201242368 六、發明說明： 【發明所屬之技術領域】 應 統 本發明係關於-種智慧娜像監控錢物件行跡追㈣統特別是指 用於具有龐大數雜及鱗_舰之轉難⑽像監控系s 【先前技術】 影像辨識技術在影紐齡統巾已被廣泛軸，目衫數朗是對單 一攝影機影像進行影像觸(諸如車牌辨識、車流細、人臉偵 識、位耗測、人員追縱、方向性移動_、電子圍釀 '物件遺留、攝影 機遮蓋、人臉侧等），待辨識完錢，將影料徵(諸轉牌號碼、車輛 顏色、車輛_、車輛廠牌贼、大/小/機車數量、物體妓、物體大小、 人員資訊、人員合法性、方等）資訊作儲存及顯示，喊有對不同攝影 機之間的鱗顺結果作更姐地_，無驗fA#的運算資源。 影像辨識的行跡追蹤技術大多_在單隻攝影機所拍下的影像，將辨 識出的物件行跡以—連串相對於影像平面的座標表示，進而建構出在單一 畫面中的物件移動軌跡，其細範圍有限，在中華民國專利證號㈣142 揭露-種賴職取像與追财法，健合雙攝频作取像與目標影像追 縱，其中雖然、是利用辨識雙攝影機的影像來進行物件的追縱，但二攝影 機的攝影場景是相關，也就是雖朗了雙攝影機但是攝影的地點場景2 相同的，ϋ此可追輸件畴跡範難舰關_場景中。 而其餘習知跨越攝影機的物件行跡追蹤技術又常有不同攝影機之間必 201242368 須有部份場景重4的_ ’才麵法進行物件行跡追蹤。其應雜圍也必 須偏限於-固定區域’且攝韻架設的密缝必—般情況更高，也就 是相同範圍必須要練比-織況更多的攝韻才麵法完成跨攝影機的 追蹤’這樣的行蹤追蹤系統勢必增加建構的成本。 是以’擴增影像辨識結果的應用細以及解決習用行蹤追縱技術之問 題及缺失，即為本發明所亟欲研究改善之方向所在者。 【發明内容】 本發明之目的即在於提供一種智慧型影像監控系統物件行跡追縱系 統’係利用跨越不同攝影機的影像辨識結杲，為其建立時間與空間的關係， 進而建構出辨識目標的行蹤資訊，並將行蹤資訊標示於電子地圖當中及 依序播放轉將行跡資訊舰覺化現，並可㈣摘視覺化的影像轉 錄為單彡觀伽放纽㈣省重複行_選的日销，故發明人鑑於 上述習用方式所綱缺點，㈣加以改並經苦爾旨潛心研 究後’終於成功研發完成本智慧鄕健㈣補件行料縱系統。 本發明相較於習知影像辨識技術在影像監㈣統中的翻，更進一步 利用跨越不同攝影機的影像辨識結果，為其建立時間與空間的關係，進而 建構出辨識目標的行师訊；w㈣跨越細_件行跡追縱技 本發_额職彳谓追_枝’可林轉域賴钱密度的 狀況下’利驗件域在各卿__料行猶翻鱗，可追縱跨 峨影機的大細區_物件的行縱。 斤述本發明之目的即在於提供—種智慧型影像監控系統物件行 201242368 跡追蹤祕解決祕問題及缺失’本發鴨彻跨越不同攝影機的影像 辨識.”。果A其建立時間與空間_係’進而建構出辨識目標的行縱資訊， 並將行蹤貝从不於電子地g當巾，及依序播放影像將行跡資訊作視覺化 的呈現，並可將前述·覺化的影像轉錄鱗1職案並喊重現以節 省重複行跡篩選的時間。 達成述發月目的之秦慧型影像監控系統物件行跡追縱系統包含：影 像監控子系統、影像辨識子系統、物件行跡查詢子系統以及物件行跡視覺 化子系統等四解线。其巾像缝衫統可特各賴韻的數位 壓縮影像、提供錄影以及調閱錄影檔等功能；該影像辨識子系統可提供針 不同[縮格式的數錄細識、儲存及查詢影像辨識結果等功能；該物 件行跡查解輪㈣細樣、絲' f椒綱賴地點相 關身訊、_«觸件行_訊魏；該物件行贼覺化子錢可將物 件仃跡路輸W_t巾，帅瓣_纖働識目標的 各路攝影機的錄影影像、並可提供一改良影像錄影檔案格式，將前述的各 =攝影機的錄影影像以及行跡資訊轉錄為單-行跡影像㈣、以及播放前 述之仃跡f烟t錢在奸_巾重聽件行跡路線。 本發明所触之智_職監㈣錄件行跡追 技術相互峨_繪咖： 、其糊 =^術絲輸⑽顺跡哪咖是對單一 ^ 行鄕_，細觀成後，郷雜«訊作儲存 異浪費大=1:_機0之間的影像辨識結果作更有效地利^ ，源，本發明更進一步利用跨越不同攝影機的影像辨 201242368 識結果，為其建立時間與空間的關係，進而建構出辨識目標的行縱資訊。 2.影像辨識的行跡追蹤技術大多顧在單隻攝影機所拍下的影像，將辨識 出的物件行跡以-連串相對於影像平面的座標表示，進而建構出在單一 畫^的物件移動軌跡，其制範财限;在中華民國專利證號咖… 揭露-種韻賴取賴追蹤綠，雜合魏職翁像與目標影像 、’其令雖然疋利用辨識雙攝影機的影像來進行物件的追縱，但其兩 攝韻的攝影場景是相同的，也就是雖然用了雙攝影機但是攝細地點 %景疋相同的’因此可追縱物件的行跡範圍僅侷限於同—場景中；而其 餘習知跨越攝影機的物件行跡追縱技術又常有不同攝影機之間必須有 部份場景重叠的限制，才有辦法進行物件行跡追縱。其應用範圍也必須 侷限於-固定區域，且攝影機架設的密集度必須比一般情況更高，也就 疋相同範圍必須要架設比-般情況更多的攝影機才有辦法完成跨攝影 機的追蹤’祕的練追猶統勢必增加建構的成本；而本發明的跨攝 影機仃跡追縱的方式，可在不需增加攝影機架設密度的狀況下，利用物 見在各攝景彡機的地點作為行跡建構的基準，可追縱跨越攝影機的大 範圍區域裡物件的行蹤。 3·本發縣合目輕像監㈣統市場展膽，翁慧郷像辨識像功 月b納入影像監控系統中，除辨識結果可供查詢之外，更進一步擴大的影 像辨識結果的綱制，針對大量鱗觸結果進行分析，從中建構出 物件的行蹤，並利用分析出的物件行蹤資訊將其視覺化，將行跡路線在 也圖服務或是在自訂地圖中標示出來以及依照物件出現的位置，將相 對應的攝一影像#段依序播放，方便追縱物件；本發明亦可進一步將 6 201242368 前述的影片片段及物件行蹤資訊，轉錄成單一樓案，節省再一次進行物 件行蹤資訊的查詢時間；智慧型影像監控系統物件行跡追縱系統可應用 於智慧型影像監控系統之跨不同攝影機(地點）的物件追縱，可協助追縱 犯案車輛、絲車輛、失縱人口、嫌疑犯、者、可疑物件或失竊物 件的行蹤’有助於增進未來數年將影像監控系統推廣。 上列詳細說明乃針對本發明之—可行實施例進行具體說明，惟該實施 例並非用以關本伽之專利卿，凡未雌本發明技藝精神所為之等效 實施或變更，均應包含於本案之專利範圍中。 知上所述’本案不僅於技術思想上確屬創新，並具備習用之傳統方〉 所不及之上述多項功效，已充分符合新酿及進步性之法紐明專利要 件，差依法提Μ請，騎貴局核准本件發明專利巾請案，以勵發明 至感德便。 【實施方式】 請參關—所示，係本發明智慧型影像監控系統物件行跡追縱系統的 系統架構圖。 請參閱圖二為本發明智慧型影像監控系統物件行跡追料、統之系統模 組圖’其中該影像監控子系統10包含前端取像模组u '影像收集模組… 影像錄影齡13 «像飾14 ;郷像賴餘㈣_像辨識模 組2卜影像賴縮難22及觸結料料庫23 ;該物件行㈣解系統 30包含嶋輯副触31、攝職_⑽、行跡筛選模组 33及物件彳了34 ;雜件彳侧找好纽㈣含細影像轉錄 201242368 模組4卜行跡影像播案回放模組42及物件行跡影像檔案資料庫43。 本發明中的物件行跡追蹤的功能主要由分析具有影像辨識功能的影像 監控系統所收⑽資料來完成物件行跡的建構，分析的結果將存入資料庫 中以便查詢，再利用分析的結果來進行物件行跡查詢以及物件行跡視覺化 的功能。本發明中影像監控系統所收集的資料包含了攝影機地點相關資 訊、鄰近攝影機資訊'攝影機影像及影像辨識結果。 本發明t首先所要收集的資料為攝影機地點訊以及鄰近攝影機 的資訊。攝影點資賴收集方式可以區分為兩種：第—種方式為由影 像收集歡丨2收集具備TcP/IP傳輸捕的前端設備(如㈣嶋)所回傳 的地點資料.帛—種方以本發賴供攝影機聽資纟鍵構模組Μ讓使用 者手動輸入攝影機地點的資料。 當採用第-種方式時，本發明在具備TGp/Ip傳輸功㈣前端設備安装 GPS疋位備’當影像收集模組12在第一次連上前端設備時，前端設備 會透過TCP/IP的方式將從GPS粒設備所雜到的經緯度倾傳送給影像 收集模組12;影像收雜組12會將收集到的攝影機地點資料存人攝影機地 點資料庫32中。此方式適用於攝影機佈建於廣大區域的路口影像監視系 統，GPS的經緯度資料可以配合現有的電子地圖服務來顯示的位置，故本發 月具備有在現行電子地圖服務上顯示物件行蹤的功能。 第-種方式主要使用在攝影機架設的位置不適用於使用Gps定位設備 時如在攝影機架設在大型賣場或是大樓中，在這種狀況下不同樓層的攝 先機得到的經緯度資料有可能重覆，因此不適合使用Gps經緯度的資料來 辨別攝域的地點；本㈣提供-攝纖地點資織構_31，讓使用者 201242368 -可使用-方便賊用者介面，自行建構攝雜地_請，並將使用者提 供的攝影機地點的資料存入攝影機地點資料庫32中，這裡的攝影機地點可 以使用和電子地_對應的位置雜何枝標示攝賴地關表示方式來 紀錄’故本發明也具備有在使用者自行提供的電子地圖上顯示物件行縱的 功能；在此方式中的電子地圖在資料庫中也存有—索引可供查詢。 本發明㈣近攝織資訊財式根職職魅的衫方式也分 為兩種··第-種方式為攝影機地點資料使用Gps經緯度資料表示時，利用 計算的方式來計算攝影機地點是利目鄰，並依照計算的結果建立每一攝影 機的鄰近攝影機的資訊，並存放於攝影機地點龍庫32中；第二種方式為 攝影機地_4_者祕的物__频置時，攝賴地點資 訊建構模組31提供-方便的使用者介面來提供使用者自行輸入每一攝影機 的鄰近攝影_資料，並存人攝韻地點㈣庫32中。 