201231927 六、發明說明： 【發明所屬之技術領域】 本發明大致上係關於一種用於分析探測資料之方法，且 更特定言之係關於-種將探測跡線匹配至_數位向量地圖 之方法。 【先前技術】 如圖1中所示之現實中的道路係用於維持地面上的交通 流。如圖1中所示，交通流可全部在一個方向上或雙向。 電子地圖（亦稱作數位地圖或數位向量地圖）越來越2地被 旅行者用於協助各種導航功能，諸如判定旅行者及/或車 輛之總體位置及定向；尋找目的地及地址；計算最佳路 線；及提供即時駕駛指導…數位地圆係經組態以儲存在 一座標系統内空間關聯之複數個線段。如圖2中所示之各 線段係由藉由形狀點連接在一起之一邊序列構成。 如圖3中所示，當與現實中的一道路關聯時，線段通常 表不道路中心線。在數位地圖内，該複數個線段係分別彼 此拓撲相連或不相連。一般而言，相連線段彼此鄰接使得 /〇著現貫中的道路之交通係通暢地自一線段直接行進至下 一線段。另一方面，拓撲不相連之線段不直接相連，且因 此沿著現實中的道路之交通一般無法在不相連的線段之間 直接行進。舉例而言，圖9描繪拓撲不相連之重疊線段 及L4。另一方面，線段L1、。及。係彼此拓撲相連。 由於收集及處理道路資訊成本較高，故數位地圖的製作 及更新係昂貴。在已創建-數位地圖之後，由於道路幾何 153368.doc 201231927 形狀隨時間而改變，故使_資訊保持時新成本較高。車 輛探測資料可用於使道路網路保持時新。車輛探測資料 (亦稱作探測跡線)係位置資料連同一時戳及可能的額外資 料（如速度、加速度、航向（heading)、精確度等）之一序 列。個別位置資料係稱作跡線黑占。一_跡線通常表示一 小汽車、腳踏車、行人等的移動。通常在二維或三維座標 系統中表示該等跡線點。該時戮可以一隱含方式表示或可 從該探測資料中省略。在其中僅需儲存第一跡線點之時間 的情況中，通常使用等距時間間隔。若並不關注時間，則 可省略時間資訊。因此’在最簡單的情況中，一探測跡線 僅含有位置資料。可使用不同的方法從探測資料產生一新 網路。可以容易地提煉及擴展一數位地圖系統中之一道路 網路的方式使用可獲得之新探測資料，前提是來自此資料 之跡線可適當地匹配至一地圖。 地圖匹配演算法對數位地圖製作者而言係一項關鍵技 術。幾乎所有的探測資料應用需要地圖匹配演算法，包含 屬性採擷（例如，速度量變曲線）、網路產生、網路提煉及 變化之探測。此外，每個導航器件中需要地圖匹配演算法 以探測其目前在導航地圖上之位置。已存在不同的地圖匹 配方法。一種方法可區分線上與離線地圖匹配演算法。對 於線上演算法，僅可獲得當前及先前的GpS點。與之相 較，離線演算法可額外地使用一些或甚至所有將來的GPS 點。 此外’一種方法可區分完整與不完整的地圖匹配演算 153368.doc 201231927 一完整的地圖匹配將各跡線點分配至任意線段。使用 此方法， 一跡線點可能與所匹配的線段相距甚遠。因此， 必須確保數位向量地圖係完整。若非如此，必須允許一跡 線點在任意情況下不會被分配至一線段。允許此等未匹配 點之演算法係歸類為不完整的地圖匹配技術。 針對不同的演算法類別’存在不同的地圖匹配方法。針 對增量網路產生’申請人已開發出在2009年10月22日所申 凊之PCT/EP2009/063937中更全面描述的n點地圖匹配技 術。此技術（亦稱作Viae Novae演算法）運行非常快速，且 在大多數情況下吾等會得到非常好的結果。但是，對於具 有短的中間道路元素的複雜十字路口或交又口，201231927 VI. Description of the Invention: TECHNICAL FIELD OF THE INVENTION The present invention generally relates to a method for analyzing probe data, and more particularly to a method of matching probe traces to a _ digit vector map. [Prior Art] The road in reality as shown in Fig. 1 is used to maintain traffic flow on the ground. As shown in Figure 1, the traffic flow can all be in one direction or in both directions. Electronic maps (also known as digital maps or digital vector maps) are increasingly used by travelers to assist with various navigation functions, such as determining the overall location and orientation of travelers and/or vehicles; finding destinations and addresses; Good route; and provide immediate driving guidance... The digital circle system is configured to store a plurality of line segments spatially associated within a standard system. The line segments as shown in Fig. 2 are formed by a sequence of edges joined by shape points. As shown in Figure 3, when associated with a road in reality, the line segment typically represents the road centerline. In a digital map, the plurality of line segments are respectively connected to each other or are not connected. In general, the connected segments are adjacent to each other such that the traffic of the existing road is smoothly traveling from one segment to the next. On the other hand, line segments that are not connected by the topology are not directly connected, and therefore traffic along the road in reality cannot generally travel directly between disconnected line segments. For example, Figure 9 depicts overlapping line segments and L4 with unconnected topologies. On the other hand, line segment L1. and. The systems are connected to each other in topology. Due to the high cost of collecting and processing road information, the production and updating of digital maps is expensive. After the created-digit map, since the shape of the road geometry 153368.doc 201231927 changes over time, the new cost is higher when the _ information is maintained. Vehicle detection data can be used to keep the road network up to date. Vehicle probing data (also known as probing traces) is a sequence of positional data with the same time stamp and possible additional information (such as speed, acceleration, heading, accuracy, etc.). Individual location data is called trace black. A _ trace usually indicates the movement of a car, bicycle, pedestrian, etc. These trace points are typically represented in a two- or three-dimensional coordinate system. The time may be indicated in an implicit manner or may be omitted from the probe data. In the case where only the first trace point needs to be stored, an equidistant time interval is typically used. If you don't pay attention to the time, you can omit the time information. Therefore, in the simplest case, a probe trace contains only location data. Different methods can be used to generate a new network from the probe data. The new probe data available can be easily refined and extended in one of the digital map systems, provided that the traces from this data are properly matched to a map. Map matching algorithms are a key technology for digital map makers. Almost all probing data applications require map matching algorithms, including attribute mining (eg, speed-volume curves), network generation, network refinement, and detection of changes. In addition, a map matching algorithm is required in each navigation device to detect its current location on the navigation map. There are different map matching methods. One method can distinguish between online and offline map matching algorithms. For online algorithms, only current and previous GpS points are available. In contrast, offline algorithms can additionally use some or even all future GPS points. In addition, a method can distinguish between complete and incomplete map matching calculus. 