201113835 六、發明說明: 【發明所屬之技術領域】 法 本發明與監視_纽錢,_是_智絲像偵測方 【先前技術】 目刖市面上的監視攝影系統,可清晰地龄 旦炎 f ’同步地顯示於一監視器’以供人員 ς烹 s;rr件發生過程。例如,當 播放衫帶’綠認物品失竊時間及 旻』置複 攝物_無法提供 上等 僅下f效的影像。然而針對攝影機本身的防護,都 剪匕斷m偵測回路只能達成攝影機與錄影機之影像傳輸線遭 檣制為Γ低人工監看的負擔,一般數位錄影機内建移動偵測 作比對影像,判斷影像中是否有快速移動物件。 變 因短暫之天氣變化、人員移動、光源 Ϊ晝Ϊ 具有實際應用效益,因此,仍需要人工監 【發明内容】 人力具有制關題,㈣要持續以 為解決上述問題,本發明實施例提出一種智慧影像偵測方 201113835 ίΐΐ用於—智慧型影像侧系統,該智慧懸像偵測系統包 ^撬=視主機及至少一攝影機。該方法包含下列步驟:以該攝 2娜-場景之影像_ ’透過該監視域將該影像圖框切 二J複數個區域;以該監視主機逐一檢查各區域,記錄各該區 兮;=始之背景區域;每間隔一取樣時間後,以該攝影機擷取 彡像圖框,傳送至該監視域,_監視主機將該影 if框切分為魏健域;靖監控域逐—檢查各該區域, 始之背景區域相同之區域記錄為背景區域,不同之區域 3錄為物件區域;以該監視主機分析物件區域並建立物件模 f析物件特徵;依據該物件特徵,該監控主機將該物件 二··疋義為物件,透過分析該物件的變化,判斷一事件發生的 種類’並進行對應動作。 #刑實施例更提出—種智慧影像侧方法,應用於一智 純,該智慧郷像_祕包含-監視主機及 識。該方法包含下列步驟:以該攝影機· 一場景 像^框,透過碰視主機將郷賴购分為複數個區 旦視主機逐一"^查各區域,記錄各該區域為初始之背 i —取樣時峨,明攝影機娜該場景之影像 J區域:之區:記錄為背景區域:不各=域== =,,以該監視线分析物件區域_场件麵並分析:件 =徵,依_物件特徵,該監控主機將該 lit細㈣輸’ _—梅細類,^ 本發明實施例更提出-種智_影像_祕,包含一龄 現主機及至少-攝影機,肋執行前述之智慧影像偵測方法二 本發明實施縦糾-的 監視主機’用以使該監視主機執行前述之象 201113835 留或遺失發出馨針對靜態物進行摘測’對物件異常遺 勿、。同時可學習場景因外在環境的變化5或是 心略常·祕動之巡·備人員,而降低發出誤報的Ϊ率。戈疋 【實施方式】 露之U:第2圖」所示,為本發明實施例所揭 备姑t Γ (像偵彳方法,該方法應用於—智慧型影像侧 110 以分1〇内安裝一智慧影像監控整合軟體,用 訊至各f框(Fr3me)的變化,傳送結果或警 於監視主'i 位攝影機120,直接連接 視主播110八把々ΐί數動祕像至監視主機110,以供監 、番邻+Λ—刀^各圖框。數位攝影機120亦可透過TCP/IP等 通訊協疋,財_路或無_路連接於監視主機。 為傳統類比攝影機130,連接於—錄影機140, 己錄傳統類比攝影機130所榻取之動態影像,在 ίη一二數他號後傳送至監視域11G,以供監視主機 110分析各圖框。 ~ 包含圖」所示,本㈣揭露之智_像侧方法’ 以攝影機擷取一場景之影像圖框(S210),透過該監視主 機110將影像圖框切分為複數個區域(S22〇),如「第3圖 所示。 以該監視主機11G逐-檢查各區域,記錄各區域為初始之 背景區域(S230)。此一步驟係用以確認該被監視之場景中, 有哪些物件為固定不動的背景,有哪些是被後續加入的物件。 ,每間隔一取樣時間後,以攝影機擷取該場景之影像圖框, 傳送至監視主機110,以監視主機11〇將該影像圖框切分為複 201113835 數個區域(S220),如「第4圖」所示。 以該監控主機11〇逐一檢查各區域,將與初始之背景區埴 J9錄為背景區域(S23G),不同之區域記錄為物^ ίΪ丄?县i,t「第5圖」所示。此-步驟係用以確認該被 s視之%景中,有哪些物件為固定不動的背景, 續加入的物件。 疋饭设 以該監視主機110分析物件區域並建立物 (S250),以進行學習更新,並分析物件特徵。 、 杜特徵,該監控主機n°將該物倾域定義為物 ===過分析該物件的變化’判斷事件發生的種類, 並進行對應動作。 如「「第5圖」」所示,當該物件為新增物件時, ,,紀錄該物件為一遺留物,並標記遺留物出現之時間&。 若該場景巾不應有遺留物出現,則該監控主機發出一異當 遺留物警訊至保全主機2〇〇。 、 如「第6圖」所示,當原有之物件不再出現,則該監控主 遺失物’並標記遺失物遺失時的時間^。 參 為需放置於該場景,則該監控主機100發出 遺留物遺失警訊至該保全主機2〇〇。 親主ί I 之外,攝雜之影像紀錄也可_錄存於監 機’透聯驗域110分析—場景之影像紀 =象ί錄之一影像圖框,加以切分為複數 G t i ί。ϊΐ再依據時序,透過該監控主機110掘取該 如像、.·己錄之另-練_,加以分析比對之0使用者 定區域進行搜尋’針對場景内新增之遺留物 偵測辨識,並標記時間點及註記,方便使用者快 201113835 透過智慧影像監控整合軟體,當姻監視範圍内有異常物 1消失或滯㈣,可發提出警告。例如在公眾場所反恐任務 時’可對監視場景中被放置的爆裂物標記為異常遺留物,再發 出警訊通知警備人員前往察看。避免人力地毯式 是巡邏班㈣間差造成失誤。 ^㈣$ 如「第7圖」所示,當物件所佔有的物件區域與背景區 的比例超過-門植值時,該物件可能為一遮蔽鏡頭之遮蔽物, 例如喷漆,此時該監控主機11〇發出一鏡頭遮蔽之警訊至該保 如「第8圖」所示,當區分之區域經比對後與初始之背景 區域及物件區域完全不同時,新的場景將整的翻斷為物件^ 域’形成佔騎有區分區域社物件。_攝影機已經完全被 轉向或是完全失焦,使得鏡頭完全擷取到另一個不同之場景, 此時該監控主機110發出鏡頭轉向之警示至該保全主機2〇〇。 除了即時監測之外,攝影機之影像紀錄也可以被錄存於監 視主機或是錄影機,透過該監控主機11〇分析一場景之影像= 錄’-時序擷取該影像紀錄之一影像圖框,加以切分為複數個 ^域分析^使用者可針對已錄存之影像指定特定區域進行搜 尋’針對場景内攝影機遭到以轉向/失焦/喷漆/以物遮蔽等進行 镇測辨識’並標記時間點及註記,方便使用者快速檢索及確認。 4實務上’攝影機因為設置場所的問題,可能存在風力造成 之背景變化’例如:水蚊波浪、觀之晃動,或是光影之變 化。、因此,分析物件區域時,該監視主機11〇同時會分析物件 區域相對於對應之触之背景區域的變化量及變化持續的時 間,變化持續時間超過一時間門檻值時,才界定該區域影像為 物件區域。當影像變彳b量超過—變化服值,才界定該物件區 域,-物件則透過―學習程序,將物件區域作為新的初始 之背景區域。前述之學習程序是為了避免背景中有快速移動而 201113835 不停留之物體’例如一般通行之人員、水面之波浪、樹葉之晃 動。如此一來’便不會因為上述事件造成誤報。 此外場景的明暗變化,例如太陽隨時間移動後,對場景中 各區域也會出現全面性之變化而被全部標記為新的物件區 域’但對應背景區域的變化量小於變化門檻值,透過一學習程 序’將新的物件區域作為新的初始之背景區域。如此一來,便 不會因為上述環境變化造成誤報。 【圖式簡單說明】201113835 VI. Description of the invention: [Technical field of invention] Law of the present invention and monitoring _ New Money, _ is _ Zhisi image detection party [Prior Art] Seeing the surveillance camera system on the market, can clearly clear the age of inflammation f 'synchronized display on a monitor' for the person to cook s; rr pieces occur. For example, when playing the shirt with the ‘Green Recognized Items’ theft time and 旻 置 置 _ _ 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法 无法However, for the protection of the camera itself, the detection loop of the camera can only achieve the burden of the image transmission line