To address the issues of accuracy degradation and computational delay in extracting small spot centers in the field of industrial high-speed visual measurement, a high-speed real-time spot localization method for small-sized spots is proposed. A Region of Interest (ROI) extraction algorithm based on sliding window brightness consistency is designed and implemented in a Field Programmable Gate Array (FPGA) to improve detection speed. A small spot center extraction algorithm combining distance-weighted least-square fitting and a Signal-to-Noise Ratio (SNR)-based adaptive weight adjustment mechanism is introduced to enhance the localization robustness of small spots under varying lighting and noise conditions. Experimental results show that the proposed method has achieved a spot center localization error of not more than 0.05 pixels, with a frame rate of 160 frames per second, significantly outperforming traditional methods in processing speed. This method to a great extent meets the high-precision real-time localization requirements of small spot centers in industrial high-speed visual measurement.