M419956 五、新型說明： 【新型所屬之技術領域】 別是=1=種具智慧型辨識功能之超廣角攝影機，特 廣二峨正、晝面分割與影像辨識模組的超 【先前技術】 製的錄影包括有攝影機與後端影像擷取與錄 個角度的影像，：取攝—個特定位置或是某 時監控:或是：::=用監視主機’_即 由方;&現錄影設備中的攝影機拍攝角 的地這也形成不法之:可__ 的攝影機，術應用了能夠拍攝較廣角度 而利用較廣角度攝影機拍攝 廣，失真的情況會愈嚴重，特別是在著角度欲 【新型内容】 本創===== 範圍的好處，將二二廣角拍攝較廣 定義-個影像辨識的功能。£域各個區域可以各自 根據本揭露書所描述的具智慧型辨識功能之超廣角攝 4/19 M419956 衫機貫施例之一，此設置於一空” 拍攝超廣肖雜之超廣肖鏡⑽H 攝f彡機包括有 生影像訊號之影像擷取模組、執〜与收峻廣角影像並產 據處理單元、能夠執行影像校正、二^輕理之影像數 與行為辨識單元，裝置更包括有斬二副與執行辨識的物件 處理之影像訊號的記憶單元日子影像數據處理單元所 元。 提供通訊連線之通訊單 具智棻刑辨细. ，冰心逋訊單 具令,¾型辨識功能之和度么 ::ί生的影像，的功能提供遠蠕存取 喊。 透過網路存取裝置内的影像 根據另—實施例， ==立的“同時可辨識功能之超廣 供-二同樣具有影像處理 =生:;同時提供終端通以，機所產生的影像 所產=晝面。 ㈣㈣覽多個超廣角攝影機 攝影機中貝、施例之一，具智#刑 理、物件遍::Γ種處理晶片，内;：識功能之超廣角 午與仃為辨識單亓 円。卩執行上述影像备 辨識等的程序。早…執行的影像校正、晝面象= f實施方式J :ξ:·要是3能之超廣角攝影機 超廣角的拍攝功能可以降的超廣角攝影機 -人滅控的場所因為拍攝 •5/J 9 技彳二全事件的問題’其中提出的影像處理 超廣角攝影機有=3=^1值得-提的是，由於 可分割為多個區域 )，各」固區域可以各自定義一個影像二功能在個別校正 廣角攝影二意:為具有具智慧型辨識功能之超 —逬2=:==::: 達成這 攝影機H)直接經由網路14傳送出谷由超廣角 圖中所示透過網路連接外部通轉置建 像處理系統可應用在一個監控系統中，利 ilii較無死角的監控環境，特別應用-種:備㈣二 頭的攝純’或是其他可拍驗輕圍或是全 到較遼闊的監視範圍，超廣角影像:」接 近囡形的域，但影像會產生像差 、接 域，需要透過校正的技術來取得正常的=二以夕的區 形的影像經校正步驟後，會形成—方形影像。b +接近圓 圖愤廣㈣雜1G射執行影像處理的軟體或带路 ㈣影像處理晶片，可自身執行影像_取、“ _序:：^==^=_像處 硬體手段進行縣紅、分顺傳送錢1G㈣的軟體或 在另-實施例中，如圖2A所示之本創作具智慧型辨識 6/19 功能之超廣角攝影機之系統實施例示意圖，其中提供一額 外主機28負責儲存接收自多個超廣角攝影機^ η幻傳 =像，亦可再經由網路2Q傳送出去。此例中:超廣角攝 =機21，22, 23或疋主機28皆可提供網頁内容或特定 網頁或特定軟體觀觀看所_的即時影像或 疋歷史影像。 此例所不㈣部超翻攝影機21，22，M419956 V. New description: [New technology field] =1==Super wide-angle camera with intelligent identification function, super-pre-existing system The video includes images of the camera and the back-end image captured and recorded at an angle: taking a specific location or monitoring at a certain time: or:::= using the monitoring host '_ is the party; & The camera shooting angle in the device is also illegal: the camera can be used to shoot a wide angle and use a wide angle camera to shoot wide, the distortion will be more serious, especially in the angle [New content] The benefits of the range ===== range, the 22-wide wide-angle shooting is widely defined - an image recognition function. Each area of the £ domain can be one of the super wide-angle 4/19 M419956 shirts with intelligent identification function described in this disclosure. This is set in an empty space. The camera includes a video capture module with a raw image signal, a digital camera with a wide-angle image and a processing unit, an image correction and a motion recognition unit capable of performing image correction, and the device includes斩Second pair of memory unit day image data processing unit with image signal processing for performing object recognition. Communication of communication connection is provided with wisdom and punishment, and Bing Xinxun single order, 3⁄4 type identification function And degree:: ί 的 image, the function provides far-reaching access shouting. Accessing the image in the device through the network according to another embodiment, == standing "simultaneously identifiable function super wide supply - two same With image processing = raw:; at the same time provide the terminal through, the image produced by the machine produced = face. (4) (4) Viewing a number of super wide-angle cameras One of the cameras, one of the examples, with wisdom #刑,物物::ΓProcessing wafers, inside;: Super wide-angle of recognition function Noon and 仃 for identification single 亓.