影像收集模組12從前端取像模組u中的影像收集設備收集各種壓縮 格式的攝域數位影像’影像錄職組13將收細雜雜，並將攝影機 :及時_軸人賴_14巾作為查娜咖。前端影像收 集。又備使用現㈣各種影級缝術(_壓歸像齡卡、硬體壓縮影像 娜卡、前端攝影機壓縮、IPCamera、Vide〇s⑽r、賺、臓)來收集 經侧·爾:雜，其働料·PEG、麵侧销、 嶋嶋__細3，麵影模組13 4庫ΓΓ成—錄’並舰影像轉的時_及攝職資訊在影像 貝枓庫14中建立索弓|，以利影像調閱。 本發月針對$慧型影像監控系統各路攝影機的錄影影像資料進行影像 201242368 辨識，由於經由不同取像設備所收集的攝影機影像會有不同的壓縮格式（如 麗、MPEG4、η· 263、η. 264等），所以透過影像錄影模組13所取得的錄影 機影像資料，必須再經由影像解壓縮模組22根據不同的壓祕式進行解壓 縮，來得到可錢行雜_的雜·f料，影像職· 21即可針對 原始影像㈣進行各郷像織，並顧__件職(諸如車牌號碼、 車輛顏色、車輛麵、車輛廠牌型號、大/小/機車數量、物體位置、物體 大小、人員資訊、人員合法性、方向性等），連同時間相關資訊(辨識物件 出現的時間戮記)和空間資訊(前端攝影機的地點資訊)存入辨識結果資料 庫23中。 /行跡查詢子系統3〇根據物件行跡料已_完成，蚊妓進行物件 行跡的建構動作：在伽者欲查_物件行跡林在於物件行跡資料庫％ 中之情況下’行跡筛選模組33會依照辨識結果及攝影機地點來建構物件行 跡，並將物件行跡資料交由物件行跡視覺化子系統4()顯示及播放影像最 後並將影像資料以及物件行跡資料交由行跡影像轉錄模組41轉錄成—特殊 影像錄影齡格式崎_，行跡影像轉賴組41會在轉錄_時在物件 行跡麵咖庫43建均丨，以供使_詢已經建構完成的物件行 跡；在使用者欲查詢的物件行跡已存在之情況下，物件行跡視覺化子系統 40即直接賴觸簡案断物件行跡的視覺化動作。 行跡篩選模組33針對辨識結果資料庫23進行查詢，本發明利用物件 出現在各卿_地點作為物件行跡建構的鮮，由於資料巾包含物件出 現的時間戰記及前端攝影機地點資訊，依此即可將辨識目標的行跡建構出 來，行跡資訊即為—組依照時間排序的倾列〔物件特徵，物件所在攝影 201242368 機地點，物件出現時間戳記〕，行跡筛選模組33會將所得的行跡資訊儲存 於物件仃跡資料庫34中。查詢的方式如下，行跡騎模⑽挑選出鄰近 時間中出現具有相似物件特徵的資料列並依時間排序；鑑於查詢的結果 有可能因為物件描述相近，造成查詢結果包含錯誤的資料列，此時行跡筛 賴組33會-併查詢攝影機地點資料庫犯中的鄰近攝影機的資訊，來確 認行跡資料列是否呈現地點相__，藉由確認此關係來 過遽掉不合理的龍列，來增加行跡資料的正轉。 物件仃跡視覺化子系統4〇在行跡:纽建構絲之後，即舰使用者所 想要查詢的物件行跡做視覺化_作，將物件行跡資訊中的攝影機地點依 照順序在電子地圖上標示出來，並在各個物件出現線段，以及同時調閱物 件行跡中各路攝影機的錄影影像片段並錢時間順序播放出來。標示物件 灯跡的方式錢攝影機地點㈣的絲方式分為兩種：第—種方式為將物 件行跡標示於電子地圖服務中，攝影機地點資料使用Gps經緯度資料表示 時，即可標示在财的電子地瞧務巾（如G_e 第二種方式為將 物件行跡標示於自行建構地时，攝影機地輯料若是使用者自行建構的 資料’則可標示於使用者自行建構的電子地圖中（如大樓平面圖）。 行跡影像觸额41會將觸各職賴的職龍以及行跡資 料’進行轉錄成特殊影像錄影格式的單—錄，節省重複篩選物件行跡 的時間μ了跡f彡賴案回放触42可將祕祕箱雜祕式錄影棺案 中的影像回放’以及將物件行着訊由觀中取出，將物件行跡重現於電 子地圖服務或使用者自行建構的電子地圖中。 有鐘於傳_紹觀無法紀跡邮彡顏齡柄則，無法符合 201242368 本發明所需要的_舰’目此本發明改良錄織案格式以符合本發明中 的功能需求。改隨的特殊影騎影格式的縣齡建立流程為： 步驟-：將賴閱到的影像資料加上自訂獅(Header)，寫成一個錄影資 料檔，自訂檔頭裡包含物件行跡的資料，用來將物件行跡繪製在 電子地圖上；如果影像監控系統的電子地圖是屬於使用者自行建 構的，播頭也將包含相對應的電子地圖資料索引值，可供播放時 將物件行跡標示在使用者自行建構的電子地圖中。 步驟二：每-段屬於不同攝影機的影像資料最前方會有—個特殊的^此 用來紀錄該攝韻崎下來的_格式，#_不同影像壓縮格 式的影片片段時’將會產生這一特殊格式的header。因此此特殊 格式可以具備在播放時切換影片解壓縮dec〇der並連續播放的功 能。 步驟二·在轉雜的過程中會同錢立該屬於職雜物件特徵及物件行 跡的開始_和結束時間索引存人物件行跡影像觀資料庫43 中，以方便查詢。 本發明中的行跡雜觀喊模組42為可播放行跡影像轉賴组Μ所轉 錄的具特殊影像錄影格式的彡儲，播放的步驟如下： 步驟一：魏餐行跡轉錄觀巾__扮㈣，她其巾的物件行跡 的資料和相對應的電子地圖資料索引值。如同物件行跡視覺化子 系統40標示物件行跡的方式，依照攝影機地點資料的表示方式， 將物件行跡標示在電子_服射，或是仙者自行建構的電子 地圖中。 201242368 步驟二：讀取包含影片壓縮格式資訊的檔頭，依照影片壓縮格式(c〇dec)解 壓縮(decode)影像資料並播放影片；之後不斷讀取檔案中的接下 來的資料，如果在一次讀取到包含影片壓縮格式資訊的檔頭，則 依照讀取到的影片壓縮格式(Codec)來切換dec〇der來解壓縮接下 來讀取到的影像資料，並播放。 步驟三：不斷重複步驟二直到檔案中的資料讀取完畢。 因此當使帛者拉查詢已雜分析和賴賴物件⑽時，物件行跡視覺 化子系統40可節省再次物件行跡建構的時間，快速重現物件行跡。 本發明_上述步驟，除可從智慧型監㈣統的影像辨識結果中 建構出大範圍空間中的物件行跡外’同時可將行跡資訊標示於電子地圖服 務或使时自行建制電子細當巾’及依賴放雜將行师訊作視覺 化的呈現，並可將前述·覺化的影像轉錄鱗—錄影财並喊重現以 節省重複行跡篩選的時間。 以上所述係藉由實施範例說明本發明之特點，其目的在使熟習該技術 者能瞭解本發明之«財施，”蚊本麵之專利朗，故，凡 其他未脫離本發__之精神所完叙等效修飾雜改，储包含在以 下所述之申請專利範圍中。 201242368 【圖式簡單說明】 請參閱有關本㈣之詳細綱及其_，將可進—步瞭解本㈣之技術内 容及其目的功效；有關附圖為： 圖-為本㈣智慧鄉雜㈣統物件行跡追蹤纽之线架構圖。 圖二為本發明智慧型影像監⑽統物件行跡追縱纽之系統模組圖。 【主要元件符號說明.】 10 影像監控子系統 11 前端取像模組 12 影像收集模組 13 影像錄影模組 14 影像資料庫 20 影像辨識子系統 21 影像辨識模組 22 影像解壓縮模組 23 辨識結果資料庫 30 物件行跡查詢子系統 31 攝影機地點資訊建構模組 32 攝影機地點資料庫 33 行跡篩選模组 34 物件行跡資料庫 40 物件行跡視覺化子系統 41 行跡影像轉錄模組 42 行跡影像檔案回放模粗 43 物件行跡影像檔案資料庫 50 攝影機201242368 VI. Description of the invention: [Technical field to which the invention belongs] The present invention relates to a kind of wisdom and image monitoring of the object of the money (4), especially for the use of large numbers and scales _ ship transfer (10) image monitoring Department s [Prior Art] Image recognition technology has been widely used in the image of New Age, and the number of eye-catching glasses is to image the single camera image (such as license plate recognition, traffic flow, face detection, bit consumption, personnel). Chasing, directional movement _, electronic enveloping 'remains of objects, camera cover, face side, etc.), to be identified after the money, will be recorded (the turn number, vehicle color, vehicle _, vehicle brand thief, Big/small/number of locomotives, objects 妓, object size, personnel information, personnel