Object detection is known as a computer vision technique used to identify objects in images or videos. Object detection is performed on images of predefined objects or real-time images with the help of the programs used. In this study, object detection was performed by image processing using underwater images. Open cv resources and python are used for programming. The bullseye operating system is used as the operating system for Raspberry pi, and since the program is set according to this operating system, it may give some errors when trying to make the same program with other operating systems. Two different objects were introduced in the study. These are fish and sunken ship, and four different images are used for object identification. During the presentation, real-time images were taken from the webcam, the image was converted to gray, and then the object whose name matched the image previously added to the library was introduced. It is aimed to reduce the processing power of the raspberry pi and the loss of time during processing by turning the received images to gray. In the next step, the names in the 'coco.names' library are compared with the name of the object retrieved from the webcam. The object is defined when the image matches after comparison. The defined picture is framed and displayed as a printout with the object name. SSD deep learning method and MobileNet artificial neural network are used for object recognition. There are multiple methods of object detection.
Anahtar Kelimeler: Underwater Image Processing, Deep Learning, Artificial Neural Networks