卩 Execute the above program such as image preparation. Early...Improved image correction, 昼面== f Implementation J: ξ:·If the 3 can't super wide-angle camera super wide-angle shooting function can be lowered by the super wide-angle camera - the place where people are controlled because of shooting • 5/J 9 The problem of the second full event of the technology's proposed image processing super wide-angle camera has =3=^1 worthwhile - because it can be divided into multiple regions), each "solid region" can define an image two functions in individual Correction of wide-angle photography: For the ultra-逬2=:==::: with intelligent identification function to achieve this camera H) directly through the network 14 to send out the valley through the network as shown in the super wide-angle diagram The transposed image processing system can be applied to a monitoring system. Lilii has no dead angle monitoring environment, and special application-species: preparation (four) two-headed pure 'or other can be tested light or full to the wider Surveillance range, super wide-angle image: "Close to the shape of the dome, but the image will produce aberrations, the domain, the need to use the correction technique to obtain the normal = two-night image of the region after the correction step, will form - Square image. b + close to the circle picture anger wide (four) miscellaneous 1G shot to perform image processing software or lead (4) image processing chip, you can perform image _ take, " _ sequence:: ^ = = ^ = _ like hardware means to carry out county red, The software of the 1G (4) software is distributed separately or in another embodiment, as shown in FIG. 2A, a schematic diagram of a system embodiment of the ultra-wide-angle camera with intelligent identification of the 6/19 function is provided, wherein an additional host 28 is provided for storing and receiving. From a number of super wide-angle cameras ^ 幻 传 = = image, can also be transmitted via the network 2Q. In this example: super wide-angle camera = 21, 22, 23 or 疋 host 28 can provide web content or a specific web page or Viewing the real-time image or the historical image of the specific software view. In this example, the (four) super-turning camera 21, 22,
统，各個超廣角攝影機自身即具有影 像處理的功I，内稍組可參考圖4的描^中可 =正經超廣角鏡頭拍攝的影像，此類影像通常i有像差， 接近鏡片周圍的部份；經校正的影像可二先行儲存 於攝影機_記憶體内，再根據需要執行其他程序，如全 面分割、影像辨識與特定事件的處理。 旦 超廣角攝影機21，22,23各自更具有網路 無f或是树手段频_1連接In the system, each super wide-angle camera itself has the function of image processing, and the inner group can refer to the image captured by the super-wide angle lens of FIG. 4, which usually has aberrations and is close to the surrounding part of the lens. The corrected image can be stored in the camera_memory first, and then other programs such as full segmentation, image recognition and specific event processing can be performed as needed. Once the ultra wide-angle cameras 21, 22, 23 each have a network, no f or tree means frequency_1 connection
j置（如圖中手持式裝置27、終端電腦25)可以直接遠 後的靜態或動態影像，，）直接取得經影像處理 在此實關巾，除了各自超廣㈣滅具有自身影像 j、^存與通訊功能外，多部超廣角攝影機4叹23所 中像訊號可由另—外部主機28健存，經整理可由圖 =庫29所儲存’以便日後調閱，並可透過網路20提 仏^而存取的服務，遠端通訊裝置可以藉此㈣劉覽 超廣角攝影機之影像資料。 。 ^慧型辨識功能之超廣角攝影機所建立的系統可繼 、，·貝參考圖2Β所描繪之—實施例示意圖。 、’ 7/19 此實施例以連接網路20 式終端通訊裝置透過網路2〇递=角攝影機21為例，各 的手持式裝置27與終端電腦超廣角攝影機21，如圖示 腦25分別經由網㈣存 ^寺^置?與終端電 。 、增角攝影機21所處理的影像 超廣角攝影機21自身可# 例令，超廣賴影機21 的超廣角影像，此 電路實現的校正模組231、2ί=體（包姉體）或是 、記憶模._ '伺服器模組;35:2二 二2;; 可包= 述的全部模組或是其二貫施例 ，經景取ίϊ 取仔自身所拍攝的靜態或動態影像 組式的超廣角影像訊號，經校正模 切割為多個“面、::;二:组232將影像晝面 = 影像辨識程序― 像管 王序中的δΚ说’並可儲存最終獲得的影 施例二’攝：，行各辨識步驟之程式碼。在進-步的實 ，以#料冰"幾可包括伺服器模組235與通訊模組236 更對外通訊並提供存取服務。 j之各種影像處理程序之相關可能實施的方式描述 景;像處理於以下所述知方式)，州 衷""方法可以應用已知的習知技術，比如，校 旦231將用於校正超廣角影像訊號之失真像差，其中 :处理方法可參考習知技術，如美國專利第7489335號 =如^國專利公開第2010/0201833號所揭示的影像校正方 /’其中利用矩陣等數學方式轉換影像中各項素，將影像 8/19 校正為正常無失直旦 面4==二分割超廣角影像訊號為多個子晝 ，辨識處_組施例1影像分割為四個子畫面 別執行〜或夕、’可根據需要設定多個子畫面訊號分 理時，可利二己情二步驟’而其中影像訊號在執行影像處 處埋之影像訊號以、，且234作為儲存經校正、分割與辨識 場影像，由於超廣角影像將涵蓋較廣的現 如接近門口的地方！Γ竟空間可能有不同的監控需求，比 而定，為;甚至針對特 而需要進行物件遺失的_ =疋位置可能擺設物品’ 口 m車輛車__=如拍攝停車場、車輛出入 的功能1型=識功能之超廣角攝影機提供外部存取号，像 “像==义，即：= 飼服器模組235提供一網^:衫像。