legality, side, etc.) Information for storage and display, shouting for a more sloppy result between different cameras _, no test fA# Computing resources. Most of the image tracking techniques are based on the images captured by a single camera. The identified object traces are represented by a series of coordinates relative to the image plane, and the object movement track in a single picture is constructed. The scope is limited. In the Republic of China Patent No. (4) 142, it is revealed that the image is taken and the image is pursued. The image is captured by the image of the dual camera. Chasing, but the camera scene of the second camera is related, that is, although the dual camera is the same, but the location of the scene is the same as the scene 2, this can be traced to the difficult track of the ship. The rest of the object tracking technology that crosses the camera often has different cameras. 201242368 There must be some scenes with a weight of 4 _'s method to track the object track. It should also be limited to the -fixed area and the tight seams of the rhythm must be higher, that is, the same range must be more than the woven condition to complete the cross-camera tracking. ' Such an whereabouts tracking system is bound to increase the cost of construction. It is based on the application of the results of the amplification of the image identification and the problem and the lack of the application of the tracking and tracking technology, which is the direction of the research and improvement of the present invention. SUMMARY OF THE INVENTION The object of the present invention is to provide an intelligent image monitoring system object tracking system that utilizes image recognition across different cameras to establish a relationship between time and space, and then constructs a path of identification. Information, and mark the whereabouts information in the electronic map and play it in sequence to turn the track information into a ship, and (4) transcribe the visualized image into a single-view gamma-new (four) provincial repeat line _ selected daily sales, Therefore, inventors in view of the shortcomings of the above-mentioned conventional methods, (4) to change and after painstaking research, 'finally succeeded in the development of this wisdom and health (four) patch line vertical system. Compared with the conventional image recognition technology in the image monitoring (four) system, the invention further utilizes the image recognition results across different cameras to establish a relationship between time and space, and then constructs a line of information for identifying the target; w(4) Across the fine _ pieces of traces to trace the skills of this hair _ 额 彳 彳 追 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The large area of the camera _ the vertical direction of the object. The purpose of the present invention is to provide a kind of intelligent image monitoring system object line 201242368 trace tracking secret solution problem and the lack of 'the hair of the duck across the different camera image recognition." Fruit A its establishment time and space _ 'There is to construct a line of information to identify the target, and to trace the line to the electronically, and to play the image in sequence to visualize the track information, and to transcribe the aforementioned image of the image. The job case is re-created to save time for repeated track screening. The Qin Hui-type image monitoring system object tracking system that meets the purpose of the month of the month includes: image monitoring subsystem, image recognition subsystem, object track query subsystem and objects Four lines of line visualization subsystem, etc. The towel image can be used to compress digital images, provide video and access video files. The image recognition subsystem can provide different