This paper presents an approach for camera-only localization, by aligning onboard observations to a pre-built reference feature map. By using only lane markings as features, the proposed method provides a lightweight alternative to multi-sensor and dense-map-based approaches. The framework is evaluated on a scaled autonomous vehicle platform in a controlled environment. A drone-based orthomosaic is used to create an aerial reference map. Onboard camera images are transformed into a bird’s eye view (BEV) representation to match the perspective of the reference map. Lane features are extracted and represented as point clouds, and alignment between local observations and the global map is performed using the Iterative Closest Point (ICP) algorithm. Experimental results demonstrate the feasibility of lane-based map alignment for localization, while also highlighting sensitivity to initialization.