其令攝影機2】内的 飼服器（WW serve小用’實務上可為一網頁 2丨内經校正、分割與辨識處理取；廣角攝影機 21更可透過通訊模組236傳 〜3廣角攝影機 影像訊號。通訊方式並不=處理之 路、有線網路，或是行動通 式’可包括無線網 使用終端通訊裝置透過通訊用者可以 過其中词服器模組235建立之廣角攝影機2】，透 與辨識處理之影像訊號，包括“、=觀看經校正、分割 歷史或即—端==像'動應的 9/19 M419956 圖3接著顯示為超廣角影像之處理程序之實施例示意 圖。 圖中頦示一超廣角鏡頭31，可為上述魚眼鏡頭、全景 拍攝鏡頭，或特定可拍攝超廣角影像的鏡頭，藉此取得一 超廣角影像33，實施例可為接近圓形的影像。超廣角影像 33接著透過分割處理，經校正後產生分割畫面％，其中'包 括有多個子晝面’不以圖f數目為限。 由於超廣角影像33可能涵蓋不同多個不同需求的區域 ，因此可以利用軟體、韌體或硬體（晶片）手段分割為處 理不同事件處理步驟的子晝面。特別的是，這些手段之實 施例可於本麟之超廣肖攝影機喊成，並料外部處二 〇 此例中的分割畫面3 5同時可拍攝四種態樣的環境，如 具有物件301的子畫面’此處可以執行物件遺失的辨識流 程’透過此子畫面中各像素與穩定不便的背景影像間的差 異’如連續減同-定點晝面，每張影像射景影像之像 素值相減（可先灰階化），藉此可以識別出物件3〇1是否有 變動，藉此達成物件遺失監控的目的。此類技術可支考美 國專利公開第20_174425賴揭露_件遺失偵測方法 ’其先定義好物件位置，與各物件之間（與環境）的關聯 性丄再價减否位置與關聯性改變，若超出—定範圍則判 斷遺失。 ' 另有—子畫面用於辨識車牌，如車牌搬，可能是拍攝 -個空間的車輛出人D，車牌辨識可根據車牌的特徵，如 ，像中具有-塊白色或特定顏色的長方形，其中記裁有文 予’可以據此騎為車牌。此類技術如美國專利公開第 10/19 M419956 2008/0212837號所揭露的車_識技術，其巾 =接著經數位化為影像，再辨識其中字母。另如：國 第仏131唬所揭露的智慧型照相 ;位=出特定區域，偵測角度、位置、亮度等= 素，即利用影像比對辨識出其中字母。 兄口 另有子晝面顯示有人303在出入口，此可為 的用途’即攝影機持續拍攝—個出入口 景影像的轉，彳⑼透郝攝影像與背 P产，寺疋對應位置的改變，並判斷改變的 再:2一門禮，可以判斷為入侵行為。 連續影像軌跡304的備測’也就是透過 樣也是透過前後連續旦置:二㈣繪出移動的執跡，同 間的變化。、’’、〜像的受化，匈斷物件位置與隨著時 所能識功能之超廣角攝影機 判斷二:6!態樣，像的變化來 例功能方棟圖揭露#所提出的超廣角攝影機之實施 機二仏=;為:攝超廣角影像的超廣角攝影 。超廣角攝影機具備間之現場影像 可擷取超廣角影像的鏡頭，、可包括;ϋ’如魚眼鏡頭等 是連續多巾貞之超廣肖影 _超廣肖影像訊號或 影像擷取模組402 接:卢態影像;裝置更包括 .逆稞超廣角鏡頭模組401，可為一 11/19 M419956 重成像”產生超廣角影像讯號的光感測器 頭模組401所拍攝之影像，甘连吐公推㈣《廣角鏡 W…一 產景_號，實施例可包 括有CCD、CMOS等影像感測器。 角摄取模組402產生超廣角影像訊號後，由超廣 角^機40内影像數據處理單元4δ接收，此影像2 理早tl48電性連接影像擷取模組4〇2， ^ 中央處理單元，用於接收影像訊=仃衫像處理的 差與分割影像訊號為多個子畫面訊號，二像 電路單元的訊號往來與溝通。 貝貝▲置内各 根據實施例之一，影像數據處理單元 角影像訊號.像差之校正模組421，此可為;^正， 〕或是硬體實_池，特則正Τ =括章刃體 攝的影像，成為-般視覺可辨識的影像鏡頭所拍 影像數據處理單元48包括分割 面之分割模組仍，畫面分割係依據實際畫 同的子畫面，而各個子晝面㈣區分不 ㈣，各觸倾可參 相關内容的描述。 g〒圖5與 超廣角攝影機40包括一物件與 連接影像數據處理單元48，物件與 電性 能是設定上述多個子晝面訊號之個類42主要功 影像數據處理單元48接著針對各種辨=亚能配合 步驟，如以下圖5之描述，各個子書面特定辨識 個辨識步驟。 —對應執行一或多 值得-提的是，本創作提出的超廣角攝 行的各種影像處理程序，包括各個子 二機40内部執 ~ 個別辨識步驟 ΙΟ/ΙΛ ^像;k正、畫面分料影像處理，可由—處理晶c J貫現。 據處角，機40包括—記憶單元44’電性連接影像數 句❹ 記憶單S 44 4影像數據處理單元48執行 於μ·Γ^暫存記憶體’也就是處理中的影像訊號可暫存 。：。己中’用於校正、影像比對等步驟的暫存區 據處單元™於影像數 48 ’此可為非揮發性記憶體（_-V〇latile 妹=形式’用於儲存經拍攝之影像訊號，亦可儲存 ;=二，事件處理後的影像訊號，提供外部褒置存 所干m广例中’如上述圖2A所述之實施例，本圖4 另攝影機4。同樣可將經處理的影像訊號傳遞於 接影=:1==6，通訊單元”性連 遞出去可透過網路5。傳 個終端通訊裝置連線，外式與外部一或多 單元奶存取裝置内的影像資料。、域過網路手段經此通訊 根據再—實施例，超廣 裝置存取其㈣存的影像資料，外部通訊 像數據處理單元4 _ ]用圖中電性連接影 廣角攝影機-内影早=供利用網頁存取超 記載執行-纟顯伺 45特別可具有 此飼服器單元45接^式（如胃W s^er)之記憶體。 ^各種終料减置縣，轉端通^ 13/19 M419956 裝置應具有㈣路5〇㈣透過此师舰程·送的影 畫面的通訊模組。 豕 根據上述本揭露書所描述應用超廣角攝影機之運作， 其中主要是根據所取得的超廣角影像訊號進行校正處理、 影像分割與特定辨識流程，以下圖5與相關描述即揭 用系統執行的各種事件處理程序。 如圖5所描述其中超廣角影像處理方法實施例流程。 ==如步驟S5〇1 ’透過設置於特定環境 胸像可為單張靜態影像， 或疋，張4或連續的影像，經影像感測後產生影像· ，如^驟挪，影像可經如CCD、CM〇s等影像感測器^ 生汛唬，取彳于的影像可暫存至記憶單元。 …超廣角影像易於周圍部份產生嚴重的像差，因此 ，校正步驟，如步驟S5Q5，經影像失真校正後，可二 得較佳可分_f彡像，，正咐彡像峨 : :視=::，述_學方法進行轉換 σ刀別儲存至s己憶單元中，各個子畫面將設 化 識功能，如步驟S509。也就是，超廣角影像可能涵=辨 冋辨識需求的區域’子畫面的切割可以依據實際 刀副為一或多個子畫面。每個子晝面所涵蓋的影 仃 進行不同的影像處理，根據需求產生結果。值得字 j述步驟S505的校正步驟可在切割子晝面的步驟 之前或是之後’依據攝影機的設計而定。. 14/19 M419956 再如步驟S511，各個子晝面將可分別執行相同（可能 有相同辨識功能，但物件不同的步驟）或不同的辨識步驟 - ，辨識步驟可參考以下實施例描述。 ' 物件遺失辨識： 開始步驟如圖5所示之影像處理步驟，包括拍攝超廣 角影像、產生影像訊號、失真校正與分割多個子畫面等。 校正與分割步驟可交換。 物件遺失辨識步驟可由子晝面之一執行，流程先設定 • 一參考背景影像，也就是影像固定的背景晝面，暫存至記 憶單元，以此為基礎，並可先行設定一物件位置，同樣作 為參考背景。之後連續擷取同一分割子畫面之影像，連續 影像先暫存至記憶單元，每張影像與參考背景影像比對， 比對前後子晝面影像，根據連續多個子晝面影像之差異得 出其中之一物件之變動狀況，當差異表示所設定物件被移 動時，或是大於一門檻時，則可判斷物件之變動狀，比如 可藉此判斷該物件是否遺失。 • 車牌辨識： 主要程序是先設定其中之一子晝面為車牌辨識功能， 此功能包括擷取子晝面影像，並暫存至記憶單元，此時， 可先設定車牌可能位置，或是透過影像處理，根據車牌的 特徵（如白底黑字、黃底紅字）進行辨識，經辨識出車牌 位置後，再辨識其中字母。 物件追蹤辨識： 流程先設定一子畫面為物件追蹤功能，物件追蹤的步 驟接著設定參考背景影像，將參考背景影像暫存至記憶單 元，再連續擷取同一子晝面之影像，同樣暫存至記憶單元 15/19 元中儲存的連續影像將與參考 透過衫像間的差異，由連續^像比對，可 緣出物件移動執跡。4出一移動物件’再描 入侵辨識： V驟同樣先設定—子畫面為入侵判斷…一 苓$景影像’暫存至記憶單元，接著連又疋— t影像’同樣暫存至記憶單元，將記憶;二:晝 與參考背景影像比對，可在特定位置二: -入侵行為。化寸侧出一人形物件，即可判斷有 角攝=所超廣 依此提 同子畫面與對應，識功能" 且無死角的各種事件處瑝能力。 •二:創作之較佳可行實施例’非因此即 内容所為=:;化=:創:說明書及圖示 ，合予陳明。 