201120817 六、發明說明： 【發明所屬之技術領域】 、本發明揭露-種㈣影像色織訊之比對處理系統與 方法’特別是色度空間轉換，並僅保留色彩資訊作為 影像比對的基礎，轉除可能產生錯誤辨識的問題。 【先前技術】 智能視頻監控（Intelligent Vide〇Surveillance，IVs)么 統是:種由閉路電視（CCTV)網絡、主控台的監視器和ς 全人貝構成。在積極發展中的智能視頻監控系統，能夠透 過數位影像分析，將所拍攝的影像進行影像特徵化、差異 =對等影像處理流程，找出移動影像或是分析出有意義^ 資訊’進而達成自動追縱、判斷，輔助安全人員找出可疑 對象。 在習知技術t，有使用閉路電視錄影配合安全人 監控模式，此類監贿式除了安全人員現場可以判斷出可 3象，通常是在特定事件發生後，透過檢視錄影内容， 4可罪監視器上的晝面找出有用的資訊。 ^著，經數位化後，傳統閉路電視改成了數位攝影機， 使得容易透過數位控财現自動_、判斷的功能 節省了影像儲存的空間與成本。 目前，安全監㈣技術進—步實現了智能視頻監控， 除了產生了多種的監控應用外，更應用到網路傳輸的技 術’包括執行目標物件制、追縱、異常事件判斷，並 而觸發警報，讓監㈣工作更強大而無遠弗屆。 在進行安全監控的狀況下，最重要之一的事情是對移 201120817 動物件的偵測，由於在數位攝影機的拍攝下，連續的影像 訊號是由複敫幀（frame)的圖像組成，先建立參考背景， 再連續比對複數幀影像，藉以得出在特定時間内產生的變 化，旎因此分析出發生於被監控區域的事件，接著透過擎 報機制告知相關保全人員。 因此’影像之間的晝素差異就變成影像分析的基礎， 可麥考美國專利公開第2008/0031493號（公開於2〇〇8年2 月7日）所揭露用於監控攝影機上的物件特徵辨識方法，201120817 VI. Description of the Invention: [Technical Fields of the Invention] The present invention discloses a method and method for comparing image processing of image color-dye-sense (especially chromaticity space conversion), and retains only color information as a basis for image comparison. , remove the problem that may cause misidentification. [Prior Art] Intelligent Vide〇Surveillance (IVs) is composed of a closed-circuit television (CCTV) network, a monitor of the main console, and a full-body shell. In the active development of intelligent video surveillance system, through digital image analysis, the image can be image characterized, difference = peer image processing flow, find moving images or analyze meaningful ^ information to achieve automatic chase Longitudinal, judgment, assist security personnel to identify suspicious objects. In the conventional technology t, there is a closed-circuit television video with a security person monitoring mode. In addition to the security personnel, the type of the bribery can be judged by the security personnel, usually after the occurrence of a specific event, by viewing the video content, 4 guilty surveillance Find the useful information on the inside of the device. ^ After the digitalization, the traditional closed-circuit television has been changed into a digital camera, which makes it easy to automatically control the space and cost of image storage through digital control. At present, the safety supervision (4) technology has realized intelligent video surveillance, in addition to generating a variety of monitoring applications, the technology applied to the network transmission 'including the implementation of target object system, tracking, abnormal event judgment, and triggering alarm Let the supervisor (4) work more powerfully and without far-reaching. In the case of security monitoring, one of the most important things is the detection of the moving animal parts of 201120817. Because of the shooting of the digital camera, the continuous image signal is composed of the image of the frame. A reference background is established, and the complex frame images are continuously compared to obtain the changes generated within a certain time period, thereby analyzing the events occurring in the monitored area, and then informing the relevant security personnel through the intelligence reporting mechanism. Therefore, the difference in the difference between the images becomes the basis of the image analysis. The characteristics of the objects used to monitor the cameras are disclosed in the US Patent Publication No. 2008/0031493 (published on Feb. 7, 2008). Identification method,
其中即如第一圖顯示透過色彩圖表進行物件特徵辨識的方 法。 如步驟S101，先由一連串變動中的影像中辨識出其中 的物件，再透過彩色圖表產生器，根據每幀中物件的顏色 與壳度的資訊產生出色彩圖表（color histogram，步驟 ^103)。再如步驟sl〇5，根據色彩圖表中的色彩分佈，將場 景與物件區分為不同的類型（步驟sl〇5)，之後可以為了能 f凸顯物讀徵’進行特定色彩空間的座標轉換（步驟 107)，再計算出各種類型中色彩分佈 咖），以稱出物件特徵（步驟S111)。 ^々驟 此例中’係透過色彩分佈的分析得出連續場景的變化 物件監控攝影機可以根據 泉老^於—般智能視頻監控的技術f是透過過濾背景得出 色^像4以數值分析來制變動狀況，但—般透過彩 生ς的變化來找出移動的變動’可能會被環境影響而產 各種可技術提出能夠透過簡化彩色影像來降低 %生决報的問題’包括利用灰階影像將不必要的 201120817 色彩排除，利用其中黑與白的變化找出變動的物體。但灰 階影像仍有可能被光線亮度的變動影響而產生誤報。 【發明内容】 習知技術中，經過攝影機錄製的影像可能會因為環境 光線、亮度對比產生錯誤判斷的問題，本發明所揭露的使 用影像色彩資訊之處理系統、處理方法與影像分類方法僅 指貝取前後影像中的色度空間中的色彩資訊，特別是將原來 影像所處的色度空間轉換為色彩-飽和度-亮度之色度空間 (Hue, Saturation，Intensity，HSI space)，並僅擁取其中色彩 資訊’以排除其他可能造成誤判的資訊。以此增加判斷的 準確性，以產生正確且有效的訊息。 根據較佳實施例，發明揭露一種使用影像色彩資訊之 ，理方法，其中步騾係先擷取一連讀影像，並設定一彖考 &像，貝際貫細(可以開始運作時的初始幀為參考影像，再 ^行色度空間轉換’特別是將原來影像所處之色度空間轉 」為色彩-飽和度-亮度之色度空間。接著，指頁取參考影 次物件之色縫訊，排除飽和度與亮度的資訊，建立來 貝訊。 夕1 進行二取後續影像，並對後續影像中各鳩 色又工門轉換’同樣轉換為色彩-飽和度 = = 續影像中物件的色彩資訊。《參 徵：4勿件的色彩資訊的匹配運算判斷是否符合物件 變動的判斷標準。其中， 了“勿件 配，則繼續擷取後續影像 —"右判斷為匹 則發出-訊息。像，但相斷不匹配（表示有變動）， 201120817 上述步驟中’可透過比對是否匹配 比如，不匹配可能表示物件正在 ^ :中 物件的:的不匹配資訊描繪出物件像; 匹配可此表讀影機拍攝的工不 機被外物遮蔽。 