Ab­stract

To achieve a high test cov­er­age of Ad­vanced Dri­ver As­sis­tance Sys­tems, many dif­fer­ent en- vi­ron­men­tal con­di­tions have to be tested. It is im­pos­si­ble to build test sets of all en­vi­ron­men­tal com­bi­na­tions by record­ing real video data. Our ap­proach eases the gen­er­a­tion of test sets by using real on-road cap­tures taken at nor­mal con­di­tions and ap­ply­ing com­puter-gen­er­ated en­vi­ron­men­tal vari­a­tions to it. This paper pre­sents an eas­ily in­te­grable frame­work that con- nects vir­tual pro­to­types with vary­ing sen­sor per­cep­tions. With this frame­work we pro­pose a method to re­duce the re­quired amount of on-road cap­tures used in the de­sign and val­i­da­tion of vi­sion-based Ad­vanced Dri­ver As­sis­tance Sys­tems and au­tonomous dri­ving. This is done by mod­i­fy­ing real video data through dif­fer­ent fil­ter chains. With this ap­proach it is pos­si­ble to sim­u­late the be­hav­ior of the tested sys­tem under ex­treme con­di­tions that rarely occur in re­al­ity. In this paper we pre­sent the cur­rent state of our vir­tual pro­to­typ­ing frame­work and the im­ple­mented plug-ins.