Machi e visio a d AI i fe e ce pipeli es depe d o co siste t image quality ac oss a full a ge of ambie t lighti g co ditio s, yet most etwo ked came a deployme ts still ely o se so s optimized mai ly fo daylight pe fo ma ce. I pa ki g st uctu es, wa ehouse loadi g docks, pe imete fe ci g, a d co ti uously ope ati g etail e vi o me ts, illumi atio ca swi g f om seve al thousa d lux at midday to a f actio of a lux afte dusk. A came a module without exte ded low-light se sitivity p oduces oisy, motio -blu ed, o u de exposed f ames p ecisely whe a AI visio model eeds its clea est i put fo accu ate detectio a d classificatio .
Exte di g a Low Light GigE Came a i to mixed light sce es exposes two e gi ee i g p oblems at o ce. The fi st is photo sta vatio : as ambie t light d ops, a se so with a small pixel o without Back-Side Illumi ated a chitectu e st uggles to collect e ough sig als above its ead oise floo . The seco d is dy amic a ge collapse: headlights, floodlights, a d eflective su faces at ight c eate localized highlights that clip alo gside da k shadow egio s i the same f ame, a p oblem that basic auto exposu e ca ot solve because a si gle exposu e setti g ca ot se ve both ext emes at o ce.
So y’s STARVIS se so li eup add essed the low light photo collectio p oblem th ough Back-Side Illumi ated pixel a chitectu e, a d STARVIS 2 exte ds that fou datio with a edesig ed pixel st uctu e that imp oves qua tum efficie cy fu the i to the ea i f a ed spect um while educi g da k cu e t oise at lo ge exposu e times. The esult is a So y STARVIS 2 Came a platfo m that captu es usable detail at illumi atio levels whe e se so s lacki g Back-Side Illumi ated co st uctio etu mostly oise.
Fo the dy amic a ge half of the p oblem, the same se so ge e atio adds o -chip exposu e combi atio , efe ed to i datasheets as DOL HDR, which eads two o mo e exposu e values pe pixel withi a si gle f ame pe iod a d me ges them befo e the data leaves the se so package. This p oduces a Low Light HDR Se so espo se that eithe clips a headlight o loses shadow detail beside it, without the f ame-to-f ame ghosti g that affects softwa e-based multi-f ame HDR.
Pai i g a IMX662 HDR GigE Came a a chitectu e with a GigE Visio i te face completes the system-level desig . GigE Visio u s ove sta da d Ethe et i f ast uctu e, suppo ts Powe ove Ethe et fo combi ed data a d powe delive y o a si gle cable, a d scales to cable u s up to 100 mete s without a epeate , which matte s fo pe imete secu ity, wa ehouse ceili g mou ts, a d t affic i f ast uctu e whe e outi g USB o MIPI cabli g back to a host compute is imp actical.
The I ova-662CRE pai s the So y IMX662 Came a se so with a fully i teg ated GigE Visio i te face, PoE powe delive y, a d a S-Mou t (M12) le s holde with Elect omecha ical IR Filte i a compact boa d-level desig suited to both di ect i teg atio a d e closed came a deployme t. As a 2MP GigE Came a, it outputs 1920 × 1080 esolutio at up to 60 f ames pe seco d with o -chip HDR active, givi g system i teg ato s a si gle came a platfo m that cove s daylight, twilight, a d ea -total-da k ess sce es without switchi g se so s o addi g exte al illumi atio fo most applicatio dista ces.
At the co e of the module sits the IMX662 Se so , a 1/2.8-i ch optical fo mat device with 2.9-mic o pixels a d Back-Side Illumi ated co st uctio that So y ma kets u de the STARVIS 2 label fo exte ded low light a d ea i f a ed se sitivity. Because the se so was e gi ee ed as a 2MP Low Light Se so f om the g ou d up, the I ova-662CRE eaches usable exposu e at illumi atio levels whe e competi g modules typically equi e gai i c eases that i t oduce visible oise i to a AI visio pipeli e. This lets it fu ctio as a Ult a Low Light gigabit ethe et Came a suited to u lit outdoo sce es as well as a ge e al-pu pose daytime came a o the same ha dwa e.
“E gi ee s buildi g AI visio a d su veilla ce systems keep aski g us fo o e came a that behaves the same at oo a d at mid ight without a seco d se so o a softwa e patch. The IMX662’s Back-Side Illumi ated STARVIS 2 a chitectu e combi ed with o -chip HDR gives the I ova-662CRE ge ui e low-light pe fo ma ce a d full dy amic a ge i the same f ame, a d GigE Visio with PoE mea s i teg ato s ca place it a ywhe e o the etwo k without a sepa ate powe u . We built this came a to emove the t adeoffs, ot ma age them.” – Alwi Vi ce t, P oduct Ma age at Vadzo Imagi g.
A: P ofile M cove s the metadata a d eve t i te face ONVIF defi es fo a alytics, ot o -came a a alytics p ocessi g itself. It allows the I ova-662CRE to expose st uctu ed metadata, such as motio o object eve ts, to a VMS o a alytics e gi e ove the sta da d ONVIF i te face athe tha a custom p otocol.
The I ova-662CRE 2MP GigE Came a is available ow fo evaluatio a d OEM i teg atio . Evaluatio kits i clude the came a module, a facto y-calib ated S-Mou t le s with Elect omecha ical IR Filte , a PoE i jecto , a d i teg atio docume tatio cove i g Ge ICam a d Vispa ARC SDK setup, with o mi imum o de equi eme t. Co tact Vadzo Imagi g at
Vadzo Imagi g develops embedded a d machi e visio came a p oducts fo OEMs a d system i teg ato s buildi g p oductio – eady visio systems ac oss secu ity, obotics, i dust ial automatio , sma t i f ast uctu e, a d edge AI. The compa y’s po tfolio spa s MIPI CSI-2, USB, GigE, Wi-Fi, a d Se Des i te face came a p oducts, suppo ti g deployme t a chitectu es f om compact o boa d modules to dist ibuted etwo ked i stallatio s. Vadzo p ovides e d-to-e d imagi g suppo t i cludi g se so i teg atio , ISP tu i g, fi mwa e developme t, a d SDK f amewo ks th ough the Vispa ARC SDK to accele ate system deployme t. Visit Vadzo Imagi g to explo e the full embedded visio came a po tfolio.