〜角度被改艾，或是攝影 或疋可針對特定空間巾的特 是否遺失’或是特定空間是否新增？： 施例是:透過不匹配的判斷發出觸牛= 開始錄製影像，或是進行其蜩I耠衫機 能。另有實施例更透過判斷 : 出的事件，ϋ而啟_數。. ❿麵有進 施例’發明所揭露的使用影像色彩資 ，之處理线包括有#陳_影像 一 _像== 而影像處理模組由儲存單 透過色度空間轉換單元將 ;^的〜像貝讯， 空間轉換為一色私 卜=各幀原本所處的色度 、m々 飽和度^度之色度空間（HSI空⑴， 取單元排除其中的飽和度與亮度的特徵， 用匹配運算判斷參考影像中心 宁的物件的色彩特徵是否匹配。 訊自取l，己判斷物件的變化，以決定是否要發出 心匕如_出物件有異常，即發出警報訊息。么出… 201120817 【實施方式】 由於監控中的影像往往會因為各種不確定的因子產生 錯誤警報的問題》比如利用監控攝影機對特定空間進行入 "(Χ監測’ ^ 間有任何改變時’攝影機所榻取的晝面就會 有改變’因而產生警報訊息’但是卻難避免因為光影、亮 度的變化而被誤判為可疑事件的問題。本發明揭露一種使 用影像色彩資訊之比對處理系統與方法，並包括一影像分 類方法，特別是利用色度空間轉換（C〇lor Space Transformation) ’轉換原本影像所處的色度空間，且透過僅 保留色彩資訊作為影像比對的基礎，以排除可能產生錯誤 辨識的問題。 ~Among them, the first figure shows the method of object feature recognition through the color chart. In step S101, an object is identified from a series of images in a series of changes, and then a color chart generator is used to generate a color histogram (step: 103) according to the color and shell information of the object in each frame. Further, as step sl5, according to the color distribution in the color chart, the scene and the object are divided into different types (step sl5), and then the coordinate conversion of the specific color space can be performed for the f-emphasis reading (step 107), and then calculate the color distribution in each type to weigh the object feature (step S111). ^ In this example, the analysis of the color distribution by the analysis of the color distribution results in a continuous scene change object monitoring camera can be based on the old-fashioned intelligent video surveillance technology f is obtained by filtering the background image 4 image by numerical analysis The state of change, but the change in the movement through the change of the color ς ' ' 可能 可能 可能 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' Unnecessary 201120817 color exclusion, using black and white changes to find out the changing objects. However, grayscale images may still be affected by variations in light brightness and false positives. SUMMARY OF THE INVENTION In the prior art, an image recorded by a camera may cause a wrong judgment due to ambient light and brightness contrast. The processing system, processing method and image classification method using image color information disclosed in the present invention only refer to The color information in the chrominance space in the front and back images, especially the chromaticity space in which the original image is located is converted into the color-saturation-intensity chromaticity space (Hue, Saturation, Intensity, HSI space), and only the content The color information 'to exclude other information that may cause misjudgment. This increases the accuracy of the judgment to produce a correct and effective message. According to a preferred embodiment, the invention discloses a method for using image color information, wherein the step method first captures a continuous reading image, and sets a reference image and a fine image (the initial frame that can be started when operating) For the reference image, then the chroma space conversion 'in particular, the chromaticity space where the original image is located is turned into the color-saturation-luminance chromaticity space. Then, the page refers to the color stitching of the reference image object. Exclude the information of saturation and brightness, and establish the information. In the evening, the second image is taken and the color is converted to the color in the subsequent image. The same color is converted to color-saturation = = the color of the object in the image is continued. Information. "How to match: 4 match the color information of the matching operation to determine whether it meets the judgment criteria of the object change. Among them, "Do not match, continue to capture subsequent images -" right judgment is the word is sent - message. Like, but the mismatch does not match (indicating that there is a change), 201120817 In the above steps, 'the matching is matched, for example, the mismatch may indicate that the object is in the object: the mismatch information of the object: The matching image can be blocked by the foreign object. The angle is changed, or the photography or 疋 can be lost for the specific space towel. The example is: by touching the mismatch, it is triggered to start recording the image, or to perform the function of the 蜩I 耠 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 The invention discloses a method for using image color, and the processing line includes #陈_影像一_像== and the image processing module is transmitted