又化均问理包含於本創作之範圍内 【圖式簡單說明】 圖〗顯 意圖； 不為具#柄叙超—彡叙祕設置示 圖2Α > f ; 圖 , |_ * 廣角攝影機之系統實施例 示意H所7^為雜本創作超廣鱗影機之系統實施例 M419956 圖3顯不為超廣角影像之處理程序之貫施例不意圖， 圖4所示為本創作超廣角攝影機之實施例功能方塊圖 圖5描述為利用本創作超廣角攝影機之影像處理方法 實施例流程。j (as shown in the hand-held device 27, the terminal computer 25) can be directly behind the static or dynamic image,) directly obtained through the image processing in this real off towel, in addition to their respective super wide (four) off with their own image j, ^ In addition to the communication and communication functions, many super wide-angle cameras 4 sighs 23 of the image signals can be stored by another external host 28, and can be stored by the map = library 29 for later reference, and can be extracted through the network 20 ^ Access to the service, the remote communication device can use (4) Liu Yingchao wide-angle camera image data. . The system established by the ultra-wide-angle camera of the HC-type identification function can be followed by a schematic diagram of the embodiment as described with reference to FIG. , '7/19 This embodiment uses a connection network 20 type terminal communication device through the network 2 〇 = angle camera 21 as an example, each of the handheld device 27 and the terminal computer super wide-angle camera 21, as shown in the figure brain 25 Through the network (four) save ^ temple ^ set? and terminal power. The image super wide-angle camera 21 processed by the angle-increasing camera 21 can be used as an ultra-wide-angle image of the ultra-wide-angle camera 21, and the correction module 231, 2ί=body (package body) or Memory mode._ 'Server module; 35:2 22 2;; can be packaged = all modules described or their second application, through the scenes ϊ 取 take the static or dynamic image group shot The ultra-wide-angle image signal is cut into a plurality of "faces,::; two: group 232 to image the image = image recognition program - δ Κ 像 ' 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可 并可Photograph: The code of each identification step. In the step-by-step, the #料冰" can include the server module 235 and the communication module 236 to communicate externally and provide access services. The manner in which the processor is likely to be implemented describes the scene; as described in the following), the state-of-the-art method can apply known conventional techniques, for example, the school 231 will be used to correct the super wide-angle image signal. Distortion aberration, wherein: the processing method can refer to conventional techniques, such as the United States U.S. Patent No. 7,489,335 = Image Correction Method disclosed in Japanese Patent Publication No. 2010/0201833, in which the elements in the image are converted by mathematical means such as a matrix, and the image 8/19 is corrected to a normal no-loss straight surface 4 ==Two-division super wide-angle image signal is multiple sub-frames, and the identification part_group 1 image is divided into four sub-pictures to execute ~ or eve, 'you can set multiple sub-picture signals according to your needs. The second step 'where the image signal is buried in the image where the image signal is buried, and 234 is used to store the corrected, segmented and recognized field image, since the super wide-angle image will cover a wider area that is close to the doorway! There may be different monitoring requirements, depending on the specifics; even for the special need to carry out the loss of the object _ = 疋 position may be placed items 'mouth m vehicle __= such as shooting parking, vehicle access function type 1 = identification function The super wide-angle camera provides an external access number, like "like == meaning, ie: = feeding device module 235 provides