Open AccessInformation acquisition of anterior branches of fruit based on Mask R-CNN
Author(s) -
Zhuo Wang,
Lihua Xiong,
Liao Haishen,
Kang Xilong,
ChengHong Yang
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1345/4/042083
Subject(s) - centroid , deflection (physics) , standard deviation , mathematics , artificial intelligence , computer science , computer vision , geometry , statistics , optics , physics
In order to obtain the deflection information of the anterior branch to enable the end effector to perform citrus picking in the correct posture, this paper proposes a method based on Mask R-CNN model and centroid estimation for pre-fruit branch deflection information acquisition. The method marks the front of the fruit branches by means of a rectangular-like marking method, and completes the identification of the pre-fruit branches with random posture. The ROI is obtained by the discrete mask obtained for the model, The ROI area is equally divided into two sub-areas for centroid extraction, At last, The two centroids obtained are used to calculate the deflection angle of the fruiting branches. The experimental results show that the recognition accuracy of the model under the test set is 96.44%, the average recall rate is 80.71%, the average deviation of the deflection angle of the pre-fruit branches is 7.4°, and the maximum deviation of the angle is 12°.