| The high-quality temporal and spatial distribution of light in soybean canopy is of great significance to increase the yield and quality of the seeds.As machine vision perception systems and artificial intelligence technology are widely used in smart agriculture,the response law between soybean canopy plant type and light distribution has become a hot and difficult research field in this field.In this study,soybean canopy in cold regions was taken as the research object,a multi-source image acquisition platform based on Kinect 2.0 sensor was built to calculate the morphological structure of the soybean canopy based on the three-dimensional(3D)reconstruction,combining the color image and the depth image of soybean canopy.Based on the time series of soybean petiole angle,important parameters were calculated including plant height,leaf area index and canopy plant type parameters as well as light distribution law within the canopy.The main research contents are as follows:(1)A rapid,accurate and non-destructive synchronous image acquisition system of soybean canopy based on Kinect 2.0 sensor was constructed.First,soybean canopy was taken as the research object,the color image and depth image of the soybean canopy were obtained by the heterogeneous image synchronous acquisition platform.Second,the characteristics of soybean the multi-source image at the different growth stages(branching period,flowering period,knotting period,drum period)were analyzed,which provides an effective technical basis for accurate three-dimensional reconstruction of soybean canopy morphology structure and calculation of plant type parameters.(2)The calculation method of soybean canopy parameters based on 3D reconstruction was established.First,the changes of the stems,leaf,bud and pod of soybean canopy with time at different growth stages were analyzed.K-means clustering segmentation algorithm was used to identify the color image region of soybean canopy.The skeleton morphology of soybean canopy was extracted using improved morphological refinement method.The branch points and endpoints of the canopy skeleton were detected by the optimized corner extraction algorithm,and the registration method of canopy depth image and corresponding color image was established based on affine transformation algorithm,and then the morphological structure of soybean canopy skeleton with color information was reconstructed in 3D space.Second,on the3 D skeleton structure of soybean canopy,the leafstalk angles were calculated in different stages.The extreme value of the soybean canopy skeleton height was obtained using the automatic corner detection algorithm,and then the soybean plant height was calculated.The soybean canopy width was calculated using 3D point cloud projection method.The calculation method of leaf area index based on 3D point cloud segmentation method was constructed by effective pixel proportion algorithm.Finally,the calculation method of soybean petiole angle,plant height,crown width,leaf area index and other plant type parameters was realized based on 3D reconstruction.(3)The relationship between plant type parameters of soybean canopy and light intensity distribution within the canopy was studied.First,the light intensity of multiple areas was simultaneously measured based on the plant type parameters of soybean canopy by multi-channel light sensors.Second,the variation law of light intensity in soybean canopy was analyzed at different growth stages(branching period,flowering period,knotting period,drum period,etc.).Third,the variation law of plant type parameters(leafstalk angle,plant height,leaf area index and crown width)was proved at different stages.In addition,the relationship between plant type and light distribution characterized by 3D images of soybean canopy was analyzed at different growth periods,which provided theoretical basis and technical support for scientific regulation of soybean canopy structure to obtain high-quality light spatial and temporal distribution. |
PDF Full Text Request