a net ^: shirt image. It enables the feeding device in the camera 2 (WW serve small use 'actually can be corrected, segmented and identified for a web page 2; wide-angle camera 21 can also transmit through the communication module 236 ~ 3 wide-angle camera image Signal. Communication method is not = processing road, wired network, or mobile-type 'can include wireless network using terminal communication device through communication users can pass through the wide-angle camera 2 built by the word processor module 235] And the image signal of the identification processing, including ", = viewing corrected, split history or instant - end == image response 9/19 M419956 Figure 3 is then shown as a schematic diagram of an embodiment of the processing procedure of the super wide-angle image. An ultra-wide-angle lens 31 can be used for the above-mentioned fisheye lens, panoramic shooting lens, or a lens that can capture ultra-wide-angle images, thereby obtaining an ultra-wide-angle image 33. The embodiment can be a near-circular image. 33 then through the segmentation process, after correction, the segmentation screen % is generated, wherein 'including multiple sub-planes' is not limited to the number of pictures f. Since the super wide-angle image 33 may cover different multiples The same area of demand, so it can be segmented into sub-surfaces for processing different event processing steps by means of software, firmware or hardware (wafer). In particular, embodiments of these means can be shouted at the ultra-wide camera of Ben Lin. In the case of the external screen, the split screen 3 in this example can simultaneously capture four kinds of scenes, such as the sub-picture with the object 301 'where the identification process of missing objects can be performed' through each of the sub-pictures The difference between the pixel and the background image with stable inconvenience 'such as continuous subtraction - fixed point surface, the pixel value of each image of the image is subtracted (can be grayed out first), thereby identifying whether the object 3〇1 has Change, thereby achieving the purpose of object loss monitoring. Such technology can be tested in US Patent Publication No. 20_174425, which discloses the method of missing detection, which first defines the position of the object and the relationship between the objects (with the environment).丄Re-price reduction and no change in position and relevance. If it exceeds the range, it will be judged to be lost. 'Others----------------------------------------------------------------------------------------------------- D, the license plate recognition may be based on the characteristics of the license plate, for example, a rectangle having a white color or a specific color in the image, wherein the record has a text of 'can be used as a license plate. Such a technique is disclosed in US Patent Publication No. 10/19 M419956. The car_sense technology disclosed in 2008/0212837, the towel = is then digitized into an image, and then the letters are recognized. Another example: the smart camera disclosed by the country 仏131唬; bit = out of a specific area, detection Angle, position, brightness, etc. = the use of image comparison to identify the letters. The other side of the brother's mouth shows that someone 303 is at the entrance and exit, this can be used for the purpose of 'camera continuous shooting' - a turn of the entrance and exit scene image,彳 (9) through the Hao photography image and back P production, the temple position corresponding to the change, and judge the change again: 2 a ceremony, can be judged as an intrusion. The preparation of the