numbers of stitches. Recording, storing and querying the image identification results; the object traces the round (4) sample, the silk 'f peppers' location related body information, _« touch pieces _ Wei Wei; The object thief feels that the money can be used to trace the object to the W_t towel, the handsome _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The video image and the track information are transcribed into a single-track image (4), and the above-mentioned smuggling of the smuggling of the smuggling of the smuggling of the smuggling of the smuggling of the smugglers. _ Painted coffee:, its paste = ^ surgery wire lost (10) Shun track which coffee is on a single ^ 鄕 _, after the memorization, noisy «Information storage is a waste of large = 1: _ machine 0 between the image identification As a result, the invention makes more effective use of the source, and the invention further utilizes the image recognition of different images across different cameras to establish the relationship between time and space, and then constructs the line vertical information of the identification target. 2. Tracking of image recognition Most of the technology takes care of the images taken by a single camera, and the identified object traces are represented by a series of coordinates relative to the image plane, thereby constructing a moving track of the object in a single painting, and its standard financial limit; Republic of China Patent Certificate Number ... exposes - the rhyme depends on the tracking green, the hybrid Wei Shi Weng image and the target image, 'there is the use of the identification of the image of the dual camera to carry out the object tracking, but the two photographic scenes are the same That is, although the dual camera is used, the fine-grained location is the same as the % view. Therefore, the track range of the object can be traced only to the same scene; while the other object tracking techniques that cross the camera are often different. There must be some overlap between the cameras, so that there is a way to track the object. The scope of application must also be limited to the fixed area, and the density of the camera frame must be higher than the normal situation, that is, the same range It is necessary to set up more cameras than the general situation to have the way to complete the tracking of the cross-camera. The secret of the pursuit of the secret will inevitably increase the cost of construction; and the way of cross-camera tracking of the present invention can be increased without In the condition that the density of the photo frame is set, the use of the object in each location of the camera is used as a reference for the construction of the track, and the object can be traced across a large area of the camera. The whereabouts. 3. The county's eye-catching light supervisor (four) unified market exhibition gallbladder, Weng Huizhen image recognition function is included in the image monitoring system, in addition to the identification results available for inquiry, the image of the image recognition results further expanded Analyze the results of a large number of scales, construct the traces of the objects from them, and visualize them with the information of the traces of the objects, and mark the traces in the map service or in the custom map and the objects according to the objects. Position, the corresponding camera image # segments are sequentially played to facilitate tracking of objects; the present invention can further transcribe the aforementioned video clips and object tracking information of 201242368 into a single building, saving information on the object tracking again. The query time of the intelligent image monitoring system object tracking system can be applied to the object tracking of different cameras (locations) of the intelligent image monitoring system, which