An Accurate Detection Algorithm for Time Backtracked Projectile-Induced Water Columns Based on the Improved YOLO Network

During a sea firing training, the intelligent detection of projectile-induced water column targets in video is prerequisite for and critical to automatic calculation miss distance, while correct precise distance directly affected by accuracy, false alarm rate time delay detection. After analyzing characteristics columns, an accurate algorithm backtracked columns based on improved you only look once (YOLO) network put forward. The capability accuracy detecting with conventional YOLO are optimizing anchor box through K-means clustering embedding squeeze excitation (SE) attention module. area limited adopting sea-sky line gray level co-occurrence matrix (GLCM), so as effectively eliminate such disturbances ocean waves ship wakes, lower increases mAP 50 30.3%. On basis detection, back-tracking designed mean shift track images containing reverse sequence. It accurately detects at its initial appearance well pixel position images, considerably reduces delay, provide support automatic, accurate, real-time distance.