Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract: Deep learning and Machine Vision has demonstrated excellent capabilities for learning image features and is widely used for image target detection. In order to improve the performance of machine vision in peach tree detection for weeding robot, Faster Region Convolutional Neural Network (Faster-RCNN) was introduced. Resnet100 was adopted as backbone network, combined with the Feature Pyramid Network (FPN) architecture for feature extraction. The Region Proposal Network (RPN) was trained and region proposals for each feature map were created. Peach tree detection results of 1000 test images showed that the average detection precision rate was 86.41%, the recall rate was 92.15% and the mean intersection over union (MIoU) rate for instance segmentation was 85.45%. The performance of the Weeding robot tests show that the rate of missed cuts is greatly reduced, compared to traditional weeding methods.