Panoramic Measurement System On Mobile Robot For Greenhouse Plants Morphological Information Collection

With the development of modern precision agriculture,plant morphology information has become an important decision-making basis for modern precision agricultural production management in the aspects of plant growth environment control,molecular breeding,fertilizer and water management,plants yield prediction and statistics,etc.The realization of accurate online non-destructive plant morphology measurement was of great significance to promoting the development of modern agriculture in China.At present,the development of intelligent facilities for greenhouse plant cultivation in China was still in the initial stage,and it was impossible to achieve accurate measurement of greenhouse plant morphology information in a large-scale and multi-regions.Therefore,according to the structural characteristics of Chinese greenhouse,a mobile plant morphology information measurement platform was designed to achieve panoramic measurement of greenhouse plant colony morphology information.The main contents and conclusions were as follows:(1)A mobile plant morphology measurement platform was built.According to the spatial structural and environmental characteristics of greenhouse,ROS(robot operating system)robot platform based on SLAM(simultaneous localization and mapping)vision technology was selected.Through autonomous positioning and navigation,carrying acquisition system sequentially achieved each plant’s panoramic measurement of morphological information.(2)A three-dimensional(3D)point cloud implementation scheme for greenhouse plants based on Kinect v2 was proposed.The experiment object was line pepper,and the original color and depth images was captured by the Kinect v2.After alignment matching of these two,an RGB-D(red-green-blue-depth)image with the color and depth information was attained and a 3D point cloud was obtained by 3D coordinate conversion.The point cloud model of a single line pepper was reconstructed by data denoising and image segmentation.(3)An extracting system of greenhouse line peppers morphology based on the mode of point cloud was designed.A variety of morphological parameters,such as plant height,plant width,point cloud amount,point cloud surface area,point cloud volume,and projected area on the plane YOZ were extracted.The plant height and width were calculated based on the minimum external bounding box of the overall point cloud.The system calculated point cloud amount through the accumulation of line peppers.Point cloud surface area and volume were calculated by constructing point cloud space polyhedron.The projection area was calculated by constructing the 2D sealing area enclosed by the boundary of YOZ convex points set.(4)The experimental verification of the extracted line pepper morphological parameters showed that the plant height and width were linear fitting correlated with the artificial reference value,and the correlation coefficient R2 for plant height and width was 0.9883 and 0.9843,respectively.Point cloud amount,surface area,volume and projection area were used to establish regression models with fresh weight,dry weight and volume in this study of biomass.R2 of point cloud amount and fresh weight,dry weight and volume of the three index parameters for biomass were 0.9429,0.9452 and 0.9383.R2 of surface area and three index parameters of biomass were 0.9658,0.9700 and 0.9643,respectively.R2 of volume and three index parameters of biomass were 0.9430,0.9453 and 0.9301,respectively.R2 of YOZ planar projected area and three index parameters of biomass were 0.9573,0.9515 and 0.9602,respectively.(5)A 3D point cloud model of line pepper colony was generated by extracting the spatial growth position and direction angle based on artificial interaction method.Spatial information and morphological parameters were combined to generate a panoramic image for morphological information of line pepper colony.Panoramic measurement of the 3D line pepper cloud morphological information was tested and verified.The data showed that the MRE(mean relative error)of the plant colony height and width were 1.24%and 1.15%,respectively.Using 24 line peppers as samples,biomass was estimated by point cloud amount,and MRE of colony biomass was 4.13%,3.85%and 3.15%,respectively.In this study,a panoramic measurement method of greenhouse plants morphological information in a large-scale and multi-regions was proposed,which was of great significance to the intelligent information collection,decision-making and management of modern greenhouse in China.