Abstract:
Abstract: Plant height is a very important phenotypic trait in peanut breeding research. It is a key parameter not only indicating the growth state of peanut, but also calculating peanut biomass and yield. At present, the acquisition of peanut plant height in breeding research mainly relies on manual measurement, which is not only time-consuming and laborious, but also has certain subjectivity. Therefore, it is necessary to design a measurement system that can be used in the field and can quickly and accurately obtain the height of peanut canopy. In this study, a LiDAR-based field peanut canopy height information acquisition system was constructed, which was a mobile data acquisition platform designed for field conditions, and a data processing and analysis algorithm was developed to extract the height of peanut canopy. The sensor was equipped with a LiDAR (LMS291-05S, SICK) for scanning the peanut canopy, an RGB camera is used to capture the image of the peanut canopy, an encoder was used to record the moving distance of the system platform; all sensors were powered by a 24 V battery and all data were uploaded to a laptop. An experimental field was established with three peanut cultivars at Oklahoma State University's Caddo Research Station in Fort Cobb, Oklahoma state, USA in May and the data collections were conducted monthly from July to September 2015. There were 12 planting plots in the field, which were arranged in a straight line, and the length of each plot was 4.75 m, the interval between adjacent plots was 1.52 m, and the interval between ridges was 0.91 m in a plot. SWR, MCD and GA04S three different breeds of peanuts were planted in 12 different plots, each of which was repeated four times, and the plots of the same breed were not adjacent to each other. The ground-based LiDAR used for this research was a line-scan laser scanner with a scan-angle of 100(, an angle resolution of 0.25(, and a scanning speed of 53 ms. A wide aperture angle of 100( was used for LiDAR in order to ensure a complete scan of target canopy. As a result, the collected data included those from the adjacent rows. An algorithm was developed to extract the region of the interested data acquired by the system through the polynomial curve fitting method. To provide fixed reference points within each plot over the three collection periods, some metal posts were installed within the center length of each scanned row. These metal stakes were caught by the LiDAR in all the data file, in addition, noise also presented in the raw data. Therefore, a data filtering and correction algorithm was developed to eliminate the interference information. All valid canopy height data, which were processed according to the previously described preprocessing algorithms, were organized into a height matrix, that was, all the canopy height values scanned in each plot were constructed into a canopy height matrix, and then the mean heights were analyzed and calculated. The results showed that the minimum deviation of the average canopy height between obtained by the system and the manual measurement was 2%, the maximum deviation effected by topography was 32%, and the average deviation was about 11%, but the measurement deviation was gradually decreasing with the growth of peanut plants. The accuracy of this result was acceptable compared with the height of the peanut plant, and the collection of canopy height information by the system can greatly reduce working time and the input of artificial labor, and improve the efficiency of crop phenotypic information acquisition and analysis. Future research will focus on the rapid movement and manipulation of measurement system, and apply information fusion to data processing of multiple sensors.