪廟方法不太可能自此位置移動此不 良定位車道’因為其外能量將（臨時）增加以移動橫穿兩條 車道之間的高能量區。因此，此導致車道之一萎縮或在兩 條車道之間之固定距離未被維持時沿雙向車道伸展。由於 先前技術之技巧之結果，已最佳化之代表圖並不對應於現 實物質世界。 然而’錢如先前所描述之本發明之原理，該兩個平行 主動輪廓係經麵合。該等主動輪廓-起被最佳化成一單一 主動輪摩。換而言之，該等車道之位置並非單獨而是—起 破最佳化…輪廓4G之每-控制點44或形狀點係與另_主 動輪廓42之—對應控制點46連結。在最佳化期間移動—個 點44時，其對應點46與該點以以相同方式移動。此意味著 保證兩個輪廓40、42之間之坧雜+置u 巨離在最佳化期間總是被保持 固定。因此，且當然在—雔闩由 — 又向車道之貫施例中，在最佳化 延伸主動輪廓位置期間保捭 仟待用—延伸之主動輪廓所模型化 :又向車道’該延伸之主動輪廊由彼此間具有一固定距 =兩個平行開放主動輪摩組成。然而，更廣泛言之，本 發明針對用包括許多已耦人 °動輪廓之一延伸主動輪廓來 145716.doc 15 201124941 ’其中在最佳化期間 定。因此，此等部分 模型化源網路幾何形狀之一特定部分 控制點44、46之間之某些距離保持固 之相對形狀將不改變。 在此等技術之另一實例中，例如 J如’可以與先前實例中之 多條平行路之間之固定距離相同的士 * η 仰u的方式維持一圓環之中心 部分之形狀。轉至圖8更具體令夕 丹篮。之，以48總體指示現實生 活中表示一圓環的一網路元件。自 目（例如）可衍生自探測資 料或另一幾何形狀之一網路的—新咨 枓集引進複數個主動 輪廓50、52、54及56。一閉合型Φ說认史 1主動輪廓5 8對應於該網路 元件48之該圓環部分。圖8中展+ 土 ^上 _ Τ展不未經加工或最佳化前狀 況下的此等主動輪廓5〇至58。 圖9展示該等主動輪廓5()至58連同其等之欲人控制點。 為維持主動輪廓58之圓環部分之相對形狀，其控制點係與 -對角相對控制點麵合或連結，如圖9Α中之虛線式箭頭所 表不此耦合在最佳化處理程序初始前完成使得在使用前 述之能量值處理來最佳化該主動輪廓之位置後原始圓形形 狀將被維持。或者，圖9Β描述不同耦合策略，藉此橫向連 結該主動輪廓58周圍之控制點以實現相同或類似之結果。 當然’可採用其他具有不同程度效果的搞纟或連結策略。 熟習此項技術者亦將容易瞭解本發明可應用於除雙向車道 及圓環外的網路幾何形狀，但可用以使用此新穎耦合方法 來、隹持任—或多個主動輪廓之間之相對形狀。圖10為如圖 中 視圖’但其展示最佳化後之主動輪廓50至58，其 中或圓％主動輪廓58之耦合已產生期望之最終結果。 1457l6.doc 16- 201124941 就可與-閉合主動輪廓相關聯之圓環及其他特徵而+， 與圓9A及圖_所示_樣，搞合處理可發生在—單:閉 合主動輪廓之特定㈣點的範@内。因此，在^背離本發 明之精神的情況下可不必在所有情況下輕合兩個截然不二 的主動輪廓。類似地，在適當情況下可在最佳化期間搞合 兩個以上主動輪廓。 因此’ «本發明之該等原理’並非藉由被個別最佳化 的-組主動輪廓’而是藉由使該等主動輪廓聚集成一已搞 合之主動輪廓，而完成模型化一網路幾何形狀之一特定部 分，其中在最佳化期間維持正被模型化的網路幾何形狀之 特定部分之某些固定屬，卜根據已知之主動輪廊策略，可 用一主動輪廓來模型化充當一源幾何形狀的公路網路之一 伸展’該主動輪摩為藉由反覆地設法降低其能量而最佳化 其位置的一直線段。能量可被描述為包括兩個主要部分： 6X法保持忒輪廓之原始形狀的内能量及試圖移動該原始形 狀的外能量。因此，衍生自源幾何形狀的内能量阻止呈極 突然彎曲、伸展、收縮及類似情形之形式的形狀變形。另 一方面，衍生自外部源資料的外能量鼓勵該主動輪廓移動 至一更理想之位置。此等技術在針對主動輪廓之文獻中已 被詳細地描述且可直接應用於増加此等新穎耦合技術的本 發明之上下文中。 已根據相關之合法標準描述前面提及之發明，因此描述 為例示性而非限制屬性。揭示實施例之變動及修都可為熟 習此項技術者明白且落入本發明之範圍内。 145716.doc -17· 201124941 【圖式簡單說明】 圖1為一數位地圖之—例示性描繪，該數位地圖包括作 為本發明中所使用類型之數位向量地圖之一形式之代表圖 的一道路網路； 圖2為一攜帶式導航系統之一例示視圖，該攜帶式導航 系統包含用於呈現地圖資料資訊的一顯示螢幕； 圖3為彳田述使用主動輪廓來最佳化一源幾何形狀的一例 示性流程圖； 圖4A至圖4C表示網路模型化之一進展情況，其中呈探 測器痕跡之形式的源資料係用以衍生一開放型主動輪廓， 接著使用習知之技術來相對於網路元件最佳化該開放型主 動輪廟； 圖5A及圖5B表示在兩個緊密隔開之開放型主動輪廓（諸 如可衍生自（例如）探測資料）被個別最佳化且導致某些區段 中之不良對準及重疊時的先前技術結果； 圖6為兩個緊密隔開之主動輪廓之一簡化視圖，如該兩 個緊达、隔開之主動輪廓可衍生自（例如）探測資料且均包含 3亥專主動輪廓隔開的複數個控制點，且使用本發明之方 法來使此等控制點相互耦合； 圖7為如圖5A及圖5B中之一視圖，但繪示之方式為可以 該方式在最佳化處理程序期間維持該等主動輪廓之間之最 佳化的、相對形狀以便產生較佳結果的方式； 圖8繪示與具有一圓環網路元件之一數位地圖有關且在 最佳化處理程序前的初始主動輪靡之形成； 145716.doc -18- 201124941 圖9A及圖9B描述與圖8之實施例相關聯之主 制點之輕合，藉此圖9八描述控制點之一直動人輪廟及控 描述控制點之一替代橫向輕合；及 直4合且圖犯 圖Π)為如圖8中之一視圖，其展示 最佳化處理程序期間使形成迴車圓严、為以该方式在 其圓形形狀以產生一可接灸 衣的主動輪廓維持具有 【主要元件符號說明】 10 導航裝置 12 顯示螢幕 14 網路元件 16 主動輪廓 18 第一主動輪靡 18' 主動輪廓 20 第二主動輪廓 20' 主動輪廓 22 網路元件 24 第三主動輪靡 24， 主動輪廓 26 第四主動輪摩 26， 主動輪廓 28 網路元件 40 弟一·主動輪靡 42 第二主動輪靡 44 控制點 145716.doc '19. 201124941 46 控制點 48 網路元件 50 主動輪廓 5 2 主動輪廓 54 主動輪廓 56 主動輪廓 58 閉合型主動輪廓 145716.doc · 20 -The source of the velocity field is generated. The energy outside the drive wheel (four) is then determined by how the I-roll rim 16 is placed at the speed field defined by the external source (in most embodiments this is at several locations of the active ship control point) The sum of all speed (four) values). Therefore, the network element 14 determines the velocity field, which in turn determines the energy outside the active corridor 16. An active contour steering strategy attempts to repeatedly adjust the active contour 16 by attempting to reduce the inner moon b within the active contour 16 and reduce its external energy (via the velocity field) by the network element η; To a more desirable position to reposition the drive rim 16 relative to the network element 14. These techniques have been improved over time and (iv) efficient such techniques may result in the open active profile 16 being fitted to the network element 14. As shown in the figure, therefore, the well-known active contour handler is applied, and the initial active corridor 16 can be well fitted to the network element 14. Since then, further analysis steps can be continued. The editing step includes potential realignment of the network element 14 or an addition of a new TL component, etc. In the previous example, the probe line (derived from the probe data in some way or another) is used to generate The source geometry of the initial profile, which is repositioned in a velocity field determined by a network component 16 145716.doc 201124941. The clamp may, in turn, be: a network component 16 for Used as an initial geometry as the source geometry of the inner contour, the initial contour (four) measured data (some other external energy source) is re-clamped in one of the velocity fields. An effective and feasible implementation of this concept. The magic ear occurs when the new data set produces one or more active contours that