This paper investigates a task-oriented environment modeling approach for service robots to bridge the gap between environment models and robotic task execution. Specifically, a novel environment modeling approach based on object fingerprinting is proposed to build an environment model with rich task-related object knowledge. Different from the existing model, the established model not only contains fingerprint information such as categories and attributes of task-related objects, but also contains a knowledge base for reasoning and updating stored objects. In order to enable service robots to quickly locate task-related objects in challenge scenarios, an inference mechanism based on timestamps and object distances is designed. Our proposal is extremely evaluated in real-world scenarios. Experimental results demonstrate the effectiveness and feasibility of the proposed approach, the constructed model is able to support the robot to accurately reason and locate the target objects in the occluded scenario.