A grayscale centroid iterative algorithm that enables parallel extraction of the position and shape information of the light spot

Real-time laser spot off-target detection based on image analysis constitutes a critical technology for high-precision Acquisition, Tracking, and Pointing (ATP) systems. This study presents a novel gray centroid algorithm incorporating image preprocessing and finite iteration to address limitations in traditional spot center detection methods. The proposed algorithm enables concurrent extraction of spot position and radius parameters while demonstrating enhanced antiinterference capabilities and real-time performance. The workflow begins with grayscale conversion of input spot images, followed by sequential median and Wiener filtering to suppress salt-and-pepper noise and Gaussian noise. OTSU threshold segmentation is then applied to generate binary images. An initial centroid estimation derived from the gray centroid method undergoes iterative refinement to reduce computational cycles and improve accuracy. Simulation experiments reveal the algorithm's superior robustness in noisy environments, achieving an average centroid localization error below 0.5 pixels. Comparative analyses show centroid offset reductions of 45.55% (horizontal) and 75.56% (vertical) compared to conventional gray centroid methods, 54.07% and 66.98% against least-square circle fitting, and 50.13% and 69.37% relative to Gaussian surface fitting. Hardware implementation on an FPGA platform operates at 100 MHz, achieving single-frame processing within 4.8 ms and centroid calculation accuracy of 0.01%. The method not only ensures efficient image data processing and precise spot positioning but also maintains reasonable spot contour fitting. These advancements provide an innovative solution for enhancing the performance of laser beam ATP systems by balancing computational efficiency with sub-pixel precision, addressing both real-time operational demands and environmental interference challenges in practical applications.