Scene understanding is a fundamental challenge for intelligent robots, especially for social robots, which are expected to have a human-like perception, comprehension, and knowledge. This paper proposes an approach to enable robots not only to detect objects in a scene but also to understand and reason the working environments. The proposed method contains three parts, which are object detection, object semantic comprehension, and feedback on robotic comprehension. Semantic comprehension is based on dictionary definitions of objects. The category, function, property, and composition of the detected objects are analyzed. These four elements are used to assist the robot in comprehending the target object in details. The experiment part of this paper discusses the applicability of the proposed method on robots.