Study on Underwater Sea Cucumber Rapid Locating Based on Morphological Opening Reconstruction and Max-Entropy Threshold Algorithm

In China, sea cucumber cultivation is developing rapidly, but sea cucumber catching still relies on inefficient manual work. Nowadays, the acquisition of underwater sea cucumber images and locating of sea cucumber target have provided technical support for underwater sea cucumber catching robots. However, there are still some problems to be solved, such as the degradation, edge blur and low contrast of underwater sea cucumber images due to the uneven light underwater and the absorption and scattering of light by water; and the cusp noises in the sea cucumber images produced by shells, gravers, planktons and other things in natural environment. Aiming at these problems, the underwater sea cucumber rapid locating method based on morphological opening reconstruction and max-entropy threshold algorithm (OR-META) is proposed in this paper. Firstly, the morphological opening reconstruction is adopted to smooth the original sea cucumber gray image; next, the max-entropy threshold segmentation algorithm is employed to segment the smoothed sea cucumber image, and the sea cucumber region in binary image is recognized according to area characteristics; finally, the position of sea cucumber region in the image is determined utilizing its centroid. In order to obtain the best result of sea cucumber image segmentation, three typical image segmentation algorithms OR-2DMETA, OR-OTSU and OR-2DOTSU are selected to compare with the OR-META. It is observed that the OR-META is obviously superior to the other three algorithms in qualitatively analyzing the image segmentation quality and quantitatively calculating the running time. To further analyze the image segmentation quality and locating accuracy, the images segmented by the OR-META are compared with manually segmented images. The comparison shows that although the image segmentation accuracy of the OR-META is low, with average target segmentation correct rate of 68.99%, the position of sea cucumber target in segmentation is predictable, which means that it will be within the directly drawn sea cucumber region. This demonstrates that the locating accuracy of the method is high. In addition, the method also has good real-time performance. It can be concluded from the experiment that for a RGB image with resolution ratio of 1280 x 720 pixels, the average centroid Euclidean distance error of the OR-META segmented image is only 52.67 pixels, and its average running time is 0.6 s, which are qualified for the requirements of the sea cucumber catching robots.