The invention discloses an AUV (Autonomous
Underwater Vehicle)
pose correction method and
system based on perspective conversion and a storage medium, and belongs to the field of
underwater robots. The method comprises the steps that an
underwater circular
light source lamp array image is shot through a
monocular camera, and initial circle center feature points are extracted after preprocessing and contour screening; using a P3P
algorithm to preliminarily solve the relative
pose and constructing a
perspective transformation matrix; carrying out
ellipse fitting on the contour and sampling on an
ellipse, and projecting sampling points to a
virtual plane through inverse
perspective transformation; calculating an average circle center coordinate on the
virtual plane, and re-projecting the average circle center coordinate to the
image plane to obtain a corrected feature point; and finally, calculating a high-precision final
pose by using the correction point through a P3P
algorithm. According to the method, the
virtual world plane is constructed by adopting a perspective conversion-based method with relatively small
operand, the circle center coordinate is obtained in a circle correction mode by utilizing the characteristic that the inclination angle between the
virtual world plane and the real world plane is relatively small, the influence of an eccentric distance phenomenon on the extraction of the circular target feature point is reduced, and the AUV pose resolving precision is effectively improved.