153368.doc 201231927 A complete map match assigns each trace point to any line segment. With this method, a trace point may be far from the line segment being matched. Therefore, you must ensure that the digital vector map is complete. If this is not the case, a trace point must be allowed to not be assigned to a line segment under any circumstances. Algorithms that allow such unmatched points are classified as incomplete map matching techniques. There are different map matching methods for different algorithm classes. For the generation of incremental networks, the applicant has developed an n-point map matching technique that is more fully described in PCT/EP2009/063937, filed on Oct. 22, 2009. This technique (also known as the Viae Novae algorithm) runs very fast and in most cases we get very good results. However, for complex intersections or intersections with short intermediate road elements,
Novae演算法有時無法總是傳達正確的結果。因此，需要 一新穎且經改良的地圖匹配演算法。此一經改良的演算法 必^員展現一咼匹配品質，在不完整的道路網路（例如，道 路缺失、拓撲連接缺失或錯誤等）情況下運行良好；且對 於抗受品質不佳的GPS跡線非常穩固（亦即，不佳的探測資 料），諸如具有點狀雲之跡線（圖4)、具有環狀之跡線（圖 5)、具有Z字形之跡線（圓6)及路面外跡線（圖7)，其等之所 一經改良的地圖匹 有經常發生在探測資料收集中。此外， 配演算法必須結合單向及雙向網路運行。 【發明内容】 本發明提供一種用於將自探測資料導出之跡線匹配至使 用自楝測資料導出之跡線之一數位地圖（諸如一數位向量 地圆）中之一或多個線段的方法。該數位地圖較佳為一= 153368.doc 201231927 通網路，其中該等線段表示一道路、 ^ 路徑或類似物之至少 一部分。 該方法包括下列步驟：提供-數位向量_，其係經組 態以儲存在-座標系統内空間關聯之複數個線段，該等線 段分別彼此拓撲相連或不相連；提供至少一探測跡線，其 具有定位在該數位向量地圖之座標系統内之複數個空間關 聯的跡線點；及在各跡線點之一特定距離（d)内判定各線段 之-匹配候選者。該方法之特徵為：產生以該等匹配候選 者作為節點之一圖形（26);及選擇該圖形之匹配候選者之 至少一路徑作為匹配結果（36)。 新的地圖匹配方法因其固有的高匹配品質而對於增量地 圖產生程序以及若干其他的應用非常有用。本發明產生該 等匹配候選者之一匹配圖形，該匹配圖形繼而允許使用圖 形理論方法以透過該圖形找到最佳路徑。該圖形之節點係 由若干匹配候選者所構成。該匹配圖形包括各跡線點之一 單獨層級。若一跡線點無匹配候選者，則圖形之一致的層 級不含節點。本發明之一核心概念係使用一匹配圖形表示 所有匹配候選者。將該圖形簡化並對其評分。該等最佳路 徑傳達匹配結果。本發明之概念可結合作為匹配準則之縱 向距離而使用。 【實施方式】 在結合下列詳細描述及隨附圖式考慮時，更易於理解本 發明之此等及其他特徵及優點。 參考圖式，其中在若干視圖中相同的元件符號指示相同 153368.doc 201231927 或對應的部件，圖8中展示根據本發明之地圖匹配程序之 一概述。一數位地圖20提供一道路網路之資料。該數位向 量地圖20係經組態以儲存在一座標系統内空間關聯之複數 個線段。該數位地圖20内之該等線段係分別彼此拓撲相連 或拓撲不相連。各線段含有至少一邊。當—線段具有複數 個邊時，形狀點將該等邊互連。各線段終止在一接點點 處。 一探測資料源22提供至少一但更普遍為大量探測跡線。 各跡線含有定位在數位向量地圖之座標系統内之空間分佈 的跡線點之一序列^ —此類型之探測跡線通常表示一小汽 車、腳踏車、行人·等之移動。地圖匹配係在創建跡線時判 疋一旅行者正在該數位地圖的哪條線段上行進之程序。此 係藉由考慮一探測跡線中之各點以識別匹配候選者及隨後 從該等候選者中間選出最佳的可能選項而完成。 在步驟24中，逐個考慮探測跡線中的點以在該數位向量 地圖中找到與其緊密相連之一或多個預期的線段。因此， 一匹配候選者為一給定跡線點至數位地圖中存在之一特定 線段之一可能的匹配結果。一跡線點可具有若干匹配候選 者。一種判定一跡線點之一匹配候選者集合之方法係選擇 一預定義的最大距離d内之數位向量地圖之所有線。針對 各個如此選擇之線段，計算跡線點至該線之投影。所有該 等投景;ί之集合係該跡線點之匹配候選者。 本發明之該方法的特徵為產生該等匹配候選者之一圖形 之〆驟，其由圖8令之功能塊26表示。為了促進理解主題 153368.doc 201231927 廣算法，可將該圖形視為具有在數目上對應於探測跡線中 之點數目的複數個循序層級。因此，若一給定探測跡線具 有十個點，則該圖形將具有十個層級。該圖形之各層級按 順序對應於S玄圖形中之其他層級，正如各跡線點按順序對 應於其他跡線點。因此，鄰近跡線點對應於該圖形中之鄰 近層級。理解圖形層級與跡線點之間的關係之另一方式係 將第一跡線點指派至第一層級；將第二跡線點指派至第二 層級...；將最後一跡線點指派至最後一層級。 各匹配候選者係在圖形中置於與其之跡線點對應的層級 處。舉例而言，若第一跡線點存在三個匹配候選者（意指 。亥第一跡線點處在與數位地圖中之三個不同的線段相距該 預定義的最大距離d内），則將所有三個匹配候選者置於該 圖形中的層級一處。 從各層級處之一或多個匹配候選者中間識別一匹配結 果:各跡線點之匹配結果係以最大可能性對應於該跡線點 之該數位向量地圖令的一點。各跡線點最多具有一匹配結 果。若跡線點具有處於數位向量地圖外部之某一可能性， 則可斷定該跡線點不具有-匹配結果。可用該點所屬之線 及沿著此線之測鏈數（亦稱作量測值）描述數位向量地圖中 之一點。或者，可用線及此線之一形狀點描述一數位向量 地圖中之—點。 各匹配結果被視作圖形中之一節點。一或多個路徑延伸 穿過該等節點’亦即穿過各層級令之_匹配候選者。該等 路徑係由若干節點序列構成’其，各兩個連續節點係由一 153368.doc 201231927 邊連接。各路徑在方向上將各層級中之—匹配候選者連接 至下一鄰近層級中之一匹配候選者，前提是各自的兩個匹 配候選者係彼此拓撲相連。以另一方式言之，各路徑不得 意在將拓撲不相連的線段彼此連接（亦即，表示無法彼此 橫越之現實中的道路之線段）。以此方式，若滿足下列條 件，則將兩個節點連接：υ匹配候選者屬於鄰近層級；2) 兩個投影點之間存在一（局域）路線；及3)任意上述路徑中 不含該路線之穿越形狀點。 為簡潔起見，應瞭解路徑為具有—線性順序之圖形的若 干特殊部分圖形。當產生一圖形時，該圖形將含有一個或 若干個路徑。一般而言，一圖形含有兩個不同類型之物 體：節點及邊。-邊為(經排序)一對節點。邊表示節點之 間的連接。該圆形之節點為匹配候選者。若兩個匹配候選 者之間存在-拓撲連接，則藉由一有向邊連接兩個節點。 藉此僅考慮局域拓撲連接。舉例而言，不應考慮經由一繞 =之拓撲連接。—單向地圖匹配亦應考慮拓撲連接之方向 一方面’一雙向地圖匹配程序不應考慮方向感。為 了找出可能㈣配⑽者之㈣局域拓 各匹配候選者表示之線段），可 由 尋距離限制之-選路演算法^(舉例由一最大搜 由任意適當的技術(舉例而言，包含步驟28處之 一坪後方法）簡化最初含有所有可能路經之一圖形。另一 Γ=Γ配候選者與對應跡線點間之-縱向距離大 於預疋義最大縱向距離夕祕士从 離之所有節點移除。在簡化後，可 I53368.doc 201231927 對各種路徑評分30,使得可從其間選擇-最佳師32該 評分較佳係基於節點所屬之最長路徑中之連接節點的數^ 及/或平均距離。可使用其他準則。在該評分步驟％期間 或之後’若得分低於其他，則可移除邊及/或節點。此將 進一步簡化該圖形》 基於砰分結果而相對簡單地從圖形中選擇最佳路徑（步 驟32)。在大多數情況中，所選的路徑已含有各層級之一 即點’其意指已可靠地找到各跡線點之—匹配結果^。但 是在少數情況中，路徑可能不含一些層級之節點。此可為 當數位向量地圖之兩條線之間存在一缺失拓冑連接之情 況。在此情況中（其中發現一路徑不含一些層級之節點）， 較佳地將具有與所選路徑之節點相同的層級之一些或所有 節點移除，其對應於圖8之流程圖中之步驟34。移除步 驟34將進一步簡化該圖形。在移除“後，對圖形重新評分 3〇並選擇下一最佳路徑（32)。重複此移除34、重新評分3〇 及選擇3 2序列直至已找到所有跡線點之可靠的匹配結果， 或直至圖形為空。通常僅需非常少的反覆。 一闡釋性實例係由圖9至圖13提供。在此，在一數位網 路中，現實中的道路係由假設已源自一數位地圖2〇之線段 L1至L4表示。線段u、。及以係在由接點38標記之一共 同交叉口處彼此拓撲相連。另一方面，線段L4係與u、 L2及L3拓撲不相連。對應於L4之現實中的道路在路線之 L1部分下方穿越而不允許交通與。、。或。之任意者直 接相互作用。圖1〇展示此等相同的線段以至“連同點卩丨至 153368.doc 201231927 P8所表示之一探測跡線。該探測跡線源自資料源22 » 匹配候選者選擇步驟24係使用一適當演算法完成。舉例 而言’演算法可基於各個各自的跡線點或節點（P1、P2、 P3、...P8)至若干網路元素（LI、L2、L3、L4)之距離與一 預定義臨限值d之對比。一跡線點p之一匹配候選者為其點 最接近P之一網路元素。一網路元素可為p至一線段、一接 點（例如，3 8)或一形狀點之正交投影。圖丨丨展示至由四條 線段L1至L4構成之一道路網路之跡線（p丨至p8)的所有匹配 候選者。 圖形產生步驟26主要為用於產生匹配候選者之一匹配圖 形之一演算法且在圖12中以表格形式予以描繪。此方法允 許使用圖形理論方法透過圖形找到可能的路徑且隨後從其 等中選出最佳者。該圖形之節點為若干個匹配候選者。該 匹配圖形包括針對各跡線點（ρι、p2、p3、 p8)之一單獨 層級。若-跡線點無匹配候選者，則圖形之一致的層級不 含節點。 圖13展示對應於圓u及圖12中之實例之一匹配圖形。因 此，該匹配圖形俗展千& ^ 係展不為一有向圖形，其中僅用一邊連接 鄰近層級之節點，前摇县嘿^ L、> 滿足上述條件。即，若兩個投影 點之間存在一路線（亦即，一 貫際拓撲關係），則兩個鄰近 匹配候選者之間僅存在一鱼 接。取決於是否允許相反交通 方向上之一路線，可能存在一 ^ 早向或雙向地圖匹配。作為 :進若任意上述路徑中不含路線之穿越形狀 Μ圖形之兩個節點。此進-步條件應避免-形 153368.doc • 12- 201231927 狀點被穿越若干次。若一先前路徑中已含路線之一形狀 點，則必須複製一部分圖形。舉例而言，此展示於圖13中 之候選者[P4，L3]處及亦候選者[P5,L2]處。在最簡單的情 況中，吾等必須僅複製單個節點。但是，若吾等之前找到 若干層級中之穿越形狀點，則可能需要複製一真實子圖 形。 在於步驟26處產生匹配圖形後，較佳地在步驟28處簡化 該圖形。