摘要
针对当前植物生长态势监测方法存在监测准确性差和能耗高的问题,提出基于立体视觉的虚拟植物生长态势监测方法。从不同角度获取被观测虚拟植物图像,依据三角测量理念得到图像中相应像素之间的视差,获取虚拟植物三维信息,完成虚拟植物三维重建。利用中值滤波算法处理重建图像,使用一个含有奇数个像素的窗口在植物图像上移动,在每个位置针对窗口中像素灰度值按照从小到大的顺序进行排列,将处在中间的灰度值当作窗口中心像素输出值,即滤波后图像。基于滤波后图像提取出植物高度、面积、形状、颜色和纹理五部分特征。依据科学特征系统分类原则,融合提取到的信息,构建一个植物多维特征树形体系。通过改进局部保持投影方式对所得特征体系进行优化,实现监测指标形式统一。利用计算机将所得植物生长特征信息以图表形式呈现出来,实现植物生长态势实时监测。实验结果表明,上述方法监测结果准确率高、监测能耗低。
Due to low monitoring accuracy and high energy consumption of current method, a method to monitor virtual plant growth situation based on stereoscopic vision was proposed. The images of observed virtual plant were obtained from different angles. According to the concept of triangulation, the parallax between the corresponding pixels in image and the three-dimensional information of virtual plant were obtained, so that the three-dimensional reconstruction of virtual plant was completed. The median filtering algorithm was used to process the reconstructed image and a window containing odd number of pixels moved on the plant image. Then, grayscale values of pixels in the window were arranged in order from small to big at each position. The grayscale value in the middle was taken as the center pixel output value of window, namely the filtered image. Based on the filtered image, five features of plant such as height, area, shape, color and texture were extracted. According to the classification principle of scientific feature system, the extracted information was integrated to construct a tree structure system with multi-dimensional features. In addition, the feature system was optimized by improving the locality preserving projections, so as to achieve the uniform monitoring indicator. Finally, the plant growth characteristic information was presented by the graphical form. Thus, the real-time monitoring of plant growth situation was achieved. Simulation results prove that the proposed method has high accuracy and low energy consumption.
作者
韩明阳
HAN Ming-yang(Jilin Engineering Normal University,Changchun Jilin 130000,China)
出处
《计算机仿真》
北大核心
2019年第11期365-369,共5页
Computer Simulation
基金
人参生长态势的虚拟化关键技术研究(XYB201811)
关键词
立体视觉
虚拟植物
态势监测
Stereoscopic vision
Virtual plant
Situation monitoring