by the storage unit through the chromaticity space conversion unit; Convert to a color of private = the chromaticity space of each frame, m 々 saturation degree (HSI empty (1), take the unit to exclude the saturation and brightness characteristics, use the matching operation to judge the reference image center Ning Whether the color characteristics of the object match. The self-received l, has judged the change of the object, to decide whether to issue a heart, such as _ the object has an abnormality, that is, an alarm message is sent. What is going out... 201120817 [Embodiment] Due to monitoring Image often Because various uncertain factors produce false alarms, such as using a surveillance camera to enter a specific space " (when there is any change between monitoring '^, the camera will change when it is taken', thus generating an alert message 'But it is difficult to avoid the problem of being misidentified as a suspicious event because of changes in light and shadow. The present invention discloses a comparison processing system and method using image color information, and includes an image classification method, particularly using chroma space conversion. (C〇lor Space Transformation) 'Converts the chromaticity space in which the original image is located, and uses color information only as the basis for image comparison to eliminate problems that may cause misidentification.
共1P，衫像原本所處的色度空間可為一釭_綠_監色度竺 間（RGB color space)，但其中隱含有因為亮度、光影（女 太陽光變化）改變的色彩值，且相關門檻值設定困難。仓 若僅將其轉化成灰階表達，仍是以亮度為參考基準。因此 本發明特別是將原來影像所處的色度空間轉換為一色彩-食 矛度儿度之色度空間（Hue，Saturation，Intensity, HS 啊ce)’並僅擷取其t色彩（Hue)資訊，而排除其中飽禾 度，強度（梵度）的變數’為的是排除其他可能造成誤矣 的資訊’以色彩資訊增加判斷影像變化的準確性，以產生 正確且有效的訊息。 若將本發龍保全監控的領域，即可參考第二圖 像色彩資訊之處理系統之應用實施例示意圖。於建 j 設置有一或多數個攝影機23，攝影機23所擷取的 〜像可透過無線或有線的通訊方式傳遞至遠端監控中心 201120817 27,此伽無縣地"妨影像傳輸。 ,^過树__的現場影細㈣至監控中心27, 監控中心27具有影像處理的 ，轉換— (Hue)資訊，以此作= 除了 =上述監控_心27中實施影像處理外 例更可於攝影機23本機上为貝％ 換與色彩資訊操取功能'色度空間轉 遞至監控h 27。 ““統，觸處料的資訊傳 而备控的應用可包括： 測’係_攝影機鏡頭被不正常移動:、 遮蔽Γ、域貝及，係由拍攝的畫面消失判斷是否有異物 斷是否遺失^，中特定物件進行監控，判 車輛）進行監控； 曰不月物件（比如人、動物' 定的=:::測即針對所監控的空間是否有新增不明固 開關，而忽略影像不動時、.、D保全、啟動特定 電子圍籬、警戒區U : 即針對一特定範圍進；偵、過^控特定空間進行保全， 序； 4仃_獨事件，並觸發特定保全程 物件計數，透過 201120817 ‘ 物體的數量； 物件偵測，可以鎖定空間中特定物件進行 物件追蹤，偵測出影像變動，由卫， 的物件，並描繪出移動執跡，㈣^4判斷移動中 ㈣雜跡，攝域即能對此物件進行追 物件分類，可透過儲存於一資料庫 相對物件㈣訊，辦彡像概比職，分類各貝祕 以上各種顧係依據本發龍取之影像色彩（H 訊作為，斷標準’排除可能誤判的光影的因素後，再透過 匹配運算_匹配財，能提高細的準確度。 如第三圖敛述的本發明使用影像色彩資訊之處理方法 中計算匹配的步驟。 步驟開始時，如步驟S301，由一攝影機進行取像，經 數位化後’操取出影像中特定物件的江_綠-藍（RGB )資訊， 此例中’影像原處於紅'綠-藍色度空間，影像資訊即由紅、 綠、監顏色表達，經過色度空間轉換（步驟S3〇5)，轉換為 色形·飽和度-亮度之色度空間（步驟S3〇7)，並擷取當中物 件色彩（Hue)資訊，以此為參考，如步驟S3〇9，進行參 考影像與擷取影像間色彩資訊的特定匹配運算（c〇rrelati〇n operation)，判斷匹配程度（步驟S311)。其中匹配運算可 應用習知各種計算匹配的方程式，在此並不贅述。 接著’第四圖係為本發明所利用之色彩_飽和度_亮度之 色度空間不意圖。 第四圖顯示色彩-飽和度_亮度色度空間的座標形式之 一 ’此例之色彩-飽和度-亮度色度空間顯示為一由色彩、顏 色飽和度與亮度等資訊表達的圓錐體的立體座標，其中縱 201120817 向表示由最暗的黑到最亮的白，以亮度（Intensity) i表達。 而其中切面圓座標顯示在圖式右方，圓半徑表示飽和度 (Saturation) s，圓心為飽和度最低〇，而接近圓周的部份 則表示最大飽和度；而圓周由0度到360度則表示色$ (Hue )h，顏色依序由紅（Red )、黃（Yellow )、綠（Green )、 青綠（Cyan)、藍（Blue)到洋紅（Magenta)。 以上僅列舉色彩-飽和度-亮度色度空間的表達型式，實 際實施則不以這種為限制。A total of 1P, the chromaticity space of the shirt image can be RGB _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ And the relevant threshold value setting is difficult. If you only convert it to grayscale expression, it is still based on brightness. Therefore, the present invention particularly converts the chromaticity space in which the original image is located into a chromaticity space (Hue, Saturation, Intensity, HS ce), and only captures its t color (Hue). Information, and exclude the variable, the intensity (Vanity) variable 'is to exclude other information that may cause misunderstanding' to judge the accuracy of image changes with color information to produce correct and effective information. If the field of monitoring and maintenance of the present invention is used, the schematic diagram of the application embodiment of the processing system of the second image color information can be referred to. Yu Jian j sets one or more cameras 23, and the images captured by the camera 23 can be transmitted to the remote monitoring center through wireless or wired communication. 