AUTHOR=Li Zhiqiang , Xie Dongbo , Liu Lichao , Wang Hai , Chen Liqing 

TITLE=Inter-row information recognition of maize in the middle and late stages via LiDAR supplementary vision

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.1024360

DOI=10.3389/fpls.2022.1024360

ISSN=1664-462X

ABSTRACT=In the middle and late stage of maize, the light is dim and non-maize obstacles exist. When the plant protection robot uses the traditional visual navigation method to obtain the navigation information, there will be a missing of information. Therefore, this paper proposes a method of using LiDAR point cloud data to supplement machine vision for recognizing the inter-row information in the middle and later stages of maize. Firstly, we have improved the YOLOv5 algorithm based on the characteristics of the actual maize inter-row environment in the middle and late stages by introducing MobileNetv2 and ECANet, compared with YOLOv5, the frame rate of Im-YOLOv5 is increased by 17.91% and the weight size is reduced by 55.56% when the average accuracy is reduced by only 0.35%, which improves the detection performance and shortens the time of model reasoning. Secondly, we identify the obstacles (such as stones and clods) between the rows through the LiDAR point cloud data to obtain auxiliary navigation information. Thirdly, the auxiliary navigation information is used to supplement the visual information, so that not only the recognition accuracy of inter-row navigation information in the middle and late stage of maize is improved, but also the basis of the stable and efficient operation of inter-row plant protection robot is provided for these stages. Experimental results from a data acquisition robot equipped with camera and LiDAR sensor are presented to show the efficacy and remarkable performance of the proposed method.