continuous image track 304 is also transmitted through the sample before and after the continuous setting: two (four) to depict the movement of the detour, the same change. , '', ~ image of the subject, the position of the Hungarian object and the ultra wide-angle camera that can be used as the function of the time to judge the second: 6! Aspects, like the change of the example function square diagram reveals # proposed super wide angle The implementation of the camera is two-dimensional =; for: super wide-angle photography of super wide-angle images. The ultra wide-angle camera has a lens that can capture super wide-angle images, and can include; ϋ 'such as a fisheye lens, etc. is a super-wide shawl _ super wide Vision image signal or image capture module 402 Connected: Lu state image; device includes: reverse 稞 super wide-angle lens module 401, can be an 11/19 M419956 re-imaging image of the light sensor head module 401 that produces super wide-angle image signal, Ganlian吐公推 (4) "wide-angle lens W... a production scene _ number, the embodiment may include a CCD, CMOS, etc. image sensor. After the angle ingestion module 402 generates a super wide-angle image signal, the image data processing unit in the super wide-angle machine 40 4δ receiving, this image 2 is early tl48 electrically connected to the image capturing module 4〇2, ^ central processing unit for receiving video signal = difference between the image processing and split image signal for multiple sub-picture signals, two images Signal communication and communication of the circuit unit. According to one of the embodiments, the image data processing unit angle image signal. The aberration correction module 421, which can be; ^正, 〕 or hardware _ Pool, special is Τ = bracketed by the blade The image data processing unit 48 that has become the image-recognizable image lens includes the segmentation module of the segmentation surface. The screen segmentation system is based on the actual sub-picture, and the sub-surfaces (4) are not differentiated (four), each touch The description of the related content is shown in Fig. 5. The super wide-angle camera 40 includes an object and connection image data processing unit 48, and the object and electrical properties are the main types of the plurality of sub-surface signals. 48 Next, for various identification and sub-energy cooperation steps, as described in FIG. 5 below, each sub-writing specifically recognizes an identification step. - Corresponding execution of one or more worthwhile - the various images of the super wide-angle camera proposed by the present creation The processing program includes the internal processing of each sub-machine 40, the individual identification step ΙΟ/ΙΛ ^ image, the k-positive, the screen-dividing image processing, and the processing can be performed by the processing unit. According to the corner, the machine 40 includes a memory unit 44. 'Electrically connected image number ❹ Memory list S 44 4 Image data processing unit 48 is executed in μ·Γ^ scratch memory ', that is, the image signal being processed can be temporarily stored. 