of the camera and the video recorder being reduced to manual monitoring. Generally, the digital video recorder has a built-in motion detection for comparing images. Determine if there are fast moving objects in the image. Due to the short-term weather changes, personnel movement, and light source Ϊ昼Ϊ, there are practical application benefits. Therefore, manual supervision is still required. [The content of the invention] The human being has the problem of the system, and (4) It is necessary to continue to solve the above problem. The embodiment of the present invention proposes a wisdom. The image detection party 201113835 is used for the intelligent image side system, and the smart image detection system includes a video camera and at least one camera. The method comprises the steps of: cutting the image frame into two areas through the monitoring field by using the image of the camera-scenario; and checking each area one by one by the monitoring host, and recording each area; The background area; after each sampling time, the camera captures the image frame and transmits it to the monitoring domain, and the monitoring host divides the shadow if box into Wei Jian domain; the Jing monitoring domain checks each area by The area with the same background area is recorded as the background area, and the different area 3 is recorded as the object area; the object of the object is analyzed by the monitoring host and the object model is analyzed; according to the characteristics of the object, the monitoring host selects the object. · Derogatory meaning is an object. By analyzing the change of the object, it is judged that the type of an event occurred' and the corresponding action is performed. The #刑实施例 is further proposed as a method of intelligent image side, which is applied to a purely intelligent, intelligent image, including the monitoring host and the knowledge. The method comprises the following steps: taking the camera and a scene image frame, and by looking at the host, dividing the purchase into a plurality of areas, the host is tracked one by one, and each area is recorded as the initial back i. When sampling, 明, the camera is the image of the scene J area: the area: recorded as the background area: not = domain == =, analyze the object area _ field surface with the monitoring line and analyze: piece = sign, according _ object feature, the monitoring host will be thin (four) lose ' _ - mei class, ^ the embodiment of the invention further proposes - 智 _ image _ secret, including a first-generation host and at least - camera, ribs to perform the aforementioned wisdom Image Detection Method 2 The monitoring host of the present invention is configured to enable the monitoring host to perform the aforementioned image like 201113835, leaving or missing the scent to extract the static object for the object. At the same time, you can learn the scene due to changes in the external environment 5 or the staff of the patrols and secrets, and reduce the rate of false positives.疋 疋 [Embodiment] U: 2, which is shown in the figure of the present invention, is a method for detecting a t Γ (like a detection method, the method is applied to - the smart image side 110 is installed in one inch A smart image monitoring integrated software, using the change to the f frame (Fr3me), transmitting the result or alerting to monitor the main 'i bit camera 120, directly connecting the anchor 110 to the monitoring host 110, The camera can be connected to the monitoring host via TCP/IP, etc., and the traditional analog camera 130 is connected to the The video recorder 140 records the moving image taken by the conventional analog camera 130, and