若相同跡線點之兩個匹配候選者具有相同的投影 點’則兩者表示相同的匹配結果，舉例而言，此可發生在 接點處。在此情況中，吾等可藉由移除該等匹配候選者之 一者而簡化圖形。有時，一跡線幾乎平行於一線段延伸。 簡化（步驟28)亦可包含一滯後演算法之應用。若一些跡 線點之位移低於最大位移值且一些其他點之位移大於最大 位移值，則匹配狀態可能經常改變。圖14中提供此情況之 一實例，其中僅跡線點3、4、5及8、9、1〇具有低於d之一 位移值。若最大位移值設定為等於d，則僅跡線點3、4、5 及8、9、10匹配至線段N1至N8β在不使用滯後方法的情 況下，可取得由未匹配點6及7分開之匹配跡線點之兩個部 分。為了避免匹配狀態之此等頻繁變化，可使用兩個不同 的最大位移值，-較小值d及一較大值般而言，使用 該較大值D。但是在匹配圖形中之—路徑之開端及末端， 可能傾向於使用該較小值d。 如上所述，為了將滯後用於本發明之以@形為基礎之地 圖匹配，該較大距離限值D可用於建立匹配候選者。在該 153368.doc •13· 201231927 簡化步驟中，若跡線點與匹配候選者之間的距離大於該較 小距離限值d，則可移除一節點。但是’是否移除該節點 取決於該節點在圖形中之位置。此—位置取決準則可能為 節點之先前者或後繼者之數目。 如上所述，本發明提出若干種簡化一圖形之可能性。此 等簡化步驟（28)可包含但不限於實施如上所述之一種滞後 效應。另外一種可能性為使用如上所述的至下一跡線點之 航向及來自各自前一跡線點之航向。當然，可（舉例而言） 藉由建立一大角度限制（φ)及一小角度限制（ψ)而一起使用 兩種改良。取決於一節點在圖形令之位置，使用較大或較 小的角度限制。隨後，若跡線點之航向與匹配候選者處之 線之間的角度大於該角度限制，則移除一節點。 類似地’可使用其他準則，如縱向距離。取決於一節點 在圖形中之位置，可使用縱向距離之一較大或較小限制。 在此方法中，若跡線點與匹配候選者之間的縱向距離大於 所選之限制，則移除一節點。 再次返回圖9至圖13之實例，在建立路徑時，較佳地在 圖形中沒有先前者之節點處開始。此等節點可稱作起始節 點，其等在此實例中對應於跡線點P1之層級。隨後移除具 有大於臨限距離d之一距離的所有起始節點。當達到具有 小於d之一距離之一節點時，停止移除節點之程序。 對沒有任意後繼者之所有節點進行相同的操作。此等節 點可稱作末端節點，其等在此實例中可對應於跡線點以之 層級。但是，若候選者出現在圖形之第一或最後層級上 153368.doc •14- 201231927 時:則通常不使用滯後效應。在此後-情況中，不知道跡 線疋否確實終止於此。除滞後效應外，亦可使用額外準 則，如縱向距離》當‘然，針對此等額外㈣，亦可使用滞 後。 通吊不期望使用任意航向或角度準則，尤其是在航向資 料係從跡線位置（亦即’跡線點）計算的情況中。其原因在 於探測資料通常含有展現z字形_6)或環狀（圖5)等之不 ^跡線。若跡線點緊挨在一起，則航向不穩定。但是，當 提供兩品質之獨立航向資訊時，通常期望以與縱向距離相 同之方式使用一航向準則。 、下一步驟30為對匹配圖形評分。兩個準則可用於該評 )路彳二中之匹配結果之數目；及2)平均距離。主要 準則為該路徑之長度。路徑越長，匹配結果應該越好。若 兩個節點具有幾乎相同之得分，則主要使用距離準則。為 了計算平均距離，可使用各種技術，諸如：—算術平均 值均方值或任意其他的均值函數。該評分步驟在各起 始節點處開始且較佳地從頂部進行至底部。若圖形中存在 卩至底邛77支，則3平分沿循兩個分支。若已對—節點 =刀’則可比較得分並移除具有較小得分之連接。故在評 分=後’匹配圖形不含從底部向上之分支。但是從頂部至 底部，圖形仍可含有分支。在評分30後，最佳得分一 末端節點。 ' $ 了取得最佳路徑32，吾等僅須找出具有最佳得分之末 端節點。由於匹配圖形在評分3〇後不含從底部至頂部之分 153368.doc 15 201231927 支，故吾等可收集從底部向上之最佳路徑。該最佳路徑32 因此表示第-匹配結果。但是第一點或最後點可能不是最 佳的匹配結果。因此，可能希望設定路徑之特定數目的第 一及最後匹配候選者至空位狀態1徑之所有其他的匹配 候選者取得可靠狀態。 在下-步驟34中，從匹配圖形令移除路徑。亦可從圖形 中移除具有與一可靠匹配候選者相同的層級之所有節點。 在步驟34後，在步驟30處對剩餘的匹配圖形重新評分。隨 後，可在步驟32中選擇下一最佳路徑。再次對路徑節點設 定可靠及空位狀態。若空位範圍彼此重疊，則選擇局域最 佳的節點並將其設定為可靠狀態。移除具有較差得分之節 點。因此只要存在滿足一最低匹配條件之另一路徑，便可 ^复路徑選擇。自此反覆程序中選擇具有一可靠或空位狀 心之各層級處之-最可靠節點。隨後，在步驟刊中此等表 示最終匹配結果。作為此等步驟之結果，最終匹配路徑係 解析為凡叫吧山小叫叫州’叫叩…印 ^ [P7，L3]>[P8，L3]。 可使用本發明之職自料路㈣擇料巾之評分準 導出品質量測。此等可包含拓撲品質量測及/或幾何品質 量測。更具體言之，拓撲品質量測可包含一路經中之匹配 :果之數目。幾何品質量測可包含平均距離(算術平均 、均方值或任意其他均值)。與下一最佳替代匹配結果 可包括路徑令之匹配結果之數目差異（針對抬撲品 質篁測）及平均距離之差異（針對幾何品質量測）。 153368.doc 201231927 簡言之，本發明係關於一種用於將探測跡線至匹配現有 道路元素之經改良的地圖匹配方法，且其可用在一增量地 圖產生程序中，諸如（但不限於）viae N〇vae演算法。本發 明之地圖匹配方法選擇圖形之各層級中之匹配候選者之一 者作為一給定跡線點之一匹配結果或替代地將該跡線點標 §己為未匹配。該地圖匹配方法可用作為（舉例而言）於2〇〇9 年10月22日申請之PCT/EP2009/063937中所描述之方法之 一替代。地圖匹配演算法含有下列步驟：i}基於一距離準 則（亦即，若一跡線點與一道路元素之間的最短距離小於 一預定臨限）判定一探測跡線之各點可能匹配之該等道路 元素；ii)隨後將「所有可能的匹配候選者」轉化為一圖 形，其中節點描述一特定道路元素之一匹配候選者且邊指 示數位地圖中相鄰道路元素之間的一拓撲連接之程序（在 假設不在探測跡線中進行U形轉彎下使用圖形中之「複 製」節點，例如，在LI ' L2與L3之間的接點處，探測不 大可能從[P3, L1]>[P4, L3]>[P5, L1]進行）；iii)隨後（例如） 藉由使用滞後效應、縱向距離及/或其他適當的技術（值得 注意的是’無需航向及角度準則）簡化圖形；iv)隨後基於 一路徑中之匹配結果之數目（及在路徑之間一閉鎖的情況 中之平均距離）從頂部至底部對圖形評分；v)隨後識別從底 部運行至頂部之最佳路徑並將其從圖形中移除；及vi)按需 要重複評分及所選的最佳路徑直至獲得一匹配結果。 在數位地圖領域中存在本發明之許多應用。此等應用可 包含（舉例而言）探測資料之分析、增量網路產生程序（例 153368.doc •17· 201231927 如，Viae Novae項目）、自探測資料之特徵提取（例如，速 度量變曲線、交通密度、道路分類等）、資料品質檢查（例 如’一道路網路之完整性及正確性、道路屬性之真實性檢 查等）、及網路合併工具等。 在本發明及隨附申請專利範圍之範_内，不同實例實施 例之元件及/或特徵可彼此組合及/或彼此替代。 此外，可以一裝置、方法、系統、電腦程式及電腦程式 產品之形式體現上述及其他實例特徵之任一者。舉例而 吕，可以一系統或器件（包含但不限於用於執行圖式中所 不之方法之結構的任意者）之形式體現上文提及之方法的 任意者。 雖然以上#細描述中所述之實施例引用Gps，但是應注 意導航裝置可利用任意種類之位置感測技術作為Gps之一 替代（或實際上之補充）。舉例而言，導航裳置可使用基於 其他全球導航衛星之諸如歐洲伽利略（Eur〇pean GaHle〇)系 統。同樣地，纟不限於基於衛星，而可易於使用基於地面 之信標或使器件判定其之地理定位的任意其他種類之系統 起作用。 本發明之以上描述本質上係例示性而非限制性。所揭示 實施例之變化及修改可為熟悉此項技術者所瞭解且落於本 發明之範疇内。因在匕，授予本發明之保護範疇僅由下列申 請專利辄圍界定。 【圖式fal單說明】 圖1係現實中之一例示性道路之一鳥瞰圖； 153368.doc -18- 201231927 圖2係來自對應於現實中 段； 之—道路之一數位地圖的一線 圖3展示重疊在圖1之道路上方之圖2的線段； 圖4係展示具有一點狀雲形式之不良品質之—例示性探 測跡線之如圖3中的—視圖； 圖5係展示具有—環狀形式之*良品質之—例示性探測 跡線之如圖3中的一視圖； 圖6係展示具有一2字形形式之不良品質之一例示性探測 跡線之如圖3中的一視圖； 圖7係展示具有—路面外跡線形式之不良品質之一例示 性探測跡線之如圖3中的一視圖； 圖8係根據本發明之經改良之地圖匹配方法的一流程 13 · 圚， 圖9係一道路系統連同來自一數位地圖之代表性線段之 一例示性繪示，其中一路線（L1)與一鄰近路線（L4)拓撲不 相連； 圖10係來自圖9之數位網路之四條線段連同跡線點？1至 P8所界定之一例示性單一跡線線之一視圖； 圖11係展示各跡線節點P1至P8之所有匹配候選者之如圖 1 〇中之一視圖； 圖12係描繪用於判定匹配候選者之一映射策略之一表 格； 圖13係根據本發明之一有向圖形，藉此根據預定準則將 各匹配候選者置於圖形中之一層級處且在鄰近層級中之候 153368.doc -19· 201231927 選者之間建立路徑；及 圖14係用於使用滯後效應將跡線點匹配至一線段之一例 示性方法。 【主要元件符號說明】 20 數位地圖 22 探測資料源 36 匹配結果 38 接點 D 距離 d 距離 L1 至 L4 線段 N1 至 N8 線段 P1 至 P10 跡線點 153368.doc -20-Novae algorithms sometimes fail to always convey the correct results. Therefore, a novel and improved map matching algorithm is needed. This improved algorithm must demonstrate a matching quality that works well in incomplete road networks (eg, missing roads, missing topologies, or errors), and is resistant to poor quality GPS traces. The line is very stable (ie, poor detection data), such as traces with point clouds (Figure 4), loops with loops (Figure 5), traces with zigzags (circle 6) and pavement External traces (Fig. 7), such as improved maps, often occur in the collection of probe data. In addition, the algorithm must operate in conjunction with both one-way and two-way networks. SUMMARY OF THE INVENTION The present invention provides a method for matching a trace derived from a probe data to one or more of a line map of a trace derived from a survey data, such as a digital vector circle. . Preferably, the digital map is a network of 153368.doc 201231927, wherein the line segments represent at least a portion of a road, a path, or the like. The method comprises the steps of: providing a digital vector _ configured to store a plurality of line segments spatially associated within a coordinate system, the line segments being topologically connected or disconnected from each other; providing at least one detection trace, Having a plurality of spatially associated trace points positioned within the coordinate system of the digital vector map; and determining a match candidate for each line segment within a particular distance (d) of each trace point. The method is characterized by generating a pattern (26) of the matching candidates as one of the nodes; and selecting at least one path of the matching candidates of the pattern as the matching result (36). The new map matching method is very useful for incremental map generation programs and several other applications due to its inherently high matching quality. The present invention produces a match pattern for one of the matching candidates, which in turn allows the use of a graphical theory approach to find the best path through the graph. The nodes of the graph are made up of several matching candidates. The matching pattern includes one of the individual trace points. If a trace point has no matching candidates, the consistent level of the graph does not contain nodes. One of the core concepts of the present invention uses a matching pattern to represent all matching candidates. Simplify and rate the graph. These optimal paths convey the matching results. The concepts of the present invention can be used in conjunction with the longitudinal distance as a matching criterion. [Effects] These and other features and advantages of the present invention will become more readily apparent from the <RTIgt; Reference is made to the drawings in which like reference numerals refer to the same 153368.doc 201231927 or the corresponding components, and an overview of the map matching procedure in accordance with the present invention is shown in FIG. A digital map 20 provides information on a road network. The digital vector map 20 is configured to store a plurality of line segments spatially associated within a standard system. The line segments in the digital map 20 are respectively topologically connected to each other or the topologies are not connected. Each line segment contains at least one side. When a line segment has a plurality of edges, the shape points interconnect the edges. Each line segment terminates at a point of contact. A probe data source 22 provides at least one but more generally a large number of probe traces. Each trace contains a sequence of one of the spatial points of the spatial distribution within the coordinate system of the digital vector map. - This type of detection trace typically represents the movement of a small car, bicycle, pedestrian, etc. Map matching is the process of determining which line segment of a digital map a traveler is traveling on when creating a trace. This is accomplished by considering points in a probe trace to identify matching candidates and then selecting the best possible option from among the candidates. In step 24, the points in the probe trace are considered one by one to find one or more expected line segments in close proximity to the digital vector map. Thus, a match candidate is a possible match result for a given trace point to one of a particular line segment present in the digital map. A trace point can have several matching candidates. One method of determining one of the set of trace points to match the set of candidates is to select all of the lines of the digit vector map within a predefined maximum distance d. The projection of the trace point to the line is calculated for each of the selected segments. All such projections; ί's collection is a matching candidate for this trace point. The method of the present invention is characterized by the step of generating a pattern of one of the matching candidates, which is represented by function block 26 of Figure 8. To facilitate the understanding of the subject 153368.doc 201231927, the graph can be viewed as having a plurality of sequential levels that correspond in number to the number of points in the probe trace. Thus, if a given probe trace has ten points, the graph will have ten levels. The levels of the graph correspond in sequence to the other levels in the S-shape, just as the trace points correspond to other trace points in sequence. Thus, adjacent trace points correspond to adjacent levels in the graph. Another way to understand the relationship between the graphics level and the trace points is to assign the first trace point to the first level; assign the second trace point to the second level...; assign the last trace point To the last level. Each matching candidate is placed in the graph at a level corresponding to its trace point. For example, if there are three matching candidates for the first trace point (meaning that the first trace point of the Hai is within the predefined maximum distance d from three different line segments in the digital map), then Place all three matching candidates at one level in the graph. A matching result is identified from one of the plurality of levels or between the plurality of matching candidates: the matching result of each of the