can assist in the pursuit of criminal vehicles, silk vehicles, indiscriminate population, suspects, The traces of suspicious objects, stolen objects or stolen objects will help to promote image surveillance systems in the coming years. The detailed description above is intended to be illustrative of the possible embodiments of the present invention, but the embodiments are not intended to be used in the context of the patents of the present invention. The patent scope of this case. Knowing that the above-mentioned "the case is not only innovative in terms of technical thinking, but also has the traditional functions of the conventional use", it has fully complied with the newly brewed and progressive French New Zealand patent requirements. Riding your office to approve the invention patent towel request, in order to invent the invention to the sense of virtue. [Embodiment] Please refer to the system architecture diagram of the object tracking system of the intelligent image monitoring system of the present invention. Please refer to FIG. 2 , which is a system module diagram of the object tracking and tracking system of the intelligent image monitoring system of the present invention. The image monitoring subsystem 10 includes a front-end image capturing module u 'image collecting module... Image recording age 13 «image Decoration 14; 郷 赖 ( (4) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The module 33 and the object are smashed by 34; the miscellaneous pieces are found on the side of the miscellaneous item (4) with fine image transcription 201242368, the module 4 track image broadcast playback module 42 and the object track image file database 43. The function of the object track tracking in the invention mainly comprises analyzing the data collected by the image monitoring system with image recognition function (10) to complete the construction of the object track, and the analysis result will be stored in the database for query, and then the result of the analysis is used. Object track query and object track visualization. The data collected by the image monitoring system of the present invention includes camera location related information, proximity camera information 'camera image and image recognition result. The first information to be collected by the present invention is the information of the camera location and the information of the adjacent camera. The collection method of photography points can be divided into two types: the first method is to collect the location information returned by the front-end equipment (such as (4)嶋) with TcP/IP transmission capture by the image collection. 2 This is for the camera to listen to the asset key module, allowing the user to manually enter the information of the camera location. When the first mode is adopted, the present invention has the TGp/Ip transmission function. (4) The front-end device is installed with the GPS device. When the image collection module 12 is connected to the front-end device for the first time, the front-end device transmits the TCP/IP. The method transmits the latitude and longitude of the GPS granule device to the image collection module 12; the image collection group 12 stores the collected camera location data in the camera location database 32. This method is suitable for the image surveillance system of the intersection of the camera in a large area. The GPS latitude and longitude data can be displayed in conjunction with the existing electronic map service. Therefore, this month has the function of displaying the object's whereabouts on the current electronic map service. The first method is mainly used in the position of the camera rack. It is not suitable for using the GPS positioning device. For example, if the camera frame is located in a large store or a building, the latitude and longitude data obtained by the camera