201120817 27, this gamut has no county. ^^ The scene of the tree __ is fine (4) to the monitoring center 27, the monitoring center 27 has image processing, and the conversion - (Hue) information is used as the = in addition to the above monitoring _ heart 27 in the implementation of image processing exceptions For the camera 23, the machine is switched to the color information operation function 'Chroma Space' to the monitoring h 27. “The application of the information that can be controlled by the device can include: “Measurement of the camera _ camera lens is not moving normally:, occlusion Γ, domain 及, and the disappearance of the captured picture to determine whether there is a foreign object is missing or not. ^, the specific object is monitored, the vehicle is monitored); the non-monthly object (such as human, animal 'determined =::: test whether there is a new undefined switch for the monitored space, and ignore the image when not moving ,., D security, start a specific electronic fence, warning zone U: that is, for a specific range; detect, control, control a specific space for preservation, order; 4仃_ unique event, and trigger a specific security item count, through 201120817 ' The number of objects; object detection, can lock specific objects in the space for object tracking, detect image changes, by Wei, objects, and depict the movement of the detour, (4) ^ 4 judge the movement (four) miscellaneous, photo The domain can be used to classify the object, and it can be stored in a database relative to the object (four), and it can be used to classify the various secrets. H. As the standard, the rule of 'excluding the light and shadow that may be misjudged, and then through the matching operation_matching money, can improve the fineness of accuracy. As the third figure cites the invention, the image color information processing method is used to calculate the matching. Steps: At the beginning of the step, in step S301, the image is taken by a camera, and after digitizing, the river_green-blue (RGB) information of the specific object in the image is taken out, in this case, the image is originally in red' green. - Blue degree space, the image information is expressed by red, green, and supervised colors, and converted into a chromaticity space of color form, saturation, and brightness by chromaticity space conversion (step S3〇5) (step S3〇7), And taking the object color (Hue) information as a reference, as in step S3〇9, performing a specific matching operation (c〇rrelati〇n operation) of the color information between the reference image and the captured image to determine the matching degree (step S311), wherein the matching operation can apply various conventional equations for calculating matching, and is not described here. Next, the fourth figure is the chromaticity space of the color_saturation_brightness utilized by the present invention. One of the coordinate forms of the color-saturation_luminance chromaticity space. The color-saturation-luminance chromaticity space of this example is displayed as a solid coordinate of a cone expressed by information such as color, color saturation and brightness. Vertical 201120817 indicates that the darkest black to the brightest white is expressed by Intensity i. Where the circular coordinates of the facet are shown on the right side of the figure, the radius of the circle indicates the saturation (Saturation) s, and the center of the circle is the lowest saturation. 〇, the part close to the circumference indicates the maximum saturation; and the circumference from 0 degrees to 360 degrees indicates the color $ (Hue )h, and the color is sequentially red (Red), yellow (Yellow), green (Green), Cyan, Blue to Magenta. Only the expressions of the color-saturation-luminance chromaticity space are listed above, and the actual implementation is not limited by this.
本發明所利用的影像資訊係需將所擷取到的原始由紅、 綠-藍色度空間表達的色彩，轉換至色彩-飽和度·亮度之色 度空間上，並排除飽和度與亮度的變數，僅以色彩（Hue) 為變數進行影像差異辨識。此處的色彩值可以角度表:示， 由〇度到360度，每種角度表示一種顏色。 轉換方式如下’此例僅為實施例之一，但實際實施教 不限制於此。比如，亮度以i= 1/3(R>G+B)表示，為紅、綠、 藍的平均值；飽和度以s = l-(3/(R+G+B))*a表示，其中 為紅、綠、藍中的最小值；色彩則可以h = e〇sA( ((〇.5*((R-G)+(R-B))) / ((R-G)-2 + (R>B)*(G-B))a(〇.5) 達。 另有習知技術（可參考美國專利第6,665,439號）以 列數學式進行色度空間，如方程式(1)所示： imax=max(r, g, b) imin=min(r, g, b) 亮度 i=(imax+imin)/2 當 imax=imin 飽和度s=0 201120817 當 i<=(max_value)/2 飽和度 s=(imax-imin)/(imax+imin) * max_value 當 i>(max一value)/2 飽和度 s=(imax-imin)/(max一value-imax-imin) * max_value rl=(imax-r)/(imax-imin) gl=(imax-g)/(imax-imin) bl=(imax-b)/(imax-imin) 當imax=imin，則色彩h並無定義 當 r=imax ’ 則色彩 h=((bl-gl);r/3 當 g=imax，則色彩 h=((2+rl-bl)TT/3 當 b=imax，則色衫 h=((4+gl-rl) ττ/3 其中r，g，b為各晝素的原始色彩值，jmax為r, g,乜的 最大值，imin為r，g，b的最小值，包括亮度丨與飽和度s , 範圍都是由0到最大值，若以g位元的記錄方式，最大值 為255，也就疋方程式(1)中的變數max_vaiue。而色彩則以 角度表示，由0到。 利用上述方程式⑴，將每個晝素（r(x, y)，g(x，y)，b(x, y))分別轉換為 h(x, y)，s(x，y)與 i(x，y)。 第五圖係為本發明使用影像色彩資訊之處理方法實 施例流程圖，此實施例流程主要是擷取出影像色彩 資訊作為影___參考，之後透過量化，比如計算 匹配’由匹配程度判斷影像變動，能應用於各種智能 監控（IVS)系統中。 θ 此處理方法步驟一開始，如步驟S5m.