are relatively closely spaced from one another and form a particular shape. This problem can occur (for example) in two A close open active contour forms a certain one in reality - when it is fixed = part of the body or when it is formed separately in a closed gallery - when it is shaped, for example, as shown in Figure 5A, the first initiative The contour 18 and the second active contour 2〇 may correspond to a network element 22 including a lane-by-lane or two-way vehicle-total north-south highway. The first driving wheel mill 18 and the second driving wheel rim _ The soil contour 20 shows the traffic in the respective lanes~, 5 Hai 4 active contours 1, 8 and 20 can be P Xi-4 ώ # The self-detection data. The same 'two active contour 24 and the fourth driving wheel _ Corresponding to _ overall east-west network elements 28. The active rims 'profiles 18' 20, 24, 26 are presented in a certain process = unprocessed or unprocessed format, ie before being optimized relative to the digital map. A particularly significant parallel relationship is established between the first active contour 18 and the second active contour 2, and a particularly prominent parallel between the three active contours 24 and the fourth active contour 26 is established. The parallel relationship in this example represents the real part of one of the fixed-shaped objects (ie, the two-way lane road). Using the conventional optimization techniques, each of the active rims, % is individually and individually Jiahua, this (in this particular area) often leads to poor 145716.doc 201124941 alignment. For example, Figure 5B shows the unintentional and unintentional merging or shrinking of the first-drive wheel (four) and the second active profile 2G on the network element 22 due to the conventional optimization step. Similarly, the third drive wheel (4) and the fourth drive wheel temple 26 have been shrunk or merged together on the section of the East-West network component 28. These results are generally unsatisfactory and lead to details <one loss, increase in difficulty, and difficulty* require manual intervention in the active rim handler to help maintain a tightly spaced drive. Some relative shapes. A similar misalignment can also occur in other situations. For example, - or multiple active rims occurring in the area of the - ring (ie, - driving the ring) may also be a victim of poor alignment during the optimization process and require manual intervention to re-adjust the unintentional The result. The deficiencies in the optimized contouring of the active contours also occur when the new dataset that is used by the Matthew source geometry (the active wheel temple is fitted to the source geometry) is the probe data or an existing digital map. Accordingly, there is a need in the art for an improved method for realigning network elements in a digital map using active contour manipulation techniques such that certain desirable contours and/or shapes can be maintained during the optimization process. . SUMMARY OF THE INVENTION The present invention provides a deficiencies and shortcomings of using pre-existing techniques by using active contour manipulation techniques to use f-alignment-features in digital maps. Provide a map of the location corresponding to the land-reality of the real world. The digital map and right /, at least one corresponds to one of the feature network elements in reality. A new data set corresponding to the current characteristics of the 盥, /, and side network components is also provided.兮雄1 °玄..同路 or the new data set is treated as 145716.doc 201124941 at least the first - active contour and the second active wheel temple formation - the relative shape of the H, the moving contour and The contour and the second initiative in the province-active two according to the specific application for the 兮 兮 网 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 。 。 。 。 。 。 At the same time, the network and the other new data set are treated as a foreign energy, 'β γ force r .