1286027 玖、發明說明： 【發明所屬之技術領域】 本發明係針對高速公路、道路、橋樑、收費站與特定 場所的影像式多車道自由車流電子收費執法系統，做為車 輛進出電子收費影像執法機制或純粹影像執法機制使 用，亦可結合VCU/車牌黑名單資料庫，做為贓車查缉及 使用偽VCU的查緝等應用。 【先前技術】 針對傳統的影像執法系統，其為了取得完整的全景影 像，必須分成多部攝影機分別自不同的角度進行取像，但 是此法所取得的多組影像相關性較低，不易相互補正。故 傳統的影像執法系統如果在光學辨識（OCR，Optical Character Recognition)工作失敗，或對於電子收費系統 的應用而言，其後端的搭配性（Matching)工作失敗時，操 作員必須從大量的失敗圖形檔當中找出關聯性，但因不同 攝影機係自不同的角度取像，故比較沒有連續空間的感 覺。 【發明目的】 本發明為解決此一問題提出使用接圖（Image Registration)的技術將同一時間多台攝影機之個別影像 加以整合成環場影像，讓操作員在進行人工比對的時候， 能夠比較有連續空間的感覺，也可以簡化操作員人工比對 1286027 的時間。 【發明内容】 本影像執法系統由取像單元、車牌辨識單元、接圖單 元及車型辨識單元所組成。取像單元包含前照（後照）感應 單元、前照（後照）CCD(Charge-Coupled Device)攝影機、 '衫像摘取卡（Frame Grabber )、DI0(Digital Input/Output) 卡等所組成。 前照（後照）感應單元負責偵測進入前照（後照）拍攝 區的車輛，並將感應訊息藉由DI0卡傳送給影像擷取卡以 $發前照（後照）C C D攝影機拍取多組單張前照（後照）車 輛影像，以便利用車牌辨識單元擷取車輛的車牌號碼。並 利用接圖（image registration)技術，將所有車道之前照 (後照）的影像整合成單一的前照（後照）環場影像供操作 者便於進行分析比對。 接圖單元使用特徵（feature)為基礎與區域為基礎的 方式混合運用，其中包含可見特徵（Visual Feature)、代 數特徵（Algebra Feature)、統計特徵（Statistical Feature)及轉換係數特徵（Transform Coefficient 12860271286027 玖, invention description: [Technical field of invention] The present invention is an image-based multi-lane free-vehicle electronic toll enforcement system for highways, roads, bridges, toll stations and specific places, as an electronic law enforcement mechanism for vehicle entry and exit Or pure video enforcement mechanism, can also be combined with VCU / license plate blacklist database, as a search and use of fake VCU. [Prior Art] For the traditional image law enforcement system, in order to obtain a complete panoramic image, it must be divided into multiple cameras to take images from different angles. However, the multiple images obtained by this method have low correlation and are not easy to complement each other. . Therefore, if the traditional image law enforcement system fails in the operation of optical character recognition (OCR), or for the application of the electronic toll collection system, the operator must fail from a large number of failure graphics when the back-end matching work fails. The correlation is found in the file, but because different camera systems take images from different angles, there is no feeling of continuous space. SUMMARY OF THE INVENTION In order to solve this problem, the present invention proposes to use an image registration technique to integrate individual images of multiple cameras at the same time into a ring image, so that the operator can compare the artificial comparisons. The feeling of continuous space also simplifies the operator's manual comparison of the time of 1286027. SUMMARY OF THE INVENTION The image law enforcement system is composed of an image capturing unit, a license plate recognition unit, a connection unit and a vehicle type identification unit. The image capturing unit includes a front (backlighting) sensing unit, a front-view (backlighting) CCD (Charge-Coupled Device) camera, a 'frame grabber' (Frame Grabber), and a DI0 (Digital Input/Output) card. . The front (backlighting) sensing unit is responsible for detecting the vehicle entering the front (backlighting) shooting area, and transmitting the sensing message to the image capturing card by the DI0 card for the front photo (backlighting) CCD camera. Multiple sets of single front (backlight) vehicle images to capture the vehicle's license plate number using the license plate recognition unit. Using image registration technology, all front-facing (backlit) images of the lanes are integrated into a single front-back (backlit) surround image for easy comparison by the operator. The connection unit uses a feature-based and region-based approach, which includes visual features, Algebra features, statistical features, and conversion factor features (Transform Coefficient 1286027).
