Researchers at Embrapa Instrumentation (SP) have developed a method that uses image sensors and artificial intelligence to detect the cartridge caterpillar, one of the most destructive pests in corn cultivation. The system analyzes digital images to identify the caterpillar both on leaves and corn cobs, reducing reliance on traditional, labor-intensive, and subjective observation methods.
Corn is among the world’s most widely cultivated cereals and the cartridge caterpillar can cause losses of up to 70% of production, according to Embrapa researchers. The caterpillar attacks crops during both vegetative and reproductive stages.
A Promising Alternative
The methodology was detailed in the journal Electronics in the article Computational Intelligence Approach for Fall Armyworm Control in Maize Crop by Alex Bertolla and Paulo Cruvinel. The authors noted that discrepancies between current detection methods and desired results motivated the development of an alternative for early pest detection in cultivated areas, according to a press release.
Bertolla explained that the study focuses on recognizing and classifying dynamic patterns of the cartridge caterpillar, which attacks not only corn but also other crops such as soybeans and cotton. The approach can provide agronomists and laboratories with more accurate pest detection and monitoring tools.
To simplify image capture, a basic camera can be attached to agricultural implements to record caterpillar presence while working in the field. The camera need not be costly — it only needs to produce sufficiently high-resolution images for effective analysis.
Process Integration
The method combines digital image and signal processing, multivariate statistics, machine learning, and computer vision. “Machine learning enables systems to learn from custom training data to automate and solve related tasks. Within this framework, deep learning is a specialized branch of machine learning that uses artificial neural networks composed of purpose-specific layers,” explains Cruvinel.
Bertolla adds that the computational algorithm can analyze digital images of the caterpillar at various growth stages on corn plants, determining both its development stage and frequency of occurrence in the field. The program was developed in Python, a high-level programming language widely used in data science and machine learning applications.
Basis for Future Studies
Cruvinel and Bertolla note that the quality of the results supports the recommendation to use a deep learning–based structure. The data were evaluated for accuracy, precision, processing time, and hardware performance within the proposed scenario.
“The method also delivered promising results in terms of hardware performance and processing time, which could support future efforts to develop an embedded version of this technology for direct integration into agricultural implements,” Cruvinel explains.
For future work, they suggest incorporating additional artificial intelligence techniques for pest detection and pattern classification, including unsupervised learning approaches. They also propose using a multispectral camera mounted on an Unmanned Aerial Vehicle (UAV) to enable real-time monitoring and detection in the field.
