Target fruit positioning segmentation method and system

• Main Authors: LU YUQI, JIA WEIKUAN, ZHANG WEIKE, WEI JINMENG, DING YANHUI
• Format: Patent
• Language: Chinese; English
• Published: 17.05.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a target fruit positioning segmentation method. The method comprises the following steps: acquiring a target fruit image; image feature extraction is carried out on an obtained target fruit image, and after feature fusion is carried out through FPN, a feature map is generated; according to the generated feature map, utilizing regression branches to generate category confidence and a regression frame, and further generating an initial segmentation mask by mask branch prediction; and obtaining position information of the target fruit and a rough mask segmentation result based on a Mask RCNN model according to the initial segmentation mask. According to the invention, through the backbone network and the post-processing module, the problems of low target fruit identification accuracy and rough mask edge in a complex orchard environment are solved, and the requirement of real-time operation of the picking robot is met. 本发明提供了一种目标果实定位分割方法，包括获取目标果实图像；通过对获取的目标果实图像进行图像特征提取，通过FPN进行特征融合后，生成特征图；根据