丨 ", (eXternal energy s〇uixe). The internal energy value (mtemal energy val) and the external value (can energy value) of the first - the active wheel temple and the second active wheel gallery, and a speed field is Assigned to an external energy source. The improvement scheme includes: coupling the first active contour of the Shiyang poor to the second main gallery, and simultaneously using the 祛m-single contour manipulation strategy to compare the outer moving energy temple with the outer energy source and the first active contour a position in which, by not changing the relative shape between the first active contour and the second active contour during the optimization process, by attempting to reduce the first active contour associated with the external energy source and the first The respective internal energy of the active profile is adjusted to a more desirable position along with the positions of the first active turret and the second active rim relative to the external energy source. Correspondingly, each of the 5th first active wheel grinding and the second driving wheel temple (and if the third driving wheel temple and the fourth active contour are applicable) has the assigned internal energy and an outer energy. The position of the active contours is optimized in the external velocity field. The velocity field can be defined by the network elements such that the active lanes (from probe data or some other type of data set) are optimized based on the location of the network elements. Or, conversely, it may also be that the main pure 言, such as Yan Hai, may be a network segment that moves along an external velocity field defined by the probe data (ie, the first active contour and the second primary 145716.doc) -10· 201124941 The initial position of the contour can be the same as the location of the network element). The present invention enables the specific part of the network geometry being modeled to be maintained during the most period by aggregating the two active contours into a specific part of the network geometry. ±. This is in stark contrast to the prior art techniques in which the main profile of the towel is optimized and results in a non-consistent alignment. In the implementation of the present invention, the model dragon has two parallel public one-way lanes or lane-by-lane roads. Sa is made by making the two lanes aware of each other and treating the two together as an active contour whose position is optimized, so that the two lanes (each lane is used - the open active contour representation of the known) Staying fixed. In this way, the fixed distance between two parallel lanes corresponding to the situation in the current world is maintained during all phases of optimization. In another embodiment, the method of the present invention models a ring or a return ring with a nearby road that has been lightly driven to rotate, wherein the center portion is maintained during the optimization. . Of course, many other applications will be apparent to those skilled in the art, and more or less than two active profiles' can be manipulated in this manner to maintain the shape of certain objects in reality during the optimization process. [Embodiment] These and other features and advantages of the present invention will become more apparent from the consideration of the appended claims. The reference numeral 丨' in which the same numerals indicate the same or corresponding 4 points in all views is schematically illustrated in Fig. 3 by a method of optimizing the source geometry by using a driving rim. The source geometry function block 3〇 represents a new asset 145716.doc 201124941 material set. This new data set can, for example, be a digital map type of the source area. Or scoop ~ a piece of data in the real land can be derived from the probe data or a number of related content in the new text of a _ ^ 言 亥 亥 源 几何 几何 : : : : : : : : : : : : : : : : : : : : : : : The road feature is classified as the external energy source ' " for the source geometry and expression. In fact, the source geometry is, and the general concept of this is a kind of repetitive optimization & new success.", broken into active contours and directly restored by Han Fu into a new geometric shape For a network in the π-digit map, the network feature will be treated as the source number: the more dependent state is outside the mutual % AL Ab θ 且 shape and This probe data is taken as. External...The original function block 32 represents a more reliable data set that can be derived from a digital map and from a related source. The digital map contains at least one million (four) components corresponding to the characteristics of the Bayer (such as roads, roads, and other such features). Providing the one or more active profiles in thousands or source geometries, each of which is assigned an internal energy value according to techniques well known to those skilled in the art of driving wheel modeling. - External energy value. The internal energy system is derived only from the source geometry' and the external energy source is derived from an external energy source. The active contours associated with the present invention are spaced apart from each other and form an opposing shape with each other. The active wheel temples correspond to specific network elements and can, for example, model the lanes of a highway or the movement of a vehicle through a return circle. The optimization of these active contours is performed in function block 34 using known techniques. In this step, the active contours are optimized according to the energy values of the active contours associated with the assigned velocity field of an external energy source. 