transmits it to the monitoring domain 11G after the ίη一二数号, for the monitoring host 110 to analyze each frame. ~ Included in the figure, this (4) The image of the image is captured by the camera (S210), and the image frame is divided into a plurality of regions (S22〇) by the monitoring host 110, as shown in FIG. The monitoring host 11G checks each area one by one, and records each area as an initial Scene area (S230). This step is used to confirm which objects in the monitored scene are fixed and which are subsequently added. After each sampling time, the camera captures The image frame of the scene is transmitted to the monitoring host 110, and the monitoring host 11 divides the image frame into a plurality of areas of 201113835 (S220), as shown in "Fig. 4". Each area is checked one by one, and the initial background area 埴J9 is recorded as the background area (S23G), and the different area is recorded as the object ^ Ϊ丄 县 county i, t "figure 5". This - step is used to confirm In the view of the % view, which objects are fixed backgrounds and continued objects are added. The rice cooker is configured with the monitoring host 110 to analyze the object area and establish objects (S250) for learning update and analyzing object characteristics. , Du characteristics, the monitoring host n° defines the object dumping area as the object === over analysis of the change of the object 'determine the type of event occurrence, and perform the corresponding action. As shown in ""5th figure", When the object is a new item, Recording the object as a remnant and marking the time of occurrence of the remnant. If the scene towel should not have any remnants, the monitoring host sends a special remnant warning to the security host 2 . As shown in Figure 6, when the original object no longer appears, the main lost item is monitored and the time when the lost item is lost is marked. ^ When it is required to be placed in the scene, the monitoring host 100 issues a legacy. Lost warning to the security host 2 . In addition to the ί I, the image record of the video can also be recorded in the surveillance machine 'through the analysis of the domain 110 analysis — the image of the scene = one of the ί The image frame is divided into multiple G ti ί. Then, according to the time series, the monitoring host 110 is used to dig the image, and the other recorded _ is analyzed, and the search is performed on the user-defined area. And mark the time point and note, so that the user can quickly integrate the software through the intelligent image monitoring 201113835. When there is an abnormality 1 disappearing or stagnation (4) within the monitoring range, a warning can be issued. For example, in the case of anti-terrorism missions in public places, the explosives placed in the surveillance scene can be marked as abnormal remnants, and then a warning is sent to notify the guards to go to the inspection. Avoiding the human carpet type is the mistake caused by the patrol class (four). ^(4)$ As shown in Figure 7, when the ratio of the object area occupied by the object to the background area exceeds the value of the door, the object may be a shield for the shadowing lens, such as painting, and the monitoring host 11〇A warning shot of a lens is sent to the image as shown in Figure 8. When the differentiated area is completely different from the initial background area and object area, the new scene will be completely turned into The object ^ domain 'forms a part of the community. _ The camera has been completely turned or completely out of focus, so that the lens is completely captured to another different scene. At this time, the monitoring host 110 issues a