on different floors may be repeated. Therefore, it is not suitable to use the Gps latitude and longitude data to identify the location of the field; this (4) provides - the location of the filming material _31, so that the user 201242368 - can be used - convenient thief user interface, self-construction of the ground _ please, The information of the camera location provided by the user is stored in the camera location database 32, where the camera location can be recorded using the location corresponding to the electronic location, and the location of the camera is recorded. There is a function of displaying the vertical line of the object on the electronic map provided by the user; in this way, the electronic map is also stored in the database - the index is available for query. The invention (4) close-up weaving information wealthy style roots and enthusiasm shirts are also divided into two types. · The first type is the camera location data using Gps latitude and longitude data representation, using the calculation method to calculate the camera location is the profit neighbor And according to the calculation result, the information of the adjacent camera of each camera is established and stored in the camera location Longku 32; the second way is the camera location _4_ The secret object __ frequency, the location information The construction module 31 provides a convenient user interface to provide the user with the proximity of each camera to the camera and to store the location (4) in the library 32. The image collection module 12 collects the digital image of the various compression formats from the image collection device in the front-end image capturing module u. The video recording group 13 will collect fine and fine, and the camera: timely _ axis people _ _ towel As Chana. Front-end image collection. Also use the current (four) various shadow level stitching (_ pressure back age card, hardware compression image Naka, front camera compression, IPCamera, Vide〇s (10) r, earn, 臓) to collect the side of the er: miscellaneous, and then 働Material · PEG, face side pin, 嶋嶋 __ fine 3, face mask module 13 4 ΓΓ ΓΓ — 录 ' 并 并 并 并 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及Eli images are read. This month, the video image data of each camera of the smart image surveillance system is imaged 201242368, because the camera images collected by different imaging devices will have different compression formats (such as Li, MPEG4, η·263, η 264, etc., so the video image data obtained by the video recording module 13 must be decompressed according to different compactness modes by the image decompression module 22 to obtain a miscellaneous f Material, video job · 21 can be used for the original image (4) for each image weaving, and consider __ part (such as license plate number, vehicle color, vehicle surface, vehicle brand model, large / small / number of locomotives, object position, Object size, personnel information, personnel legitimacy, directionality, etc., together with time-related information (identification of the time at which the object appears) and spatial information (location information of the front-end camera) are stored in the identification result database 23. / trajectory query subsystem 3 〇 according to the object trajectory has been completed, the mosquito cockroach to carry out the construction of the object track: in the case of gamma _ _ object track forest in the object track database % in the case of the track screening module 33 The object track will be constructed according to the identification result and the location of the camera, and the object track data will be transferred to the object track visualization subsystem 4() to display and play the image, and the image data and the object track data will be transferred to the track image transcription module 41 for transcription. Into