Feature) ° 【實施方式】 請參閱圖一所示，係本發明所提供之系統架構流程實 施例： a. 以前照（後照）感應單元負責偵測進入前照（後照） 拍攝區的車輛； b. 將偵測到車輛的感應訊息藉由DI0卡傳送給影像 擷取卡，藉以觸發前照（後照）CCD攝影機拍取多組單張前 照（後照）車輛影像； c. 利用車牌辨識單元擷取車輛的車牌號碼。 前述方法由於取像單元藉由前照（後照）感應單元觸 發取像，所以可以避免因不同的車速而導致的取像失敗或 因大小車混合所導致的取像失敗。此外，為避免產生同一 台車多次取像導致取像資料過多而造成後端比對機制的 困擾，所以我們採用鎖定（1 ock)的方法來解決此一問題。 所謂鎖定是指當前照（後照）感應單元感應到車輛進入感 應區後自動取像，並鎖定住取像單元，直到前照（後照） 感應單元感應車輛離開感應區。 圖二為本發明之四車道的CCD攝影機架設位置實施 例。如圖所示，我們分別將上行的前照CCD攝影機命名為 UF卜5，上行後照的CCD攝影機命名為UB1〜5，下行的前 l286〇27 照CCD攝影機命名為卯卜5，下行的後照CCD攝影機命名 為DB1〜5。在正常行駛的情況下，如果上行車輛感應區1(3) 觸發，則UF2(UF4)會進行取像；如果是下行車輛感應區 4(6)觸發，則DF2(DF4)會進行取像。當車輛變換車道時， 如果使上行車輛感應區2觸發，則UF1、UF3、UF5會進行 取像；如果使下行車輛感應區5觸發，則DF1、DF3、DF5 會進行取像。為避免徒增後端人工處理機制的困擾，如果 車牌辨識程式能成功辨識UF3或DF3的車牌影像，則 UF卜UF5或DF1、DF5並不會使用；相對地，如果車牌辨 識程式無法成功辨識UF3或DF3的車牌影像，則會自動對 UF卜UF5或DF1、DF5進行車牌辨識，並將結果送交後端 人工處理機制進行自動比對。 由於多車道自由車流可能會同時摻雜大貨車、小轎 車，則小轎車的車牌很有可能被大貨車遮住，所以，我們 提出使用後照的方式來加以補強。亦即，當谓測到平均車 速過慢，則自動啟動後照功能，換言之，在塞車的情況下， 如果上行車輛感應區1觸發，則UF2、UB2會進行取像； 如果上行車輛感應區3觸發，則UF4、UB4會進行取像； 如果是下行車輛感應區4觸發，則DF2、DB2會進行取像； 如果是下行車輛感應區6觸發，則DF4、DB4會進行取像； 當車輛變換車道時，如果使上行車輛感應區2觸發，則 1286027 UF卜UF3、UF5、UBl、UB3、UB5會進行取像；如果使下 行車輛感應區5觸發，則DF1、DF3、DF5、DB1、DB3、DB5 會進行取像。為避免徒增後端人工處理機制的困擾，如果 車牌辨識程式能成功辨識UF3或DF3的車牌影像，則 UF卜UF5、UB卜UB5或DF卜DF5、UB1、UB5並不會使用； 相對地，如果車牌辨識程式無法成功辨識UF3、UB3或 DF3、DB3的車牌影像，則會自動對UF1、UF5、UB1、UB5 或DF1、DF5、DB1、DB5進行車牌辨識，並將結果送交後 端人工處理機制進行自動比對。 此一方法形同第一位專家先利用UF2、UF3、UF4或 DF2、DF3、DF4的前照影像來進行檢驗車牌取像是否成 功’如果不成功，才動用第二位專家針對un、UF5或DF卜 DF5的前侧照影像進行檢驗，如果還不成功，則動用第三 位專家針對UB2、UB3、UB4或DB2、DB3、DB4的後照影像 進行檢驗車牌取像是否成功，如果不成功，再動用第四位 專豕針對的後侧照影像UB卜UB5或DB1、DB5進行檢驗， 最後一道關卡，則是操作員這個第五位專家來進行人工比 對。運用這種多重專家的方式可有效減少取像失敗的機 率’大幅提昇電子收費系統的可靠度。雖然會增加取像資 料里’值利用多重專家的方式可加速系統運算時間。使用 刖後照搭配的方式，可有效解決塞車時大貨車遮蔽小轎車 1286027 的問題，也可避免為規避繳費而將車距拉小所導致之取像 失敗。由於傳統的執法系統，為了取得完整的全景，所以 分成多台攝影機分別取像，但是如果後端Matching失 敗，則操作員必須從大量的圖形檔當中找出關聯性，但比 較沒有連續空間的感覺。本發明為解決此一問題提出使用 接圖（Image Registration)單元的技術將同一時間多台 攝影機之個別影像加以整合成環場影像，讓操作員在進行 人工比對的時候，能夠比較有連續空間的感覺，也可以簡 化操作員人工比對的時間。 此外，本發明提出利用雷射感測器作為車型辨識單 元，利用雷射感測器所連續取得之形象（profile)資訊加 以判斷車型，如果系統應用在電子收費系統，則系統可依 據影像執法系統所判斷出來之車輛車型與VCU所設定的 車型來做判斷，如果不同，則影像執法系統就可以逕行開 立罰單。 【特點及功效】 本發明與其它習用技術相互比較時，更具有下列之優 (1)運用這種多重專家的方式可有效減少取像失敗的 機率，大幅提昇電子收費系統的可靠度；雖然會增加取像 資料量，但利用多重專家的方式可加速系統運算時間。 1286027 (2) 使用鎖定（lock)機制可有效防止重覆扣款及重覆 取像，減少後端自動及人工比對機機的負擔。 (3) 使用前後照搭配的方式，可有效解決塞車時大貨 車遮蔽小轎車的問題，也可避免為規避繳費而將車距拉小 所導致之取像失敗。 (4) 即使車輛以高速（可達250Km/hr)通過收費車道 通訊區，亦能正確無誤的完成自動扣款與影像執法交易程 序，達到自動收費之功能。 (5) 使用接圖（Image Registration)的技術將同一時 間多台攝影機之個別影像加以整合成環場影像，讓操作員 在進行人工比對的時候，能夠比較有連續空間的感覺，也 可以簡化操作員人工比對的時間。 (6) 利用感測器進行車型辨識，利用感測器所連續取 得之prof i 1 e資訊加以判斷車型，如果系統應用在電子收 費系統，則系統可依據影像執法系統所判斷出來之車輛車 型與VCU所設定的車型來做判斷，如果不同，則影像執法 系統就可以逕行開立罰單。上列詳細說明係針對本發明之 一可行實施例之具體說明，惟該實施例並非用以限制本發 11 1286027 明之專利範圍，凡未脫離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。 【圖式簡單說明】 圖一為本發明之系統架構流程實施例；以及 圖二為本發明之CCD攝影機之架設位置實施例。 【主要部分代表符號】 UF1〜UF5 CCD攝影機 UB1〜UB5 CCD攝影機 DF1〜DF5 CCD攝影機 DB1〜DB5 CCD攝影機 1〜5車輛感應區 12[Embodiment] Please refer to FIG. 1 for an embodiment of the system architecture process provided by the present invention: a. The front-light (backlighting) sensing unit is responsible for detecting the vehicle entering the front-lighting (backlighting) shooting area. b. The detected information of the detected vehicle is transmitted to the image capture card by the DI0 card, thereby triggering the front camera (backlighting) CCD camera to take multiple sets of single front (backlight) vehicle images; c. using license plate recognition The unit retrieves the license plate number of the vehicle. In the foregoing method, since the image taking unit triggers the image taking by the front (backlight) sensing unit, it is possible to avoid image failure due to different vehicle speeds or image failure due to mixing of large and small cars. In addition, in order to avoid the trouble of causing the back-end comparison mechanism caused by too many image acquisitions caused by multiple images taken by the same car, we use a locking method to solve this problem. The so-called locking means that the current photo (backlighting) sensing unit senses that the vehicle enters the sensing area and automatically takes the