摘要：Crop breeding is the focus of global agricultural hi-tech companies and requires phenotypes screening after growing and cultivating. However, rice phenotypic traits are not easily obtained with high-throughput in the field due to the complexity of unstructured environment. This paper presents a design solution of the field robot to construct visual space for measurement of rice traits. The truss system equipped with manipulator modules and vision system locates the object by the position information recorded in the phase of growing seedlings. 3 manipulators including the rice-separating manipulator, the height measuring manipulator and the panicle-expanding manipulator were designed by imitating people’s actions such as pushing adjacent rice, expanding the panicle and rubbing rice long in order to reduce the overlap. Combined with them, 3 imaging sensors including the CCD camera, the structured light sensor and the laser sensor were introduced to measure a phenotype quantitatively. The simulations were designed to obtain workspaces of manipulators. The results showed that the volume of reachable space of clapboards of the rice-separating manipulator was 1.6 × 10-3 cubic meters, the effective movement of the height measuring manipulator was 1300mm and the maximum work area of the panicle-expanding manipulator was 32500 square millimeters. The results are consistent with those of manual operations. In conclusion, this paper describes an effective and reliable approach to acquiring phenotypic traits with high-throughput in the field.