the special image video age format _, the track image transfer group 41 will be in the transcription _ time in the object track surface coffee library 43 to build the average, in order to make the _ have been constructed object track; in the user's query In the case where the object track already exists, the object track visualization subsystem 40 directly visualizes the visual action of the object track. The track screening module 33 queries the identification result database 23, and the present invention uses the object to appear in each _ _ place as the object track construction, because the data towel contains the time warfare of the object and the front camera location information, The track of the identification target is constructed, and the track information is the group's time-ordered dumping (object feature, object location photography 201242368 machine location, object timestamp), and the track screening module 33 will store the obtained track information. In the object trace database 34. The query method is as follows. The track riding mode (10) selects the data columns with similar object features in the adjacent time and sorts them according to time; since the result of the query may be because the object descriptions are similar, the query result contains the wrong data column, and the track is at this time. The screening group 33 will - and query the information of the neighboring cameras committed by the camera location database to confirm whether the track data column presents the location __, and confirm the relationship to eliminate the unreasonable dragon column to increase the track. The forward rotation of the information. The object tracking visualization subsystem 4 is in the track: after the construction of the wire, the object track that the ship user wants to query is visualized, and the camera locations in the object track information are marked on the electronic map in order. And the line segments appear in each object, and at the same time, the video images of the cameras in the object track are read and played in time order. The way to mark the light trails of the object The location of the money camera (4) is divided into two types: the first method is to mark the object track in the electronic map service, and the camera location data is indicated by the Gps latitude and longitude data, which can be marked in the electronic The ground floor towel (such as G_e, the second way is to mark the object track in the self-constructed place, if the camera material is constructed by the user), it can be marked in the user's own electronic map (such as the floor plan of the building). The track image touch amount 41 will be transcribed into a single video recording of the special job video format in the job title and track data, saving the time for repetitive screening of object traces. The video playback in the secret video of the secret box will be taken out and the object will be taken out of the view, and the object track will be reproduced in the electronic map service or the user's own electronic map. Shaoguan can not be traced to the post of Yan Ling, can not meet the requirements of the 201242368 invention, the invention of the improved recording and weaving format to comply with the present invention Can be demanded. The process of establishing the age of the special shadow riding format is: Step-: Add the image data that you have read to the custom lion (Header), write a video data file, and include the object in the custom file. The trace data is used to draw the object trace on the electronic map; if the electronic map of the image monitoring system is constructed by the user, the broadcast header will also contain the corresponding electronic map data index value, which can be used for playback. The track is marked in the user's own electronic map. Step 2: Each piece of image data