In　the　realm　of　big　data　and　cloud　computing,　the　autonomous　vehicle　is　an　extremely　promising　and　complex　research　topic,　which　is　driven　primarily　by　computer　systems　that　employ　AI　technology,　radar,　GPS,　visual　computing,　cloud　computing,　and　other　technologies　to　operate.　autonomous　vehicles　use　LIDAR　to　sense　road　information,　and　highly　accurate　maps　are　integrated　with　V2X　(vehicle　networking　technology)　to　recognize　relevant　information　such　as　traffic　lights　and　speed　limit　signs.　However,　in　some　remote　areas　where　map　information　needs　to　be　completed,　recognition　of　traffic　lights　and　traffic　signs　cannot　rely　on　positioning　to　achieve　this.　This　paper　uses　the　Efficientdet-d1　target　detection　algorithm　built　on　Pytorch　to　simulate　autonomous　vehicles　sensing　pedestrian,　vehicle,　traffic　light,　and　traffic　sign　information.　This　target　detection　algorithm　uses　EfficientNet-B1　as　the　backbone　network　and　enhances　the　feature　extraction　process　using　four　stacked　BiFPN　modules.　The　method　involves　using　the　open　source　big　dataset　LISA　traffic　light　dataset　and　GTSRB　German　traffic　sign　dataset　to　train　the　model.　Considering　the　uneven　distribution　of　samples　in　the　dataset,　these　classes　are　distributed　into　three　Efficientdet-d1　target　detection　frameworks　for　pedestrians　and　various　types　of　vehicles,　traffic　lights,　and　traffic　signs,　respectively.　A　multi-threaded　approach　allows　the　three　detection　processes　to　be　executed　simultaneously.　The　detection　results　are　stored　in　a　queue　before　aggregation　for　mapping　to　improve　the　speed　of　single-image　execution.　The　experiments　show　the　three　prediction　networks　used　achieved　better　results　overall.　The　method　proposed　in　this　paper　is　a　practical　guide　for　autonomous　vehicles　to　make　road　condition　judgments　during　driving　in　remote　areas.