warning of the lens steering to the security host 2〇〇. In addition to the instant monitoring, the image record of the camera can also be recorded in the monitoring host or the video recorder, and the image of the scene can be analyzed through the monitoring host 11 to record the image of the scene. Divided into a plurality of ^ domain analysis ^ Users can specify a specific area for the recorded images to search for 'cameras in the scene with steering / out of focus / painting / object shielding, etc." and mark Time points and notes are convenient for users to quickly search and confirm. 4 In practice, the camera may have background changes caused by wind due to problems in the installation site, such as water waves, shaking, or changes in light and shadow. Therefore, when analyzing the object area, the monitoring host 11〇 simultaneously analyzes the amount of change of the object area relative to the background area of the corresponding touch and the duration of the change, and defines the area image when the change duration exceeds a time threshold value. Is the object area. The object area is defined when the amount of image change b exceeds the value of the change, and the object is used as the new initial background area through the learning process. The aforementioned learning procedure is to avoid the object that does not stop in the background, such as the general person, the wave of the water surface, and the sway of the leaves. As a result, there will be no false positives caused by the above incidents. In addition, the light and dark changes of the scene, such as the sun moving over time, will also show a comprehensive change in each area of the scene and be marked as a new object area 'but the amount of change corresponding to the background area is less than the threshold of change, through a learning The program 'takes the new object area as the new initial background area. As a result, false positives will not result from the above environmental changes. [Simple description of the map]
第1圖為本發明實施例之智慧型影像偵測系統之系統方 塊圖。 第2圖為本發明實施例之智慧影像债測方法之流程圖。 第3圖為本發明實施例中,該監視主機將影像圖框切分 複數個區域之示意圖。 ’ ^第4圖為本發明實施例中,該監視主機逐一檢查各區域 記錄各區域為初始之背景區域之示意圖。 第5圖為本發明實施例中,該監視主機逐一檢查各區域 將與初始之背景區域相同之區域記錄為背景區域 記錄為物件區域之示意圖。 °° 之示圖為本發明實施例中’該監控主機標記物件為遺失 第7圖為本發明實酬巾,當物件所佔 景區域的_超過-門檻值之示意圖。㈣物件&域與 第8圖為本發明實施例中’區分之區 背景區域麟魏域完全糊之轉目。_後與初始 【主要元件符號說明】 100 監控主機 201113835 no 監視主機 120 數位攝影機 130 傳統類比攝影機 140 錄影機 200 保全主機FIG. 1 is a system block diagram of a smart image detection system according to an embodiment of the present invention. FIG. 2 is a flowchart of a smart image debt testing method according to an embodiment of the present invention. FIG. 3 is a schematic diagram of the monitoring host dividing the image frame into a plurality of regions according to an embodiment of the present invention. 4 is a schematic diagram of the monitoring host checking each area of each area as an initial background area one by one according to an embodiment of the present invention. In the fifth embodiment of the present invention, the monitoring host checks each area one by one to record the same area as the initial background area as a background area and records it as an object area. The diagram of °° is the embodiment of the present invention. The monitoring host marked object is lost. Figure 7 is a schematic diagram of the _ excess-threshold value of the scene of the present invention. (4) Object & Field and Fig. 8 is a sub-division of the area in the background area of the present invention. _After and initial [Main component symbol description] 100 Monitoring host 201113835 no Monitoring host 120 Digital camera 130 Traditional analog camera 140 Video recorder 200 Security host