基于三维视觉的海洋生物区域种类识别仿真

摘要

Traditional recognition method for species of marine organism district always ignores strengthen treatment for edge feature of the marine organism,which leads to poor classification precision.In order to solve the problem,based on three-dimensional (3D) vision,a new recognition method is proposed in this article.Edge curve of image of the marine organism is projected into space plane and space recovery is carried out for surface information of structure of the marine organism.3D coordinate of image border is obtained and edge pixel of the image is recognized.Gradient of each effective pixel is calculated.Direction and intensity of the edge structure are introduced to enrich edge feature of homogeneous region of the marine organism.Simulation results show that the method has high recovery precision of 3D coordinate of measure point under area self-occlusion.It improves matching precision and recognition effect of edge of the marine organism.%对海洋生物区域种类的识别,能够对海洋生物种类构成和丰度进行有效识别和定量分析.对海洋生物区域种类进行识别,需要结合边缘结构的方向和强度,丰富海洋生物同质区域的边缘特征,完成生物种类识别.传统方法将图像第一卷积层特征图划分为多个子块图像,计算海洋生物子块图像矩阵的奇异值,但忽略了对海洋生物边缘特征的强化处理,导致分类精度偏低.提出基于三维视觉的海洋生物种类识别方法.将海洋生物图像边缘曲线投影至空间平面,对海洋生物结构表面信息进行空间恢复,获得海洋生物图像边缘三维坐标,将海洋生物图像的边缘像素识别出来,计算各个有效像素的梯度.加入边缘结构的方向和强度,丰富海洋生物同质区域的边缘特征.实验结果表明,该方法对于区域自遮挡下的测点三维坐标恢复精度较高,提高了海洋生物边缘的匹配精度以及识别效果.