image, and locks the image capturing unit until the front (backlighting) sensing unit senses that the vehicle leaves the sensing area. Fig. 2 is a view showing an example of the position of a four-lane CCD camera frame according to the present invention. As shown in the figure, we named the up-front CCD camera as UF Bu 5, the CCD camera as the back-illuminated UB1~5, and the first l286〇27 as the CCD camera as the 55, after the downlink According to the CCD camera, it is named DB1~5. In the case of normal driving, if the uplink vehicle sensing area 1 (3) is triggered, UF2 (UF4) will perform image capturing; if the downstream vehicle sensing area 4 (6) is triggered, DF2 (DF4) will perform image capturing. When the vehicle changes lanes, if the uplink vehicle sensing area 2 is triggered, UF1, UF3, and UF5 will perform image capturing; if the downlink vehicle sensing area 5 is triggered, DF1, DF3, and DF5 will perform image capturing. In order to avoid the trouble of adding manual processing mechanism at the back end, if the license plate recognition program can successfully identify the license plate image of UF3 or DF3, UF UF5 or DF1, DF5 will not be used; in contrast, if the license plate recognition program cannot successfully identify UF3 Or the license plate image of DF3 will automatically identify the license plate of UF Bu UF5 or DF1, DF5, and send the result to the back end manual processing mechanism for automatic comparison. Since the multi-lane free traffic may be mixed with large trucks and small cars at the same time, the license plate of the car is likely to be covered by the big truck, so we propose to use the backlight to reinforce. That is, when the average vehicle speed is measured to be too slow, the backlight function is automatically activated. In other words, in the case of traffic jam, if the uplink vehicle sensing area 1 is triggered, UF2 and UB2 will perform image capturing; if the uplink vehicle sensing area 3 If triggered, UF4 and UB4 will perform image acquisition; if it is triggered by the downlink vehicle sensing area 4, DF2 and DB2 will perform image capturing; if the downstream vehicle sensing area 6 is triggered, DF4 and DB4 will perform image capturing; In the lane, if the uplink vehicle sensing area 2 is triggered, 1286027 UF, UF3, UF5, UB1, UB3, UB5 will perform image capturing; if the downstream vehicle sensing area 5 is triggered, DF1, DF3, DF5, DB1, DB3, DB5 will take the image. In order to avoid the trouble of adding manual processing mechanism at the back end, if the license plate recognition program can successfully recognize the license plate image of UF3 or DF3, UF U, Ub5, UB, UB5 or DF DF5, UB1, UB5 will not be used; If the license plate recognition program cannot successfully identify the license plate image of UF3, UB3 or DF3, DB3, it will automatically identify the license plate of UF1, UF5, UB1, UB5 or DF1, DF5, DB1, DB5, and send the result to the back end manual processing. The mechanism performs automatic comparisons. This method is similar to the first expert to use the front image of UF2, UF3, UF4 or DF2, DF3, DF4 to check whether the license plate image is successful or not. If not successful, the second expert is used for un, UF5 or The front side of the DF DF5 is scanned for image. If it is not successful, the third expert is used to check whether the license plate image is successful for the back image of UB2, UB3, UB4 or DB2, DB3, DB4. If it is not successful, Then use the fourth back-to-back image UB UB5 or DB1, DB5 for the final inspection. The last level is the fifth expert of the operator to perform the manual comparison. The use of this multi-expert approach can effectively reduce the chances of image failures' dramatically increasing the reliability of electronic toll collection systems. Although it will increase the value of the image data, the