，+由一攝影機進 行拍攝’制是記錄（儲存至纖體或是特定暫存區）連 201120817 續的影像，接著，如步驟S503，透過影像處理單元執二 度空間轉換，如上述實施例，特別是將影像原處的色= 間轉換為本方法實施例所應用的色彩-飽和度-二空 空間。 色度 利用上述色彩-飽和度-亮度之色度空間所描述的旦 包括有色彩（Hue)、飽和度（Saturad〇n)與強度或= (Intensity)轉數，如步驟S5〇5，實施例僅掘取 定物件的色# (Hue)資訊，㈣除飽和度與強度 ^ 其中影像中⑽件可為❹者所選取，或是透過自動偵^ 變動的内容而判斷出。再如S5Q7，由最初在 下的 影像色f資”建立參考資訊，作為比對的基礎下的 在貫際實施上，連續影像有複數個巾貞（fram :试 :所謂色度空間轉換或是練的色彩資訊皆以帕為基礎單 接考，財法實蝴繼物取_ ，於步驟加中進行色度空間轉換，於本即 ::像原處的色度空間轉換為上述色彩_飽和度之 :巧。，後，如步驟灿，擷取 =配程度判斷參考資訊與後續影像^^配運异配 (步驟S515)。 貝。孔疋否匹配 於此判斷步驟中，主要是 ί，影像的色彩資訊間的匹配得出關二J像== 异則可參考其他習知技術，主要的#度而匹配運 參考資訊與後續色彩資訊間的差異。以適當地量化 上述透過匹配運算判斷關聯的程度，可透過設定另一[ 201120817 Π檻值來決定匹配程度是否符合—特定判斷標準 值可用以調整判斷標準，可解釋為敏感度。舉例 ： 匹配門檻值低（接近最小匹配程度），容易判斷為匹配 示敏感度低；若門禮值高（接近最大匹配程度），表〒 匹配的可能較低，則敏感度較高。 衣不判斷 步驟中，若判斷為匹配，則繼續顧取後 ，牛，藉Μ判斷參考影像與後續影像間的 也就如步驟S517，判斷是否超矛度 間的判斷步驟中，主_編炎本旦^時間在此延遲時 聯程度，也是就判斷與後續影像間的關 間，此延的叙β轉—延遲時 比如偵測過程中有特定被允許的人員進:成:丄 Τ ’以此延遲時間作為排除暫時被允許的事件：若不匹： 息（步驟st)的 了延遲時間或以上，則發出訊 間，判斷排除此輕主古右不匹配的情況並無維持此延遲時 驟（步驟S509)。日事件’則繼續進行擷取後續影像的步 過程:相關人員將進行處理，料後續處理 如解除罄報（牛^迷。之後’將解除此訊息相關資訊，比 行取像LlrS52l)，則回到最初步驟讀，重新執 又二間轉換、擷取色彩資訊、建立Λ考， 並後_取影像與計算匹配等步驟。建立參考貝几 特別的是，# & 算與延遲時間=上述參考影像與後續影像間的四配運 攝影機拍攝的若判斷為關聯程度較低，可能表示 置與角度被改變，或是攝影機被外物遮蔽。 201120817 或是，可針對特定空間中的特定物件進行偵 物件遺失，此影像_聯程度（參考匹配程度與時間 將表，來’包括色彩資訊產生差異’並且維持了一段:士 間。或是利用不匹配與維持了一段時間的資訊判斷= 間是否新增不明其他物件。 、一 另-實施例是，若上述影像間關聯程度較低，則發 觸發訊息’比如觸發攝影機開始錄製影像，或是進 定警報等功能。另有實施例更 。 則後關聯程度較低，而判斷有進出的事件， 進而啟動計數程序。 另有實施例如第六圖所示本發明使用影像色彩資訊之 處理方法_件分類實施例流程圖。 第五圖所料利用影像巾㈣資訊作為崎 此例則應用於物件分類。 件：:1::’:::601 ’進行取像，接著針對影像中特定物 件或疋工間擷取出物件色彩（Hue)資訊（步驟_3)。接 著’如步驟⑽5’引人―資料庫，此 影像物件的絲特徵，特別是經過轉換至色彩如Γ度亮L ^色度娜请像的色彩特徵，透過比對（步驟S607)，能 夠將所獅_像巾的物件撕錢（步驟謂），如流程 圖所示，將物件分類A _ Ε,Ι / μ 马頒別—（步驟S611)、類別二（步驟 則）與類別三（步驟S615)，或是更多類別。 £用像處理方法的系統可參考第七圖顯示的應 蚩 2禮。α —系統之專父佳實施例至少包括能擷取連續 ^ 、 早70丨’有—電性連接取像單元701的類比數 201120817 ，轉換早tl 703，通常透過鏡麵取的晝面為類比訊號，將 透過類比數位轉換單^3數位化，並暫存於相互電性連 接的暫存記舰犯。暫存記憶體705通常為設置於系統前 端的快閃記憶體（flash)或是動態隨機記 憶體（RAM)，或 可應用硬碟的中特定區塊。 系統中處理核心為處理單元7〇7，其電性連結各電路， 如圖所不’包括有暫存記憶體7()5、儲存單元、通訊單 tl 713與影像處理模經715。處理單元7〇7用以處理所擷取 的，t像’並由其_擷取出參考影像與後續影像。其中 儲子單7L709用於儲存參考影像與—段時間内擷取的後續 二像’可連接或疋内建一資料庫7ιι，根據上述實施例，資 711何別錢有各中影像、物件的資訊，記錄複數個 /類別的影像資訊，包括影像物件中經過色度空間轉換 k擷取的色彩（Hue)特徵。 、 另有通Dfl單元713’能以有線或無線通訊手段連接遠端 工，機房7G’控制機房7()能遠端控制此使用影像色彩資訊 =免理系統’並接收由該系統所傳遞的訊息，如應用於智 ，視頻監控系統上，透過控制機房7()的監控，可即時收到 各地保全系統傳遞的訊息。 4影像處理模組715由上讀存單it期得到影像資 -系、4主要透過景;像處理模組715中的色度空間轉換單 =71執订色度空間轉換，於較佳實施例中，將連續影像甲 貞原本所處的色度空間轉換為一色彩_飽和度-亮度之色 度空間。 名^過其中色彩特徵娜單元72擷取經色度空間轉換後 、彩（Hue)特彳政（可針對特定物件），並排除其中飽和 16 201120817 度與亮度的待彳致。 且根據其心==: 否 i:出 接著，第八圖至第十_圖顯示為=出吕似息。 施例，但實際實施並不限於此。本毛明之各種貫The image information utilized by the present invention is required to convert the original color expressed by the red, green, and blue degrees into a chromaticity space of color-saturation and brightness, and exclude saturation and brightness. Variables, image difference recognition is only performed with color (Hue) as a variable. The color values here can be angled: show, from twist to 360 degrees, each angle represents a color. The conversion method is as follows. 'This example is only one of the embodiments, but the actual implementation is not limited thereto. For example, the brightness is expressed as i = 1/3 (R > G + B), which is the average of red, green, and blue; the saturation is represented by s = l-(3/(R+G+B))*a, Where is the minimum of red, green, and blue; color can be h = e〇sA( ((〇.5*((RG)+(RB)))) / ((RG)-2 + (R>B) *(GB))a(〇.5) 达. Another conventional technique (refer to U.S. Patent No. 6,665,439) performs a chromaticity space in a mathematical formula, as shown in equation (1): imax=max(r, g, b) imin=min(r, g, b) brightness i=(imax+imin)/2 when imax=imin saturation s=0 201120817 when i<=(max_value)/2 saturation s=(imax- Imin)/(imax+imin) * max_value when i>(max-value)/2 saturation s=(imax-imin)/(max-value-imax-imin) * max_value rl=(imax-r)/( Imax-imin) gl=(imax-g)/(imax-imin) bl=(imax-b)/(imax-imin) When imax=imin, the color h is not defined when r=imax 'the color h= ((bl-gl);r/3 When g=imax, the color h=((2+rl-bl)TT/3 When b=imax, the color shirt h=((4+gl-rl) ττ/ 3 where r, g, b are the original color values of each element, jmax is the maximum value of r, g, 乜, imin is the minimum of r, g, b, including brightness 饱和 and saturation s The range is from 0 to the maximum value. If the g-bit is recorded, the maximum value is 255, which is the variable max_vaiue in equation (1), and the color is represented by angle, from 0 to. Using the above equation (1) Convert each element (r(x, y), g(x, y), b(x, y)) into h(x, y), s(x, y) and i(x, respectively. y) The fifth figure is a flow chart of an embodiment of a method for processing image color information according to the present invention. The