trace points is a point at which the maximum likelihood corresponds to the digit vector map of the trace point. Each trace point has at most one matching result. If the trace point has a certain probability outside of the digit vector map, it can be concluded that the trace point does not have a -match result. A point in the digital vector map can be described by the line to which the point belongs and the number of chains along the line (also known as the measured value). Alternatively, the line and one of the line shape points describe the point in a digital vector map. Each matching result is treated as one of the nodes in the graph. One or more paths extend through the nodes', i.e., through the various levels of _ matching candidates. The paths are composed of a sequence of nodes, each of which is connected by a 153368.doc 201231927 edge. Each path connects the matching candidates in each level to one of the next adjacent levels in the direction, provided that the respective two matching candidates are topologically connected to each other. In another way, each path is not intended to connect segments that are not connected to each other (i.e., segments representing real roads that cannot traverse each other). In this way, two nodes are connected if the following conditions are met: υ the matching candidate belongs to the adjacent level; 2) there is a (local) route between the two projection points; and 3) the above path does not include the The route crosses the shape point. For the sake of brevity, it should be understood that the path is a special part of the graph with a linear sequence of graphics. When a graphic is produced, the graphic will contain one or several paths. In general, a graphic contains two different types of objects: nodes and edges. - Edges are (sorted) a pair of nodes. The edge represents the connection between the nodes. The node of the circle is a matching candidate. If there is a -topological connection between the two matching candidates, the two nodes are connected by a directed edge. This only considers local topology connections. For example, topology connections via a wraparound = should not be considered. - One-way map matching should also consider the direction of topological connections. On the one hand, a two-way map matching procedure should not consider the sense of direction. In order to find out (4) the (4) of the (4) local extension of each matching candidate representation of the line segment), the distance-restricted-selection path algorithm ^ (for example by a maximum search by any appropriate technique (for example, including steps) One of the 28 ping-pong methods) simplifies the graph that initially contains all possible paths. The other Γ = the distance between the candidate and the corresponding trace point - the longitudinal distance is greater than the maximum longitudinal distance of the pre-depreciation. All nodes are removed. After simplification, I53368.doc 201231927 can score 30 for various paths so that it can be selected from among them - the best teacher 32 is better based on the number of connected nodes in the longest path to which the node belongs ^ and / Or average distance. Other criteria can be used. During or after the scoring step %, if the score is lower than others, the edges and/or nodes can be removed. This will further simplify the graph based on the result of the split and relatively simple from The best path is selected in the graph (step 32). In most cases, the selected path already contains one of the levels, ie the point 'which means that the trace points have been reliably found — the matching result ^. In the case of a number of cases, the path may not contain nodes of some level. This may be the case when there is a missing topology connection between the two lines of the digital vector map. In this case (where a path is found without some hierarchical nodes) Preferably, some or all of the nodes having the same level as the nodes of the selected path are removed, which corresponds to step 34 in the flow chart of Figure 8. The removal step 34 will further simplify the graphics. “After, re-score the graph by 3〇 and select the next best path (32). Repeat this removal 34, re-score 3〇 and select the 3 2 sequence until you have found a reliable match for all trace points, or until The graph is empty. Usually only very few repetitions are required. An illustrative example is provided by Figures 9 to 13. Here, in a digital network, the actual road is assumed to have originated from a digital map. The line segments L1 to L4 are indicated. The line segments u, and are connected to each other in a topological manner at a common intersection marked by the contacts 38. On the other hand, the line segment L4 is not connected to the u, L2 and L3 topologies. Corresponding to L4 The road in reality is on the road The L1 part traverses below and does not allow traffic to interact directly with any of ., or .. Figure 1 〇 shows these same line segments as well as "together with one of the detection traces indicated by point 153368.doc 201231927 P8 The detection trace originates from the data source 22 » The matching candidate selection step 24 is performed using a suitable algorithm. For example, the