belonging to different cameras will have a special one at the forefront. This is used to record the _ format of the camera. #_不同影像When compressing a video clip in a format, 'this header will be generated in this special format. Therefore, this special format can have the function of switching the movie to decompress dec〇der and play continuously during playback. Step 2 · Will be the same in the process of turning the mix The start and end time index of the belonging character and the object track are stored in the character image view database 43 for convenient inquiry. The sneak screaming module 42 is a sneak peek of a special video recording format transcribed by the traversing image. The steps are as follows: Step 1: The Wei dynasty transcription transcript __ Dress up (4), she The information of the object track of the towel and the corresponding electronic map data index value. Just as the object track visualization subsystem 40 marks the object track, according to the representation of the camera location data, the object track is marked on the electronic _ service, or The electronic map constructed by the immortal. 201242368 Step 2: Read the header containing the information of the video compression format, decompress the image data and play the video according to the video compression format (c〇dec); The next data, if the header containing the information of the video compression format is read once, the dec〇der is switched according to the read video compression format (Codec) to decompress the image data read next. And play. Step 3: Repeat step 2 until the data in the file is read. Therefore, when the viewer is asked to query the miscellaneous analysis and the dependent object (10), the object track visualization subsystem 40 can save time for reconstructing the object track and quickly reproduce the object track. The invention _ the above steps, in addition to constructing a large-scale space object trace from the image recognition result of the intelligent supervisor (four) system, at the same time, the track information can be marked on the electronic map service or the self-made electronic fine towel And relying on the miscellaneous to visualize the appearance of the clerk, and can transcribe the aforementioned sensational images - video and reproduce to save time for repeated track screening. The above description illustrates the features of the present invention by way of examples, and the purpose of the present invention is to enable those skilled in the art to understand the patent of the present invention, and the patent of the mosquito face, therefore, the other is not deviated from the present invention. The spirit is finished with the equivalent modification and modification, and the storage is included in the scope of the patent application described below. 201242368 [Simple description of the schema] Please refer to the detailed outline of this (4) and its _, you will be able to further understand this (4) The technical content and the purpose of the purpose; the relevant drawings are: Figure-Bui (4) Wisdom Township (four) system traces of the tracking track of the object track. Figure 2 is the system model of the intelligent image surveillance (10) system of the invention Figure 1. [Main component symbol description.] 10 Image monitoring subsystem 11 Front-end image capture module 12 Image capture module 13 Video capture module 14 Image database 20 Image recognition subsystem 21 Image recognition module 22 Image decompression mode Group 23 Identification result database 30 Object tracking query subsystem 31 Camera location information construction module 32 Camera location database 33 Track screening module 34 Object track database 40 Objects Track Visualization Subsystem 41 Track Image Transcription Module 42 Track Image File Playback Modem 43 Object Track Image Archive Library 50 Camera