use of multiple experts can speed up the system operation time. The use of the squatting and illuminating method can effectively solve the problem that the large truck shields the small car 1286027 during the traffic jam, and can also avoid the image failure caused by pulling the small distance to avoid the payment. Because of the traditional law enforcement system, in order to obtain a complete panorama, it is divided into multiple cameras to take images separately, but if the back-end Matching fails, the operator must find the relevance from a large number of graphics files, but there is no feeling of continuous space. . In order to solve this problem, the present invention proposes to use an image registration unit technology to integrate individual images of multiple cameras at the same time into a ring image, so that the operator can compare the continuous space when performing manual comparison. The feeling can also simplify the time for the operator to manually compare. In addition, the present invention proposes to use a laser sensor as a vehicle type identification unit, and utilizes the profile information continuously obtained by the laser sensor to determine the vehicle type. If the system is applied to the electronic toll collection system, the system can be based on the image law enforcement system. The judged vehicle model and the model set by the VCU are used for judgment. If they are different, the image law enforcement system can open a ticket. [Features and Efficacy] When the present invention is compared with other conventional technologies, it has the following advantages: (1) The use of such multiple experts can effectively reduce the probability of image failure and greatly improve the reliability of the electronic toll collection system; Increase the amount of image data, but use multiple experts to speed up system computing time. 1286027 (2) The use of the lock mechanism can effectively prevent repeated deduction and repeated image retrieval, reducing the burden on the back end automatic and manual comparison machine. (3) The use of front-rear and collocation can effectively solve the problem of obscuring a small car when a traffic jam occurs, and avoiding the failure of the image taken by pulling the car to avoid the payment. (4) Even if the vehicle passes the toll lane communication area at high speed (up to 250Km/hr), the automatic deduction and video law enforcement transaction procedures can be completed without any mistakes, and the automatic charging function can be achieved. (5) Using Image Registration technology to integrate individual images of multiple cameras at the same time into a ring image, allowing operators to compare the feeling of continuous space and simplify when performing manual comparisons. The time the operator manually compares. (6) Using the sensor for vehicle type identification, using the prof i 1 e information continuously obtained by the sensor to judge the vehicle type. If the system is applied to the electronic toll collection system, the system can be determined according to the vehicle model and the image vehicle system. The model set by the VCU is used to judge. If it is different, the image law enforcement system can open a ticket. The detailed description above is a detailed description of one of the possible embodiments of the present invention, and is not intended to limit the scope of the invention as defined by the present invention. It is included in the patent scope of this case. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is an embodiment of a system architecture flow of the present invention; and FIG. 2 is an embodiment of a mounting position of a CCD camera of the present invention. [Main part representative symbol] UF1~UF5 CCD camera UB1~UB5 CCD camera DF1~DF5 CCD camera DB1~DB5 CCD camera 1~5 vehicle sensing area 12