algorithm can be based on each respective trace point or node (P1, P2, P3, .. .P8) The distance from a number of network elements (LI, L2, L3, L4) to a predefined threshold d. One of the trace points p matches the candidate for the point closest to the network of P Element: A network element can be an orthogonal projection of p to a line segment, a joint (eg, 3 8), or a shape point. The figure shows all the matching candidates to the traces (p丨 to p8) of the road network formed by the four line segments L1 to L4. The graphics generation step 26 is primarily one of the algorithms for generating a matching pattern for one of the matching candidates and is depicted in tabular form in FIG. This method allows the use of graphical theory to find possible paths through the graph and then select the best from them. The nodes of the graph are a number of matching candidates. The matching pattern includes a separate level for each of the trace points (ρι, p2, p3, p8). If the - trace point has no matching candidates, the consistent level of the graph does not contain nodes. FIG. 13 shows a matching pattern corresponding to one of the circle u and the example in FIG. Therefore, the matching graphic display & ^ exhibition is not a directed graph, in which only one side is connected to the node of the adjacent level, and the former county 嘿 ^ L, > meets the above conditions. That is, if there is a route between the two projection points (i.e., a continuous topological relationship), there is only one fish connection between the two adjacent matching candidates. Depending on whether one of the directions in the opposite traffic direction is allowed, there may be an early or two-way map match. As: If any of the above paths do not contain the route through the shape of the two nodes of the graph. This step-by-step condition should be avoided - shape 153368.doc • 12- 201231927 The point is crossed several times. If a path in one of the previous paths already contains a shape point, you must copy a portion of the graphic. For example, this is shown at the candidate [P4, L3] in Figure 13 and also at the candidate [P5, L2]. In the simplest case, we must only copy a single node. However, if we have previously found crossing shape points in several levels, we may need to duplicate a real sub-pattern. After the matching pattern is generated at step 26, the pattern is preferably simplified at step 28. If two matching candidates of the same trace point have the same projection point' then both represent the same match result, which may occur, for example, at the joint. In this case, we can simplify the graphics by removing one of the matching candidates. Sometimes a trace extends almost parallel to a line segment. Simplification (step 28) may also include the application of a hysteresis algorithm. If the displacement of some of the trace points is lower than the maximum displacement value and the displacement of some other points is greater than the maximum displacement value, the matching state may change frequently. An example of this is provided in Figure 14, where only trace points 3, 4, 5 and 8, 9, 1 have a displacement value lower than d. If the maximum displacement value is set equal to d, only the trace points 3, 4, 5 and 8, 9, 10 are matched to the line segments N1 to N8β, and the unmatched points 6 and 7 can be obtained without using the hysteresis method. Match the two parts of the trace point. To avoid such frequent changes in the matching state, two different maximum displacement values can be used, the smaller value d and a larger value, as the larger value D is used. However, in the matching graph, the beginning and end of the path may tend to use the smaller value d. As described above, in order to use the hysteresis for the @shape-based map matching of the present invention, the larger distance limit D can be used to establish a matching candidate. In the simplification step of 153368.doc •13· 201231927, if the distance between the trace point and the matching candidate is greater than the smaller distance limit d, one node can be removed. But 'whether or not to remove the node depends on where the node is in the drawing. This—the location-resolved criterion may be the number of previous or successors of the node. As described above, the present invention proposes several possibilities for simplifying a graphic. These simplified steps (28) may include, but are not limited to, implementing a hysteresis effect as described above. Another possibility is to use the heading to the next trace point as described above and the heading from the respective previous trace point. Of course, two improvements can be used together (for example) by establishing a large angle limit (φ) and a small angle limit (ψ). Depending on where a node is at the graphical order, use a larger or smaller angle limit. Then, if the angle between the heading of the trace point and the line at the matching candidate is greater than the angle limit, then a node is removed. Other criteria, such as longitudinal distance, can be used similarly. Depending on where a node is in the drawing, one of the vertical distances can be used with a larger or smaller limit. In this method, a node is removed if the longitudinal