'Scratch area data unit for correction, image comparison, etc. in image number 48' This can be non-volatile memory (_-V〇latile sister = form' is used to store the captured image signal, It can also be stored; = 2, the image signal after the event processing is provided in the external storage device. In the example of the above, the embodiment shown in Fig. 2A is shown in Fig. 4, and the camera 4 is similar. The processed image can also be processed. The signal is transmitted to the connection =:1==6, and the communication unit can be transmitted through the network 5. The terminal communication device is connected, and the external and external one or more unit milk access device image data. The domain over-network means via this communication according to the re-implementation, the ultra-wide device accesses its (four) stored image data, and the external communication image processing unit 4 _ ] uses the picture to electrically connect the wide-angle camera - the inner shadow early = For the use of web page access, the ultra-documentation execution - the display device 45 can have the memory of the device unit 45 (such as the stomach W s ^ er). ^All kinds of final material reduction county, the end of the pass ^ 13/19 M419956 device should have (four) road 5 〇 (four) through the division of the ship to send the shadow of the communication module.应用 The operation of the super wide-angle camera is applied according to the above disclosure, which mainly performs correction processing, image segmentation and specific identification process according to the obtained super wide-angle image signal. The following FIG. 5 and related descriptions disclose various executions of the system. Event handler. The flow of the embodiment of the super wide-angle image processing method is as described in FIG. ==If step S5〇1' can be a single still image, or 疋, 4 or a continuous image, the image can be generated after image sensing, such as ^ 骤, the image can pass through, for example, CCD Image sensors such as CM〇s are produced, and images captured by them can be temporarily stored in the memory unit. ...the super wide-angle image is prone to severe aberrations in the surrounding part. Therefore, the correction step, such as step S5Q5, can be better divided into _f images after correcting the image distortion, and the image is: =::, the _ learning method performs the conversion σ knife and stores it in the suffix unit, and each sub-picture will set the characterization function, as in step S509. That is, the super wide-angle image may contain the area where the identification needs are determined. The sub-picture can be cut according to the actual knife pair as one or more sub-pictures. The images covered by each sub-surface are processed differently to produce results based on demand. It is worthwhile to say that the correction step of step S505 can be performed before or after the step of cutting the face, depending on the design of the camera. 14/19 M419956 In step S511, each sub-plane will be able to perform the same (may have the same identification function, but different steps) or different identification steps - the identification steps can be described with reference to the following embodiments. ' Object Loss Identification: The initial steps are the image processing steps shown in Figure 5, including shooting super wide-angle images, generating image signals, distortion correction, and splitting multiple sub-pictures. The correction and segmentation steps can be exchanged. The object missing identification step can be performed by one of the sub-surfaces, and the flow is first set. • A reference background image, that is, a fixed background of the image, temporarily stored in the memory unit, based on which the object position can be set first, also As a reference background. After that, the images of the same divided sub-picture are continuously captured, and the continuous images are temporarily stored in the memory unit, and each image is compared with the reference background image, and the front and rear sub-surface images are compared according to the difference of consecutive sub-images. The change of one of the objects, when the