Deep Convolutional Neural Networks for Weeds and Crops Discrimination From UAS Imagery

Abstract

Weeds are among the significant factors that could harm crop yield by invading crops and smother pastures, and significantly decrease the quality of the harvested crops. Herbicides are widely used in agriculture to control weeds; however, excessive use of herbicides in agriculture can lead to environmental pollution as well as yield reduction. Accurate mapping of crops/weeds is essential to determine weeds’ location and locally treat those areas. Increasing demand for flexible, accurate and lower cost precision agriculture technology has resulted in advancements in UAS-based remote sensing data collection and methods. Deep learning methods have been successfully employed for UAS data processing and mapping tasks in different domains. This research investigate, compares and evaluates the performance of deep learning methods for crop/weed discrimination on two open-source and published benchmark datasets captured by different UASs (field robot and UAV) and labeled by experts. We specifically investigate the following architectures: 1) U-Net Model 2) SegNet 3) FCN (FCN-32s, FCN-16s, FCN-8s) 4) DepLabV3+. The deep learning models were fine-tuned to classify the UAS datasets into three classes (background, crops, and weeds). The classification accuracy achieved by U-Net is 77.9% higher than 62.6% of SegNet, 68.4% of FCN-32s, 77.2% of FCN-16s, and slightly lower than 81.1% of FCN-8s, and 84.3% of DepLab v3+. Experimental results showed that the ResNet-18 based segmentation model such as DepLab v3+ could precisely extract weeds compared to other classifiers.

Explicitly Incorporating Spatial Information to Recurrent Networks for Agriculture

“…For strawberry detection [19] used a multi RBG-D camera array and fused the RGB and LAB images in a RetinaNet architecture for promising results. In 2020, [20] proposed a modified U-Net [21] architecture with separate pathways for weed and crop segmentation and, as well as [22], compared results with variants of the popular SegNet [23] and DeepLabV3 [24] models in different agricultural scenarios. Both found that the performance comparisons are strongly dependent on the data domain.…”

“…This is because of its simplicity in terms of layer operations, which requires little work to incorporate our spatial-temporal layers. However, considering the variability in performance of segmentation architectures in agriculture [20], [22] we also compare to DeepLabV3 on both datasets.…”

Weed Detection from Unmanned Aerial Vehicle Imagery Using Deep Learning—A Comparison between High-End and Low-Cost Multispectral Sensors

“…In order to meet the growing demand for food in the future, innovative solutions that increase the efficiency of agriculture and are accessible to everybody are needed [ 1 ]. Weeds are the primary cause of yield reduction [ 2 ], as they compete with crops for nutrition, water, sunlight, and space [ 3 , 4 ]. The use of herbicides is the most common and efficient tool to control weeds but leads to irreversible ecological damage such as groundwater pollution, soil contamination, and biodiversity loss [ 5 , 6 , 7 ].…”

Overlapped tobacco shred image segmentation and area computation using an improved Mask RCNN network and COT algorithm

“…In recent years, machine vision-based deep learning methods have provided advanced and efficient image processing solutions in agriculture. Deep learning methods, combined with machine vision technology, have been widely used in plant disease and pest classification, including the classification of fresh tobacco leaves of various maturity levels ( Chen et al., 2021 ); the classification of tobacco plant diseases ( Lin et al., 2022 ); the classification of wheat spike blast ( Fernández-Campos et al., 2021 ); the classification of rice pests and diseases ( Yang et al., 2021 ); the detection of plant parts such as tobacco leaves and stems ( Li et al., 2021 ); the detection of tomato diseases ( Liu et al., 2022 ); the detection of wheat head diseases ( Gong et al., 2020 ); the detection of brown planthoppers in rice ( He et al., 2020 ); plant image segmentation, such as tobacco planting areas segmentation ( Huang et al., 2021 ); field-grown wheat spikes segmentation ( Tan et al., 2020 ); rice ear segmentation ( Bai-yi et al., 2020 ; Shao et al., 2021 ); rice lodging segmentation ( Su et al., 2022 ); photosynthetic and non-photosynthetic vegetation segmentation ( He et al., 2022 ); weed and crop segmentation ( Hashemi-Beni et al., 2022 ); and wheat spike segmentation ( Wen et al., 2022 ). Deep learning methods combined with machine vision technology have been utilized in research focused on the classification of tobacco shred images.…”