As a key technology of the underwater intelligent robot’s perception,the underwatertarget automatic detection and recognition have been a practical significance and challengingresearch topic of today’s image processing domain. The underwater target recognitionsystem plays an important role in navigation, obstacle avoidance and detecting unknownobjects for underwater robots. Image pre-processing results will affect results of image targetrecognition and tracking.In this paper underwater image pre-processing including image filtering,edgedetection,image segmentation technology was researched. Underwater target images wereachieved from national defence key laboratory of autonomous underwater vehicle technology.The advantages and disadvantages of the traditional image filtering algorithms were analyzedin this paper. Non-local means filtering algorithm can achievea a better image target edgefrom particles noise image. Combining the characteristics of underwater images, impulsenoise model was established.At the same time, impulse noise in the underwater image canalso be removed. PCNN filtering was researched and simulated. Particle swarm optimizationalgorithm was applied into PCNN model to solve setting PCNN parametersAutomatically.The traditional edge extraction algorithms were studied and sorted. The activecontour algorithm has been applied to the underwater image processing. This algorithm wasfit for underwater sphere and ellipsoid target extraction.Above all, the underwater image pre-processing technologies are researched deeply.Non-local means filtering, image filtering based on PCNN, active contour algorithm proposedin this paper were tested by using real underwater image taken. The experiments results showthese algorithms are effective and performance is well. |