145716.doc -12· 201124941 1 疋 block 3 6 represents an active The repetitive nature of the contour manipulation handler, the active (four) manipulation handler produces an optimization of the shape of the source when it is run to its natural conclusions38. If the detector trace is treated as an external energy source, the active contours such as 3H are derived from the network components and vice versa. In the case of It, etc., a method for assigning an external velocity that is always derived from an external energy source to the active contours may include using a raster image having gradient values in such cases τ, The gradient values between pixels within the raster image can be considered as external forces. Such raster images can be composed of a digital map or a quadrature material. For example, the network elements can correspond to the raster image of the derived vehicle position probability density. The format is rendered. Turning now to Figure 6' shows a first active profile 4〇 and a second active profile 42, the active profiles comprising a plurality of individual control points 44, 46 that are incrementally closed along the active wheel mills. The control points or shape points represent the points of the intervention segment with respect to each other, and the intervention segments form edges and connections consistent with well-known techniques. In accordance with the methods of the present invention, the first active profile 40 is coupled to the second active profile 42 during an optimization process. Thus, the optimization according to the present invention is coupled or fixed together with the contours 40, 42 and is generally a simultaneous - single-active 35-grinding ±. The coupling of the active profiles 4, 42 is indicated by a dotted line arrow extending between the relative control points 44, 46 in ffi6. Therefore, an active contour manipulation strategy is used in which the position of the first driving wheel (four) and the second driving wheel U is optimized relative to a network element to move the coupled contours together in a repeated adjustment manner. 42 to achieve a more desirable location or locations relative to an external source of energy. As previously stated, an active control maneuver strategy includes the step of reducing the respective internal energy of the active profiles 40, 42 relative to the velocity field of the source of the I457I6.doc -13·201124941. By bringing the first active profile 40 into engagement with the second active rim, the relative shape between the lands can be maintained during the optimization process. In other words, taking the example in FIG. 6 as an example, the shapes between the active rims 4, 42 are defined by their overall parallel nature, because each contour 40, 42 is not Optimized individually or independently. After the active profiles 40, 42 are coupled or linked together, they will undergo an optimization process consistently or in concert with each other, but move in the same direction and at the same distance such that the active corridors, The particular relative shape between 42 does not change during the optimization process. Preferably, the step of coupling includes causing one of the control points 44 of the first drive rim 4 to be directly coupled to one of the corresponding control points of the first rifle. It may be desirable to couple the 44' (9) so close together in each of the control points (3) of the king's rims. By maintaining a fixed interval between the connected control points, the active contours can be manipulated relative to the external energy source. 