flow of this embodiment mainly extracts image color information as a shadow ___ reference, and then passes quantization, such as calculating a match 'by matching. The degree of image change can be applied to various intelligent monitoring (IVS) systems. θ This processing method starts from the beginning, as in step S5m., + is shot by a camera. The system is recorded (storing to slimming or specific temporary storage). The image is continued with the image of the 201120817, and then, in step S503, the second spatial conversion is performed by the image processing unit. As in the above embodiment, the color of the original image is converted to the color applied by the method embodiment. Saturation - two empty spaces. Chroma using the color-saturation-luminance chromaticity space described above includes hue, saturation (Saturad〇n) and intensity or = (Intensity) revolutions, as in step S5〇5, an embodiment Only the color # (Hue) information of the object is extracted, (4) the saturation and intensity are removed. ^ (10) of the images can be selected by the latter, or judged by the content of the automatic detection. Another example is S5Q7, which establishes reference information from the initial image color. As a basis for comparison, there are multiple frames in the continuous image (fram: try: so-called chromatic space conversion or training) The color information is based on the PA-based single test, the financial method is followed by the material _, in the step plus color space conversion, in the original:: the original color space converted to the above color _ saturation : 巧., After, such as step can, capture = match degree judgment reference information and subsequent image ^ ^ distribution mismatch (step S515). Bay. Hole 疋 no match in this judgment step, mainly ί, image The matching between the color information can be obtained by referring to other conventional techniques, the main #度, and the difference between the reference information and the subsequent color information. The degree can be determined by setting another [201120817 Π槛 value to determine whether the matching degree is met - the specific judgment standard value can be used to adjust the judgment standard, which can be interpreted as sensitivity. For example: the matching threshold is low (close to the minimum matching degree), It is easy to judge that the matching indicates low sensitivity; if the door value is high (close to the maximum matching degree), the table 〒 matching may be lower, the sensitivity is higher. In the clothing non-judging step, if it is judged to be a match, continue to take care After that, the cow, by means of judging between the reference image and the subsequent image, as in step S517, determining whether or not the step of judging between the super-sprays, the time of the main_inflammation is delayed, and the judgment is also The interval between subsequent images, the delay of the β-turn-delay, such as the specific allowed personnel in the detection process: into: 丄Τ 'This delay time as an exclusion of temporarily allowed events: if not: If the delay time or above is delayed (step st), the message is sent out, and it is judged that the delay is not maintained (step S509). The day event continues to capture subsequent images. The step process: the relevant personnel will carry out the processing, and the subsequent processing such as releasing the report (the cow will be fascinated. After that, the information about the message will be released, and the line is like LlrS52l), then return to the initial step to read, and then re-execute two rooms. Conversion, Take the color information, establish the reference, and then take the steps of matching the image with the calculation. The reference to the Bayesian is specifically, # & calculation and delay time = the four camera cameras between the reference image and the subsequent image It is judged that the degree of association is low, which may indicate that the angle and angle are changed, or the camera is obscured by foreign objects. 