distance between the trace point and the matching candidate is greater than the selected limit. Returning again to the examples of Figures 9 through 13, when the path is established, it preferably begins at the node in the graph where there is no previous one. Such nodes may be referred to as starting nodes, which in this example correspond to the level of trace point P1. All starting nodes having a distance greater than one of the threshold distances d are then removed. When a node having a distance less than one of d is reached, the process of removing the node is stopped. Do the same for all nodes that do not have any successors. Such nodes may be referred to as end nodes, which in this example may correspond to a hierarchy of trace points. However, if the candidate appears on the first or last level of the graph 153368.doc •14- 201231927: the hysteresis effect is usually not used. In the latter case, it is not known whether the trace does end up here. In addition to the hysteresis effect, additional criteria can be used, such as longitudinal distance, when ‘those, for these extra (four), lag can also be used. It is not desirable to use any heading or angle criteria, especially if the heading data is calculated from the trace position (i.e., the 'trace point'). The reason for this is that the probe data usually contains traces that exhibit a zigzag_6) or a ring (Fig. 5). If the trace points are close together, the heading is unstable. However, when providing two-quality independent heading information, it is often desirable to use a heading criterion in the same manner as the longitudinal distance. The next step 30 is to score the matching graphic. Two criteria can be used for the number of matching results in the review; and 2) the average distance. The main criterion is the length of the path. The longer the path, the better the match should be. If two nodes have almost the same score, the distance criterion is mainly used. To calculate the average distance, various techniques can be used, such as: - arithmetic mean mean value or any other mean function. The scoring step begins at each of the starting nodes and preferably proceeds from the top to the bottom. If there are 77 邛 to bottom 图形 in the graph, then 3 bisectors follow the two branches. If already - node = knife' then compare the scores and remove the connections with smaller scores. Therefore, in the score = after 'matching graphics, there is no branch from the bottom up. But from top to bottom, the graph can still contain branches. After scoring 30, the best score is the end node. For the best path 32, we only need to find the end node with the best score. Since the matching graphic does not contain the score from the bottom to the top after rating 3〇, we can collect the best path from the bottom up. This optimal path 32 thus represents the first match result. But the first or last point may not be the best match. Therefore, it may be desirable to set a certain number of first and last matching candidates of the path to all other matching candidates of the vacancy state 1 path to obtain a reliable state. In the next step 34, the path is removed from the matching graphics order. All nodes having the same level as a reliable match candidate can also be removed from the graph. After step 34, the remaining matching graphics are re-rated at step 30. The next best path can then be selected in step 32. Set the reliable and vacant state of the path node again. If the gap ranges overlap each other, select the best local node and set it to a reliable state. Remove nodes with poor scores. Therefore, as long as there is another path that satisfies a minimum matching condition, the path selection can be repeated. From this iterative procedure, the most reliable node at each level with a reliable or vacant heart shape is selected. This is then followed by a final match in the step report. As a result of these steps, the final matching path is parsed as the name of the bar called the state called '叩 叩...印 ^ [P7, L3] > [P8, L3]. The quality of the product can be derived by using the rating of the material of the invention (4). These may include top quality measurements and/or geometric quality measurements. More specifically, the top quality test can include a match in the way: the number of fruits. Geometric quality measurements can include average distances (arithmetic mean, mean square, or any other mean). The result of matching with the next best alternative may include the difference in the number of matching results of the path order (for the quality of the lift) and the difference in the average distance (for the geometric quality measurement). 