40, 42 does not change the important specific shape formed by the two active contours 4 〇, 42 by 亥 殳. ^fen as shown in Figure 7 'If you use the new processing of the present invention 1 to optimize the active contours 18, 2〇' and 24, 26, then use each to activate:: already Γ', 24. The parallel lanes indicated by 26' are not shrunk and have maintained a s solid distance during the best period of 4, which is more closely related to the real world. Thus, the contours are reported to be aligned with their respective external energy sources, in this example 曰, 22, 28. ...Special...including network elements 145716.doc -14- 201124941 Thus, the principles of the present invention are particularly advantageous where two separate, parallel active profiles can be used to model a two-way lane. Using the previous ^ technique, each lane independently optimizes its position. For example, the undesired result is shown in Figure 5B, where one lane is attracted by force and the other lane is also the same. When the misalignment of the two-way lane is greater than the fixed distance between the two lanes: It is likely that one of the lanes will reposition itself at the predetermined position of the other lane. It is not possible to move this defective positioning lane from this position using the conventional active wheel temple method because its external energy will (temporarily) increase to move across the high energy zone between the two lanes. Therefore, this causes one of the lanes to shrink or to extend in the two-way lane when the fixed distance between the two lanes is not maintained. As a result of prior art techniques, the representative map that has been optimized does not correspond to the real material world. However, as with the principles of the invention as previously described, the two parallel active profiles are face to face. These active contours are optimized to be a single active wheel. In other words, the positions of the lanes are not separate but the optimization is optimized... each of the contours 4G - the control points 44 or the shape points are coupled to the corresponding control points 46 of the other - active contours 42. When the point 44 is moved during the optimization, its corresponding point 46 moves with the point in the same manner. This means that the noisy + set u large distance between the two profiles 40, 42 is always kept fixed during the optimization. Therefore, and of course, in the case of the 雔 由 — 又 又 又 又 — — 车道 — 车道 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳The porch consists of a fixed distance = two parallel open active wheels. More broadly, however, the present invention is directed to extending an active contour with one of a number of coupled human contours 145716.doc 15 201124941 'where the optimization period is determined. Thus, some of these portions of the modeled source network geometry will remain unchanged for certain distances between control points 44, 46 that remain solid. In another example of such techniques, for example, J such as 'can maintain the shape of the central portion of a ring in a manner that has the same fixed distance from the plurality of parallel paths in the previous example. Go to Figure 8 to make it more specific. Thus, a network element representing a ring in reality is indicated by 48 overall. A new set of active profiles 50, 52, 54 and 56 is introduced from the target, for example, from a network of probes or one of the other geometries. A closed type Φ says that the active profile 5 8 corresponds to the annular portion of the network element 48. In Fig. 8, the active contours 5〇 to 58 in the case of unexamined or optimized conditions are shown. Figure 9 shows the active profiles 5() through 58 along with their desire points. In order to maintain the relative shape of the annular portion of the active contour 58, the control point is opposite or connected to the diagonal control point, as shown by the dotted arrow in Fig. 9Α, before the optimization process is started. Completion is such that the original circular shape will be maintained after the aforementioned energy value processing is used to optimize the position of the active contour. Alternatively, Figure 9A depicts different coupling strategies whereby the control points around the active contour 58 are laterally coupled to achieve the same or similar results. Of course, other methods of messing up or linking with different degrees of effect can be used. Those skilled in the art will also readily appreciate that the present invention can be applied to network geometries other than two-way lanes and toroids, but can be used to use this novel coupling method to maintain the relative-orientation of multiple active profiles. shape. Figure 10 is a view of the figure' but showing an optimized active profile 50 to 58 in which the coupling of the circle or active profile 58 has produced the desired end result. 