201120817 Alternatively, the detection object may be lost for a specific object in a specific space. Time will be table, to 'including color information to make a difference' and maintain a paragraph: between the judges, or use the mismatch and maintain a period of information to determine = whether to add other unknown items. Another example - is that If the degree of association between the images is low, a trigger message is sent, such as triggering the camera to start recording an image, or setting an alarm. Other embodiments are more. Then, the degree of association is low, and it is judged that there is an event that enters and exits, and then the counting program is started. In addition, for example, the processing method of the image color information of the present invention shown in the sixth figure is shown in the sixth figure. The fifth picture uses the information towel (4) information as the weather. This example is applied to the object classification. The piece::1::’:::601 ' takes the image, and then extracts the object color (Hue) information for the specific object or the workroom in the image (step _3). Then, as in step (10) 5' to introduce the data library, the silk features of the image object, in particular, the color characteristics converted to the color such as the brightness of the L ^ chromaticity image, through the comparison (step S607), The lion _ the towel object tears the money (steps), as shown in the flow chart, classify the object A _ Ε, Ι / μ horse award - (step S611), category 2 (step) and category three (step S615), or more categories. For the system using the image processing method, refer to the answer shown in the seventh figure. α—The expert embodiment of the system includes at least the analog number 201120817 that can capture the continuous ^, early 70 丨 'electrically connected image capturing unit 701, and converts the early t1 703, which is usually analogized by the mirror surface. The signal will be digitized by the analog digital conversion unit ^3 and temporarily stored in the temporary storage of the ship. The scratch memory 705 is usually a flash memory or a dynamic random memory (RAM) disposed at the front end of the system, or a specific block in the hard disk. The processing core in the system is a processing unit 7〇7, which is electrically connected to each circuit, and includes a temporary storage memory 7 () 5, a storage unit, a communication unit tl 713, and an image processing module 715. The processing unit 7〇7 is configured to process the captured, t-images and to extract the reference image and subsequent images from the image. The storage list 7L709 is used for storing the reference image and the subsequent two images captured in the period of time can be connected or built into a database 7 ιι, according to the above embodiment, the capital 711 has different images and objects. Information, recording a plurality of types of image information, including color (Hue) features in the image object through the chromatic space conversion k. In addition, the Dfl unit 713' can connect to the remote worker by wired or wireless communication means, and the 7G' control room 7 () of the equipment room can remotely control the use of the image color information = the unreasonable system 'and receive the transmission by the system The message, such as applied to the smart, video surveillance system, through the monitoring of the control room 7 (), can immediately receive the message transmitted by the local security system. 4 The image processing module 715 obtains the image resource system from the upper memory list, and the main image through the image processing module 715; the color space conversion unit in the image processing module 715 = 71 the fixed color space conversion, in the preferred embodiment. Converting the chromaticity space in which the continuous image of the image is originally converted into a color-saturation-luminance chromaticity space. Name ^ The color characteristic of the Na unit 72 after the conversion of the chromaticity space, color (Hue) special ( (for specific objects), and exclude the saturation of the 2011 20111717 degrees and brightness. And according to its heart ==: No i: Out Next, the eighth to tenth_pictures are shown as = out of interest. Example, but the actual implementation is not limited to this. Ben Mao’s various
