Data Sheet - cctvxanpr PC based 1-4 channel ANPR (Automatic Licence Number Plate Recognition)

Transcription

1 Data Sheet - cctvxanpr PC based 1-4 channel ANPR (Autmatic Licence Number Plate Recgnitin) the hme f Kdicm in Main features Real time PC based recrding Neural Net ANPR system Multi user peratin Event lg Intelligent search functin Fully functinal scene setup Wrks with CCTVX ANPR cameras and cctvx ANPR DVR Ideal Applicatins Parking management Petrl statin frecurt surveillance Tll cllectin Vehicle access cntrl Law enfrcement Supermarket check-uts The cctvxanpr represents the very latest in PC based ANPR technlgy prviding rck slid stability and image quality as well as super high accuracy ANPR (Autmatic Licence Number Plate Recgnitin). This feature rich platfrm gives access t perfrmance previusly unheard f as this price pint and culd be seen as the engine at the heart f any prfessinal CCTV surveillance system, when used with cctvx ANPR cameras The ANPR is based n a Neural net design in ther wrds it will quickly learn t read number plates f all types and languages that can be stred alng with pictures in the database, r stred in cnjunctin with data n ne f ur DVRs What makes this unit s unique is its teaching technlgy using a neural netwrk as ppsed t rdinary OCR (Optical Character Recgnitin) methds, making fr far higher accuracy f infrmatin. The whle system is then further enhanced by the use f remte mnitring sftware, that means pictures can be viewed and the system prgrammed in remte lcatins using a web brwser.

2 Data Sheet - cctvxanpr PC based 1-4 channel ANPR (Autmatic Licence Number Plate Recgnitin) the hme f Kdicm in Specificatins (all specs subject t mdificatin) CPU RAM HDD Muse + Keybard Mnitr Operating Sftware Intel Cre2Du 2 Gbyte DDR Gbyte Supplied VGA Mnitr needed (nt supplied) Windws XP Prfessinal Cntact cctvx Unit 7 Cygnet Business Centre Hanley Swan Wrc. WR8 0EA Tel: Fax: sales@cctvx-cctv.cm Part f the new slutins range f publicatins frm cctvx, the specialist distributr in CCTV prducts cctvx the exclusive hme f the Wrld leading Kdicm range f Surveillance prducts including DVR s and Cameras

3 the hme f Kdicm in ANPR LPR ALPR Autmatic Number / License Plate Recgnitin System Data Sheet and Specificatin The CCTVX ANPR (Autmatic Number Plate Recgnitin) system is a sftware recgnitin engine, capable integrating and fulfilling mst requirements where a vehicles number plate is used r recrded as an identifier. Ppular uses include tll-cllectin, vehicle access cntrl, parking management and law-enfrcement applicatins. Prven Technlgy The advanced ANPR/LPR (Autmatic Number Plate Recgnitin/License Plate Recgnitin) technlgy frm CCTVX sets a new benchmark thrugh its cmbinatin f best-f-breed neural netwrk sftware design by ffering the highest success in cntinuus rapid read rate fr vehicles either statinary r travelling at speeds in excess f 180 km/h. Neural Netwrk This wrld-class prduct cntinuusly imprves bth read accuracy and recgnitin speed during peratin thrugh NeuvNet Neural netwrking r sftware learning techniques embedded in its design, applying varius LineDraw and Clur Fading prprietary sftware techniques. Single Deplyment - Multiple Applicatins Multiple applicatins via a single deplyment can be achieved thrugh the use f pen DLL s, enabling easy integratin f legacy and enterprise systems t its pen architecture withut impacting the remarkable affrd ability f this prven technlgy. Mdular Design This key feature, as well as mdular design architecture, allws fr rapid deplyment in a variety f multi-lane applicatins and highlights anther advantage achieved thrugh the use f a Vista-ready, Windws based prduct that is easily scalable thrugh nrmal netwrk prtcl standards. Real-Time Cmmunicatin Remte access via telecmmunicatins further enhances the flw f infrmatin in critical reactin requirements such as instantaneus Ht Plate uplads. Ease f Use The rugged hardware design is packaged with client-driven sftware t result in a friendly user interface that can be applied in simplistic user envirnments. Rapid Recgnitin Rate The rapid and effective recgnitin prcess and adjunct applicatins are backed up by a thrugh lg prcess, runding ff with an audit trail f the event histry fr imprved infrmatin systems and management reprting.