153368.doc 201231927 Briefly, the present invention relates to an improved map matching method for detecting traces to match existing road elements, and which can be used in an incremental map generation program such as (but not limited to) Viae N〇vae algorithm. The map matching method of the present invention selects one of the matching candidates in each level of the graph as a match result for one of the given trace points or alternatively the trace point is already unmatched. The map matching method can be used as an alternative to the method described in, for example, PCT/EP2009/063937, filed on Oct. 22, 2009. The map matching algorithm has the following steps: i} determining that the points of a probe trace may match based on a distance criterion (ie, if the shortest distance between a trace point and a road element is less than a predetermined threshold) And other road elements; ii) then convert "all possible matching candidates" into a graph in which the node describes one of the specific road elements matching the candidate and the side indicates a topological connection between adjacent road elements in the digital map The program (using the "copy" node in the graph under the assumption that the U-turn is not in the probe trace, for example, at the junction between LI ' L2 and L3, the probe is unlikely to be from [P3, L1]> [P4, L3] > [P5, L1]); iii) simplification, for example, by using hysteresis effects, longitudinal distances, and/or other appropriate techniques (notable for 'no heading and angle criteria') Graphics; iv) then scoring the graph from top to bottom based on the number of matching results in a path (and the average distance in the case of a lock between paths); v) then identifying the best path from bottom to top And remove it from the graph; and vi) repeat the scoring and the best path selected as needed until a match is obtained. There are many applications of the invention in the field of digital mapping. Such applications may include, for example, analysis of probing data, incremental network generation procedures (eg, 153368.doc • 17·201231927, for example, Viae Novae project), feature extraction of self-detecting data (eg, velocity-volume curves, Traffic density, road classification, etc., data quality inspection (such as 'the integrity and correctness of a road network, the authenticity of road attributes, etc.), and network merger tools. The elements and/or features of the different example embodiments may be combined with each other and/or substituted for each other within the scope of the invention and the scope of the appended claims. In addition, any of the above and other example features may be embodied in the form of a device, method, system, computer program, or computer program product. For example, any of the methods mentioned above may be embodied in the form of a system or device, including but not limited to any of the structures for performing the methods not illustrated in the drawings. Although the embodiment described in the above detailed description refers to Gps, it should be noted that the navigation device can be replaced (or actually supplemented) with one of the types of Gps using any kind of position sensing technology. For example, navigational skirts can be used based on other global navigation satellites such as the European Galileo (Eur〇pean GaHle〇) system. Similarly, it is not limited to satellite-based, but can be easily implemented using ground-based beacons or any other kind of system that causes the device to determine its geographic location. The above description of the invention is intended to be illustrative rather than limiting. Variations and modifications of the disclosed embodiments are known to those skilled in the art and are within the scope of the invention. As a result, the scope of protection granted to the present invention is defined only by the following patent applications. [Figure fal single description] Figure 1 is an aerial view of one of the exemplary roads in reality; 153368.doc -18- 201231927 Figure 2 is from the map corresponding to the real middle section; Figure 2 is a line segment of Figure 2 superimposed on the road of Figure 1; Figure 4 is a view showing the poor quality of a point cloud form - an exemplary probe trace as shown in Figure 3; Figure 5 is a view showing a ring-shaped form Figure 7 is a view of an exemplary probe trace having an undesired quality in a zigzag form as shown in Figure 3; Figure 7 is a view of FIG. An illustration of an exemplary detection trace having one of the poor quality qualities in the form of a road surface trace as shown in FIG. 3; FIG. 8 is a flow 13 of an improved map matching method in accordance with the present invention. A road system is exemplarily illustrated along with one of representative line segments from a digital map, wherein one route (L1) is not connected to a neighboring route (L4) topology; FIG. 10 is a four line segment from the digital network of FIG. Together with the trace points? A view of one of the exemplary single traces defined by 1 to P8; FIG. 11 is a view showing one of the matching candidates of each of the trace nodes P1 to P8 as shown in FIG. 1; FIG. Matching one of the mapping strategies of one of the candidates; Figure 13 is a directed graph in accordance with the present invention whereby each matching candidate is placed at one level in the graph and in the adjacent hierarchy 153368 according to predetermined criteria. Doc -19· 201231927 Establishing a path between selectors; and Figure 14 is an illustrative method for matching trace points to a line segment using hysteresis. [Main component symbol description] 20 Digital map 22 Probing data source 36 Matching result 38 Contact D Distance d Distance L1 to L4 Segment N1 to N8 Segment P1 to P10 Trace point 153368.doc -20-