To address the issues of accuracy degradation and computational latency in extracting the centers of tiny light spots in high-speed industrial visual measurement, this paper proposes a high-speed real-time light spot localization method for small-sized spots. First, a region of interest (ROI) extraction algorithm based on sliding window brightness consistency was designed and implemented on an FPGA to enhance detection speed. Subsequently, a center extraction algorithm for tiny light spots was developed by integrating a distance-weighted least squares fitting method with an adaptive weighting adjustment mechanism based on signal-to-noise ratio. This approach improves the localization robustness of tiny light spots under varying lighting and noise conditions, enabling fast and high-precision center localization for small-sized spots. Experimental results demonstrate that in measurement scenarios involving small-sized spot extraction, the proposed method achieves a localization accuracy better than 0.05 pixels, with a processing speed approximately 36 times faster and a frame rate of 160 fps. It meets the dual requirements of high precision and low latency for real-time measurement with high-speed cameras, significantly enhancing real-time processing capabilities in long-distance, small-spot scenarios.