This　study　enhances　the　YOLOv8n　algorithm　for　nighttime　vehicle　detection　challenges.　It　preprocesses　nighttime　imagery　with　the　Retinex　algorithm　to　improve　image　quality　under　low-light　conditions.　The　introduction　of　MishDyHead,　a　dynamic　detection　header,　replaces　the　original　Detect　structure　and　significantly　enhances　vehicle　identification　accuracy　in　nocturnal　scenes.　Furthermore,　network　light　weighting　is　achieved　by　replacing　Bottleneck　with　FasterEMA_Block　and　integrating　the　EMA　mechanism　to　enhance　the　C2F　module.　Additionally,　the　novel　IoU　loss　function,　MPDIoU　(minimum　point　distance　IoU),　replaces　the　conventional　CIOU　loss　function,　further　optimizing　bounding　box　prediction.　Experimental　results　demonstrate　a　2.5%　increase　in　Average　Precision　(AP)　and　a　6.7%　reduction　in　parameters　compared　to　the　original　YOLOv8n　algorithm.