基于尺度不变特征转换算法的棉花双目视觉定位技术

发布时间：2018-01-04 10:58

本文关键词：基于尺度不变特征转换算法的棉花双目视觉定位技术　出处：《农业工程学报》2016年06期 　论文类型：期刊论文

【摘要】：为了给采棉机器人提供运动参数,设计了一套双目视觉测距装置以定位棉株。对获取的左右棉株图像进行经背景分割等预处理。求取其在8个尺度下的高斯图,通过尺度不变特征转换SIFT(scale-invariant feature transform)算法在相邻高斯差分图中提取出SIFT关键点;计算每个高斯图中关键点邻域内4×4个种子点的梯度模值,得到128维特征向量。分割右图关键点构成的128维空间,得到二叉树;利用最优节点优先BBF(best bin first)算法在二叉树中寻找到172个与左图对应的粗匹配点。由随机采样一致性RANSAC(random sample consensus)算法求出基础矩阵F,恢复极线约束,剔除误匹配,得到分布在11朵棉花上的151对精匹配。结合通过标定和F得到的相机内外参数,最终重建出棉花点云的三维坐标。结果表明,Z轴重建结果比较接近人工测量,平均误差为0.039 3 m,能够反映棉花间的相对位置。
[Abstract]:In order to provide motion parameters for cotton picking robot, a binocular vision ranging device was designed to locate cotton plant. The obtained left and right cotton plant images were preprocessed by background segmentation, and Gao Si images were obtained at 8 scales. SIFT(scale-invariant feature transformation by scaling invariant features. The algorithm extracts the key points of SIFT from the adjacent Gao Si difference graph. The gradient norm value of 4 脳 4 seed points in each Gao Si graph is calculated, and the 128-dimensional eigenvector is obtained, and the 128-dimensional space formed by dividing the key points in the right graph is segmented, and the binary tree is obtained. Using the optimal node-first BBF(best bin first algorithm, 172 rough matching points corresponding to the left graph are found in the binary tree. Random sample Consensus) algorithm to find the fundamental matrix F. By restoring the constraint of the pole line and eliminating the mismatch, 151 pairs of fine matches distributed on 11 cotton pieces are obtained, and the camera inner and outer parameters are obtained by the combination of calibration and F. Finally, the 3D coordinate of cotton point cloud was reconstructed, and the result showed that the reconstruction result of Z axis was close to manual measurement, and the average error was 0.039 3 m, which could reflect the relative position of cotton.
【作者单位】：南京农业大学工学院;
【分类号】：TP391.41
【正文快照】： 朱昒杰,朱颖汇,王玲,卢伟,罗慧,张志川.基于尺度不变特征转换算法的棉花双目视觉定位技术[J].农业工程学报,2016,32(6):182-188.doi:10.11975/j.issn.1002-6819.2016.06.025 http://www.tcsae.orgZhu Rongjie,Zhu Yinghui,Wang Ling,Lu Wei,Luo Hui,Zhang Zhichuan.Cotton pos

【相似文献】