1457l6.doc 16- 201124941 The ring and other features associated with the -close active profile +, and the circle 9A and the figure _, the engagement process can occur in - single: the specific of the closed active profile (4) Point Fan @内. Therefore, it is not necessary to combine two distinct active contours in all cases without departing from the spirit of the invention. Similarly, more than two active profiles can be combined during the optimization period, where appropriate. Therefore, 'the principles of the present invention' are not modeled by a group of active contours that are individually optimized, but by integrating the active contours into an engaged active contour. A specific portion of a shape in which certain fixed genus of a particular portion of the network geometry being modeled is maintained during optimization, and may be modeled as a source by an active contour according to known active porch strategies One of the geometric road networks stretches 'the active wheel is a straight line segment that optimizes its position by repeatedly trying to reduce its energy. Energy can be described as comprising two main parts: the 6X method maintains the internal energy of the original shape of the 忒 profile and the external energy that attempts to move the original shape. Thus, the internal energy derived from the source geometry prevents shape deformation in the form of extremely sudden bending, stretching, shrinkage, and the like. On the other hand, the external energy derived from the external source material encourages the active contour to move to a more desirable location. These techniques have been described in detail in the literature for active contours and are directly applicable to the context of the present invention in which such novel coupling techniques are added. The aforementioned invention has been described in accordance with the relevant legal standards, and thus the description is illustrative and not limiting. Variations and modifications of the disclosed embodiments are apparent to those skilled in the art and are within the scope of the invention. 145716.doc -17· 201124941 [Simplified Schematic] FIG. 1 is an exemplary depiction of a digital map including a road network as a representative map in the form of a digital vector map of the type used in the present invention. Figure 2 is an exemplary view of a portable navigation system including a display screen for presenting map information; Figure 3 is an example of using an active contour to optimize a source geometry. An exemplary flow chart; Figures 4A-4C illustrate one of the evolutions of network modeling in which source data in the form of detector traces is used to derive an open active profile, which is then compared to the network using conventional techniques. The road component optimizes the open drive wheel temple; Figures 5A and 5B show that the two closely spaced open active profiles (such as may be derived from, for example, probe data) are individually optimized and result in certain zones Prior art results in poor alignment and overlap in the segment; Figure 6 is a simplified view of one of the two closely spaced active profiles, such as the two tightly spaced, spaced apart active profiles Generated from, for example, probing data and each comprising a plurality of control points separated by a dedicated contour, and using the method of the present invention to couple the control points to each other; FIG. 7 is one of FIGS. 5A and 5B View, but depicted in a manner that maintains an optimized relative shape between the active contours during the optimization process to produce better results; Figure 8 depicts and has a ring One of the network elements is digitally mapped and the initial active rim is formed prior to the optimization process; 145716.doc -18- 201124941 Figures 9A and 9B depict the main points associated with the embodiment of Figure 8. Lightly, by means of Figure 9-8, one of the control points of the control wheel and one of the control description points is used instead of the horizontal light combination; and the straight 4 and the figure is as shown in Figure 8, which shows the most During the optimization process, the formation of the carriage return is critical, and the active contour in this manner in its circular shape to produce a moxibustion garment is maintained. [Main component symbol description] 10 Navigation device 12 Display screen 14 Network element 16 Active contour 18 first Drive rim 18' active profile 20 second active profile 20' active profile 22 network component 24 third drive rim 24, active profile 26 fourth drive wheel 26, active profile 28 network component 40 brother one drive wheel靡42 Second drive rim 44 Control point 145716.doc '19. 201124941 46 Control point 48 Network element 50 Active contour 5 2 Active contour 54 Active contour 56 Active contour 58 Closed active contour 145716.doc · 20 -