Research on modern agricultural machinery: machine learning and artificial intelligence for smart agriculture
At present, artificial intelligence is widely used in various agricultural scenarios, including intelligent perception, real-time on-site monitoring, intelligent early warning, disease and pest detection, and intelligent decision-making in crop production environments. With the help of artificial intelligence, farmers can now detect if there are any diseases and pests, whether they need to use pesticides, and whether their plant protection practices are effective. This special edition focuses on several issues that still require further study and discussion, such as agricultural unmanned aerial vehicles, crop type mapping, crop phenotypic analysis, and identification of crop diseases and pests in sustainable and smart plant protection.
1. Agricultural unmanned aerial vehicles
Agricultural unmanned aerial vehicles (AUAVs) integrate robotics, artificial intelligence, big data, and the Internet of Things. They have been widely used in a variety of agricultural operations such as seed seeding, land monitoring, crop disease and pest detection, and pesticide and fertilizer spraying. AUAV greatly improves agricultural production efficiency and frees up labor.
They are becoming a new force in precision agricultural aviation. Compared to traditional agricultural machinery, they are small, lightweight, easy to transport, and have flexible flight controls. AUAV is characterized by precise operation, high efficiency, environmental friendliness, intelligence and ease of use. However, in many cases, real-time changes in AUAV load during flight can affect its speed, accuracy, and flight path stability.
Xu et al. proposed a flight dynamics model to achieve AUAV flight trajectory stability by using PID controller and powerful T-S fuzzy control method. The model can achieve a certain stability on the flight path to resist load disturbances with different mission requirements. With AUAV recorded crop growth data, farmers can analyze their crops and make informed decisions based on accurate crop growth information.
2. Crop type mapping
Large-scale and accurate CTM plays a key role in agricultural management, including crop monitoring in the field, optimizing crop distribution, and enabling agricultural intensification for sustainable development and food security. However, it is challenging due to factors such as crop diversity, interclass spectral similarity, and within-class variability.
Traditional CTM methods rely on remote sensing images (RSIs) as a data source, but the limited availability of cloud cover and optical images during critical crop growth can hinder RSI accuracy, especially in hot and rainy areas. In addition, irregular time series and limited coverage of remote sensing data complicate CTM.
To overcome these challenges, recent research has proposed deep learning-based CTM methods that outperform traditional machine learning methods, taking advantage of advances in Earth observation satellites and deep learning techniques. For example, Bian et al. designed a channel attention U-Net model that integrates shallow CNN, U-Net, and channel attention mechanisms to improve spectral feature extraction capabilities.
This approach can better handle inconsistent availability of remote sensing data due to clouds and rain. Future research should continue to focus on addressing this issue to enable large-scale CTM for precision agriculture management and macro control of food production.
3. Crop phenotypic analysis
Overall, crop phenotypic analysis (CPA) is an important tool for understanding the various factors that affect crop growth and providing timely data to crop managers. Traditional CPA methods rely on manual operations, which are time-consuming and labor-intensive, and analysis results can be unstable and inaccurate.
To overcome these challenges, machine vision and deep learning techniques can be used to quickly and accurately analyze crop phenotypic features. Zhang et al. proposed a three-stage multi-branch self-correcting feature estimation network (TMSCNet) for CPA, which can provide scientific basis for real-time monitoring of crop growth. In addition, seed morphological analysis is important for understanding the taxonomic relationships of various plant families and genera and for developing crop varieties with higher yields and better quality. In Seki et al., image-based phenotyping was used to develop a quantitative method to measure seed morphological characteristics through deep learning, even for small crop seed sizes. This approach can accelerate the discovery of the genetic basis of small morphological traits, such as seed size and shape.
4. Crop pest control
Crop disease and pest identification is a key aspect of agriculture that can help reduce pesticide use and increase agricultural productivity in a sustainable way. Due to the low recognition rate and weak generalization, traditional recognition methods such as support vector machines, naïve Bayesian and BP neural networks are not suitable for large-area pest and disease identification in this field.
In contrast, deep learning methods based on convolutional neural networks (CNNs) have achieved remarkable results and have strong generalizations. Due to the scarcity of images of crop pests and diseases, pre-trained VGG and ResNet 50 architectures based on the ImageNet dataset are often used. In order to improve the recognition accuracy of small insect targets, S-ResNet was built based on ResNet, and the recognition accuracy of Wang et al. increased by 7%.
Deep learning methods require powerful computing power and large training datasets, which makes them difficult to deploy on mobile devices. Future research efforts should focus on developing lightweight Siamese networks and incorporating other forms of data, such as geographic location, pest history, and weather trends, to improve the accuracy and reliability of pest identification systems.
