The proliferation of illegal landfills represents a critical environmental problem that affects both ecosystems and public health. In this work, an innovative approach based on multi-agent systems (MAS) is proposed for automated landfill detection using image pattern recognition techniques. The system is composed of specialized agents that collaborate to preprocess visual data, extract relevant features and classify potentially contaminated areas from satellite and aerial images. Computer vision and deep learning algorithms are used to train detection models that allow agents to identify affected areas with high accuracy. The distributed architecture of the system facilitates scalability and adaptation to different geographical contexts. Experimental results demonstrate the effectiveness of the proposed approach, highlighting its potential as a support tool for environmental management and decision making in sustainability policies. This work contributes to the development of intelligent solutions for environmental monitoring through the synergistic integration of artificial intelligence and geospatial technologies.