Natural Landmark Feature Extraction and Memorization with Depth Camera

Robots must navigate unknown places efficiently. This requires scanning and memorizing the surroundings. These processes use a lot of time and memory. They may slow the robot's core tasks. This research aimed to develop a memorization algorithm. The researchers inspired the memorization algorithm by human cognitive processes. The algorithm used a two-wheeled robot with differential steering. Human-like object memorization has three stages. First, identify unique objects in specific contexts. Then, recognize these objects. Finally, memorize their features. The proposed algorithm was then added to a robot. It can now detect objects, mark them, and remember the object's features: ID, color, and position. In this experiment, we used a depth camera as the main sensor for object detection. We validated the research by having the robot do three learning and six testing experiments in the environment. This tested its ability to recall marked objects. Variables included the robot's viewpoint angle and environmental lighting conditions. The memorized data was then compared across tests. This assessed the quality and consistency of feature memorization. It compared the ID, landmark color, and pixel position of landmarks. Preliminary results showed the robot successfully memorized object features, with 86.95% accuracy. This research aimed to improve the robot's navigation and autonomy. It sought to do this with advanced object recognition and memory integration.