Morphological measurement for carrot based on three-dimensional reconstruction with a ToF sensor

Abstract

The three-dimensional (3D) shape information of carrot is vital for carrot grading and phenotyping analysis, which cannot be obtained accurately based on the two-dimensional (2D) image lacking depth information. Therefore, a morphological measurement method for carrot was proposed based on 3D reconstruction. The RGB-D acquisition system was composed of a Time-of-Flight (ToF) sensor and a turntable pasted with circle markers. 16 RGB and 16 depth images were captured by the Kinect sensor from different views to cover the whole carrot surface. The registration errors of point clouds from different views concentrated within 2.4 mm, and most were within 1 mm. The morphological variables (volume, length, and maximum diameter) of 136 carrots were obtained from the 3D model generated by the Poisson reconstruction method. The MAPEs between actual morphological variables and those obtained from the 3D model were all below 3%. The proposed method can be employed as a low-cost, accurate, and robust method for 3D reconstruction and morphological measurement of fruit and vegetables with few surface features.