Purpose The use of unmanned aerial vehicles (UAVs) has significantly increased in the past decade and nowadays is being used for various purposes such as image processing, cargo transport, archaeology, agriculture, manufacturing, health care, surveillance and inspections. For this reason, using the appropriate image processing method for the intended use of UAVs increases the study's success. This study aims to determine the most suitable one among the innovative methods that constitute the image processing system for a UAV to be used for surveillance purposes. Design/methodology/approach Analytical hierarchy process has been used in the solution of the decision problem to be handled in three stages, namely, platform, architecture and method. The most suitable alternative and the effect weights of these criteria results were determined at each stage. Findings As a result of this study, Jetson TX2 was determined as the most suitable embedded platform, ResNet is the optimum architecture and Faster R-convolutional neural networks was the best method in the image processing layer for a system that will provide surveillance with image processing method using UAV. Practical implications In UAV designs, where multiple hardware and software choices and system combinations exist, multi-criteria decision-making (MCDM) approaches can be used as a system decision mechanism. Originality/value The novelty of this work comes from the application of MCDM methods that are used as a multi-layered decision mechanism in UAV design.