Size estimation of sweet onions using consumer-grade RGB-depth sensor

作者：Wang Weilin; Li Changying^*

来源：Journal of Food Engineering, 2014, 142: 153-162.

DOI：10.1016/j.jfoodeng.2014.06.019

• Size estimation
• Volume
• RBG-D
• Kinect
• Machine vision
• Phenotyping

Size estimation is an important aspect of the postharvest handling of onions. This study applied the RGB-depth (RGB-D) sensor to measure the maximum diameter and volume of sweet onions, and estimated the density of onions using measured parameters. RGB-D images were acquired when onions were placed at six different orientations. The maximum diameter was calculated using both the color and depth images. The volume was estimated using the depth images. The onion diameter estimated by depth images achieved a higher average accuracy and robustness (RMSE = 2 mm) than those calculated by color images (RMSE = 3.4 mm). The predicted volume of onions showed a RMSE of 18.5 cm(3) and an accuracy of 96.3%. Results also demonstrated that it is promising to nondestructively estimate the onion density based on its depth image. The proposed methods can be applied to improve the efficacy and efficiency of size estimation in onion phenotyping and postharvest sorting/grading.

• 出版日期2014-12