t第八圖所示之遺失監控實施例。 攝影機80固定監控特定空間每 =定空間外，更可定時轉動動態掃二^_ 續影像擷取比對影像802，經比對表考t *801 ’並由後 的色彩_資訊，_斷刚像 遺失的事件。此可能觸發保全系統。4的工間中發生 第九圖顯示遺留物監控之實施例。 攝影機90同樣針對一空間92 方向，由初始環境設定參考影像9〇1，，的 比對影像902 ’經擷取色彩資訊、比封 二'貝衫像得出 不明的物品。此同樣可觸發保全*彳’此夠得出有新增 同理，如第十關示為_=_ 中攝影機10先產生原本環境中的泉私動之貫施例’其 像102中不同的影像，可以發現 像1〇1，再比對影 可能觸發保m作進-步處理。〜機被移料事件，此 比對上述參考影像與連續變孰 連續改變的影像，整合連續變動全比對影像，可以得到 、息面，可以描繪出—移[ 201120817 動軌=’如第十一圖所示，可透過 11全程監控，並判斷出有連續__，如 移動物112與移動物Π2， 。 ^不的t Loss monitoring embodiment shown in the eighth figure. The camera 80 fixedly monitors a specific space every time outside the fixed space, and can also rotate the dynamic sweeping time _ continuation image capture comparison image 802, after comparing the table test t * 801 ' and after the color _ information, _ broken just Like a lost event. This may trigger a security system. The ninth diagram of the work in 4 shows an example of the monitoring of the remnant. The camera 90 is also directed to a space 92 direction, and the reference image 902 ′ of the initial environment setting reference image 902 ′ is extracted from the color information, and the unidentified item is obtained by comparing the image of the second jersey. This can also trigger the preservation *彳' This is enough to add the same reason, for example, the tenth is shown as _=_. The camera 10 first generates the original embodiment of the spring in the original environment. The image can be found like 1〇1, and then the contrast can trigger the m-step processing. ~ The machine is moved to the event, which compares the above-mentioned reference image with the continuously changing image, and integrates the continuous change of the full-aligned image, which can be obtained, and can be depicted as a shift [201120817 moving track = 'such as the tenth As shown in the figure, it can be monitored through 11 and it is judged that there is continuous __, such as moving object 112 and moving object ,2. ^No
綜上所述，本發明訊使用影像色彩資訊之處理 處理方法與影像分類找，主要是透過將娜之影像L =轉換，取其中不被環境光影、亮暗影像判斷丁 的色办（Hue)纽，之後經過匹配運算判斷匹配程声 判斷出異常事件，其中具有較高的判斷率。 惟以上所述僅為本發明之較佳可行實施例，非因 侷限本發明之專職ϋ，故舉凡運縣發明說明書及圖八 内容所為之等效結構變化，均同理包含於本發明之銘= 内，合予陳明。 【圖式簡單說明】 第-圖係為習知技術透過色彩圖表進行物件特徵护 的方法； Τ 第二圖係為本發明使用影像色彩資訊之處理系統之應 用實施例示意圖； μ 第三圖係為本發明使用影像色彩資訊之處理方法中計 算匹配的步驟； ° 第四圖係為本發明所利用之色彩_飽和度_亮度之色度 空間示意圖； & & 第五圖係為本發明使用影像色彩資訊之處理方法實施 例流程圖； ' 第六圖係為本發明使用影像色彩資訊之處理方法的物 件分類實施例流程圖； 18 201120817 第七圖係為本發明使用影像色彩資訊之處理系統實施 例示意圖； 第八圖係為應用本發明之實施例之一； • 第九圖係為應用本發明之實施例之二； 第十圖係為應用本發明之實施例之三； 第十一圖係為應用本發明之實施例之四。In summary, the present invention uses image color information processing methods and image classification to find, mainly through the conversion of Na image L =, to take the color light (Hue) which is not judged by ambient light and dark image New, after the matching operation judges the matching process sound to determine the abnormal event, which has a higher judgment rate. However, the above description is only a preferred embodiment of the present invention, and is not limited to the full-time operation of the present invention. Therefore, the equivalent structural changes of the invention description and the contents of FIG. 8 are included in the description of the present invention. = Within, combined with Chen Ming. [Simplified description of the drawings] The first figure is a method for the object feature protection by the prior art through the color chart; Τ The second figure is a schematic diagram of an application embodiment of the processing system using the image color information of the present invention; The step of calculating the matching in the processing method of using the image color information of the present invention; ° The fourth figure is a chromaticity space diagram of the color_saturation_brightness utilized by the present invention; && A flow chart of an embodiment of a method for processing image color information; 'The sixth figure is a flow chart of an object classification embodiment using a method for processing image color information according to the present invention; 18 201120817 The seventh figure is a process for processing image color information according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS The eighth embodiment is one of the embodiments of the present invention; the ninth embodiment is the second embodiment of the present invention; the tenth embodiment is the third embodiment of the present invention; One figure is the fourth embodiment of the application of the present invention.
【主要元件符號說明】 建築物21 無線基地台25 取像單元701 暫存記憶體705 儲存單元709 通訊單元713 影像處理模組715 色彩特徵擷取單元72 攝影機 80, 90, 10, 11 空間82,92， 比對影像802, 902, 102 攝影機23 監控中心27 類比數位轉換單元7.03 處理單元707 資料庫711 控制機房70 色度空間轉換單元71 匹配運算單元73 參考影像801, 901，101 移動物112, 112’ 19[Main component symbol description] Building 21 Wireless base station 25 Image capturing unit 701 Temporary memory 705 Storage unit 709 Communication unit 713 Image processing module 715 Color feature capturing unit 72 Camera 80, 90, 10, 11 Space 82, 92, comparison image 802, 902, 102 camera 23 monitoring center 27 analog digital conversion unit 7.03 processing unit 707 database 711 control room 70 chromaticity space conversion unit 71 matching operation unit 73 reference image 801, 901, 101 mobile object 112, 112' 19