difference indicates that the set object is moved, or is greater than a threshold, the change of the object can be judged, for example, whether the object is lost. • License plate recognition: The main program is to set one of the sub-surfaces as the license plate recognition function. This function includes capturing the sub-surface image and temporarily storing it to the memory unit. In this case, you can first set the possible position of the license plate, or The image processing is identified according to the characteristics of the license plate (such as black on white and red on yellow). After the license plate position is recognized, the letters are recognized. Object tracking identification: The process first sets a sub-picture as the object tracking function, and the object tracking step then sets the reference background image, temporarily stores the reference background image to the memory unit, and continuously captures the image of the same sub-surface, and also temporarily stores to The continuous image stored in the memory unit 15/19 yuan will be compared with the reference through the image of the shirt, and the image will be moved by the continuous image. 4 A moving object 're-invasion identification: V is also set first - the sub-picture is for intrusion judgment... a 景 $ scene image is temporarily stored in the memory unit, and then the 疋-t image' is also temporarily stored in the memory unit. Will remember; two: 昼 and reference background image comparison, can be in a specific position two: - intrusion behavior. A person-shaped object can be judged by the side of the inch, and it can be judged that there is an angle of the camera = the super-wide. According to this, the same sub-picture and corresponding, the function of "features" and no dead ends of the various events. • 2: The preferred and feasible embodiment of the creation is not the result of the content =:; = = creation: instructions and illustrations, combined with Chen Ming. The rationalization of the rationale is included in the scope of this creation [Simple description of the schema] Figure 〖Intention; Not for the #柄叙超-彡秘秘 diagram 2Α >f; Figure, |_ * Wide-angle camera The system embodiment shows that H is a system for the creation of a super-wide scale camera. M419956 is not intended to be a super-wide-angle image processing program. Figure 4 shows the creation of a super wide-angle camera. Embodiment Function Block Diagram FIG. 5 depicts a flow of an embodiment of an image processing method using the present super wide-angle camera.
【主要元件符號說明】 超廣角攝影機10 網路14 超廣角攝影機21,22,23 手持式裝置27 校正模組231 辨識處理模組233 伺服器模組235 超廣角鏡頭31 分割晝面35 車牌302 移動轨跡304 校正模組42】 辨識處理模組425 超廣角鏡頭模組401 影像數據處理單元48 記憶單元44 網路50 資料庫29 網路20 主機28 終端電腦25 分割模組232 記憶模組234 通訊模組236 超廣角影像33 物件301 人303 超廣角攝影機40 分割模組423 伺服器單元45 影像擷取模組402 物件與行為辨識單元42 通訊單元46 儲存單元43 步驟S501〜S511影像處理流程 17/19[Main component symbol description] Ultra wide-angle camera 10 Network 14 Ultra wide-angle camera 21, 22, 23 Handheld device 27 Correction module 231 Identification processing module 233 Server module 235 Ultra wide-angle lens 31 Split face 35 License plate 302 Moving track Trace 304 Correction Module 42] Identification Processing Module 425 Ultra Wide Angle Lens Module 401 Image Data Processing Unit 48 Memory Unit 44 Network 50 Database 29 Network 20 Host 28 Terminal Computer 25 Split Module 232 Memory Module 234 Communication Module 236 Super Wide Angle Image 33 Object 301 Person 303 Ultra Wide Angle Camera 40 Split Module 423 Server Unit 45 Image Capture Module 402 Object and Behavior Identification Unit 42 Communication Unit 46 Storage Unit 43 Steps S501~S511 Image Processing Flow 17/19