نوع مقاله : کامپیوتر - محاسبات نرم و هوش مصنوعی
نویسندگان
1 دانشجوی دکتری، دانشگاه همدان ،همدان، ایران
2 پژوهشگر،دانشگاه پدافند هوایی خاتم الانبیاء(ص)،تهران، ایران
3 دانشیار،دانشگاه ملایر،ملایر،ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
In recent years, the very rapid development of technology in the field of UAVs (Unmanned Aerial Vehicles), along with its advantages, has brought serious threats at various social and security levels. Among these problems, the issue of unauthorized flights in protected and security area can be mentioned. Therefore, real time detection of these devices is necessary in order to quickly carry out relevant measures. In this regard, in this research, using the YOLOv5l algorithm, which is part of the latest version of one-stage computer vision algorithms, two models with SGD and Adam optimizers have been developed for the real time detection of UAVs. To develop the models in this research, a dataset containing 10046 images of different types and states of UAVs has been used. The processing of the models has been done with the Google Colab platform, which provides a free powerful processing system for developers. The models have been evaluated on four test sets of 1000 images including normal, low volume, night mode, and gray scale and also a test set including 100 images from several UAVs. According to the results, the developed model with The Adam optimizer performed better than the model developed with the SGD optimizer.
کلیدواژهها [English]