Smart Detection Of Illicit Cannabis Plantations Using Remote Sensing Technology And Machine Learning

Remote sensing technology and machine learning classifiers can be utilized to develop smart detection systems for illicit crops such as Cannabis sativa L . Machine learning algorithms for classifying medium-resolution optical satellite data can be compared to identify the best model for enhancing law enforcement’s detection of illicit crops efficiently and accurately. Remote sensing-based smart detection systems have been developed in South America and Central Asia; however, these methods cannot be used effectively for Indonesia due to high cloud coverage, geographical differences, and the smaller area of Cannabis sativa L . plantations. This research developed an agile methodology that employs backpropagation neural networks to analyze the statistical growth phenology of cannabis derived from multitemporal medium-resolution remote sensing data. Using datasets derived from Indonesian law enforcement eradication records, the method achieved 94% accuracy and a kappa coefficient of 0.9. Further, plant growth phenology based on vegetation index values from multitemporal data was used to assess the condition of identified cannabis plantations.