A method for locating soil monitoring points in greenhouse cultivation based on DeeplabV3

基于DeeplabV3的温室栽培土壤监测点定位方法

阅读:2

Abstract

Accurate positioning of water and fertilizer monitoring points in the soil of potted plants contributes to improving the level of precision agriculture management. This study proposes an automated extraction method for water and fertilizer detection points in blueberry potted plants in a greenhouse scenario. Based on the improved DeeplabV3+, a lightweight soil region segmentation network (LSRS-Net) is proposed. Firstly, the original DeeplabV3 + backbone was replaced by the lightweight MobileNetV2 to reduce computational complexity. Then, the Densely connected Atrous Spatial Pyramid Pooling (DenseASPP) module enhances multi-scale context feature fusion by embedding a Feature Pyramid Network (FPN) between the encoder and decoder to achieve cross-layer semantic fusion. In addition, the optimization algorithm based on Euclidean distance has improved the coordinates of the detection points. The experimental results show that, compared with the original model, the Mean Pixel Accuracy (MPA) and Mean Intersection over Union (MIoU) of the LSRS-Net model have increased by 2.98% and 3.84%, respectively, and the processing speed has reached 25.24 ms per frame. The extraction validity rate (EVR) exceeds 98.33%, which meets the requirements for fast and stable application in greenhouse cultivation and provides a viable solution for intelligent water and fertilizer management in agriculture.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。