4 Overview f the Traditinal VS Neural Netwrking the hme f Kdicm in Traditinal methds used fr number plate recgnitin: OCR Optical Character Recgnitin Frmula Based Recgnitin Althugh bth these methds have sme merit, bth have inherent prblems; OCR makes use f a simple methd f placing a numerical r alphabetical histrical character image ver the captured image and cmparing the tw fr accuracy. While frmula based systems make use f cmplex algrithms t cmpare the captured numerical r alphabetical character with a similar value in it database. Bth systems require a lengthy cmparisn prcess the average time t analyze a cmplete number plate is between 250ms and 500ms Breakthrughs in cmputer prcessing abilities has allwed CCTVX t mve away frm the traditinal OCR Optical Character Recgnitin and Frmula Based number plate recgnitin methds t the use f ur prprietary NeuvNet Neural Netwrk intelligence engine, ensuring a greater accuracy rate while imprving the recgnitin speed. Average speed t recgnize a number plate: between 80ms and 200ms. The CCTVX ANPR NeuvNet system will teach itself the average places in the image were it is likely t find a number plate as well as what the average number size will be, this is dne t increase reading speed. With its inbuilt training system the user can with minimal cmputer skills train the system t recgnize new characters r symbls such as thse fund in the Chinese r Arabic alphabet allwing fr rapid deplyment int new cuntries and territries. The CCTVX NeuvNet sftware will assign a cnfidence level t each character in its trained character layers, thus in fact psing a questin t the user asking fr cnfirmatin n its accuracy. Characters r symbls with a lw cnfidence level can be trained t imprve accuracy and read speeds. NeuvNet System Advantages The CCTVX ANPR systems NeuvNet apply several prprietary methds t the image in rder t btain a clear recgnizable alphanumerical character: Neural filter LineDraw Clur Fade Each time a character is recgnized crrectly the answer is stred separately in a new neural layer. New recgnitin prcesses will first questin the newest neural layer fr an answer thus imprving speed by reducing the recgnitin prcesses. This ensures increased speed and accuracy fr cameras in a statinary psitin. By redrawing each alphanumerical character hrizntally and vertically with parameters, the CCTVX ANPR system remves mst imperfectins r bstructins frm the number plate. In sme cuntries number plates are nt reflective and cntain images in the backgrund, the CCTVX Clur fade prcess will fld the image with multiple clurs and remve each individual clur until nly the alphanumerical characters f the number plate remain.

5 the hme f Kdicm in Individual Numerical & Alphabetic Character Training This wrld first, built-in prgram feature will allw the peratr access t train the system t recgnize new characters. Additinal characters r symbls such as thse used in the Chinese r Arabic alphabet can then simply be added t the system within days withut the help f CCTVX system develpers. Character Segmentatin In sme cases, a character may be partially bstructed, by an insect smear r tw hitch. Character segmentatin will then make use f a partial numerical r alphabetical character t recnstruct the character. Even with perfect alphanumerical characters present, recgnizing a segment f a character, the NeuvNet system will instantaneusly revert t its mst likely cmparisns in the neural layers thus increasing accuracy and speed. Plate Filters By prviding the CCTVX ANPR systems NeuvNet with a basic filter cntaining the likely alphanumerical sequence f the number plate greatly imprves recgnitin accuracy and speed. Several filters can be applied but in cases such as a persnalized number plate, the NeuvNet will ignre the filter and apply a Brute Frce methd fr recgnitin. Autmatic Number Scaling The CCTVX system builds its wn Aut Number Scaling prviding the NeuvNet with a likely character size. This greatly reduces internal pssesses and increases accuracy and speed. Dual Layer character Matching By quick matching a pssible character with ptential matches in the neural layers, essentially shrt-listing the alphanumerical characters and then cmparing the shrtlist with the reduced pssibilities, ur system reduces system prcesses as well as increasing accuracy and speed. (Duble-Check) Multiple Read Methd Free Flw Methd The CCTVX ANPR system is capable f prcessing 25 images per secnd, per channel. It reads each passing number plate mre than nce. By cmparing its first read with the secnd and third it btains a higher level f cnfidence in its accuracy and will display the mst likely number plate sequence. Depending n the placement f the system and the average speed f the vehicle, the multiple reads functin can be set frm tw reads t ten reads. In a case such as the law enfrcement applicatin, were the number plate is cmpared with a database, the Free Flw methd will pass all the number plates that are prcessed t a database. The database will be questined several times per the passing vehicle. This increases the number f successful database hits. ODD Line Only Methd In cases were vehicles are mving at high speeds the CCTVX ANPR system will nly acquire the ODD lines frm the image and cmpress the ODD lines int a single image, remving all ptential image smear frm the image. (2CIF)

6 the hme f Kdicm in ODD and EVEN Line Methd Fr increased accuracy the system can individually extract the ODD and EVEN fields frm the image cmpressing them int tw new images then cmpare each new image individually t display a singular result. (2CIF x 2) Mtin detectin CCTVX ANPR makes use f pure image mtin detectin. The need f external triggers and lines being cut int the rad is eliminated. Event Recrding CCTVX makes use f a prprietary event-recrding lg; each individual number plate recgnized is stred in a managed lg file with the image used and time-date stamp. The lg file features include: Search by number Plate Search by date Search by time Search by camera Individual recgnitin printing (Time, date, transactin data and image) Lg printing Individual image saving (Time, date, transactin data and image) Audit Lg The inbuilt aut-lgging system will ensure a cmplete audit trail f all user events as well as attempts by unauthrized users t execute administrative functins. The infrmatin is prvided in a User, Time, Date, Operatin frmat and can be printed r saved as a user system audit trail. System Security ANPR system cntains sensitive infrmatin and need t be prtected with adequate security features, fr this reasn CCTVX makes use f ur security grup system, allwing varius grups f user s access t levels f infrmatin equivalent t the clearance awarded by the administratr. Traditinal user grups include the

7