Teleoperation enables humans to perform tasks in complex and dangerous environments more safely and comfortably. Current vision-based methods for non-anthropomorphic robotic teleoperation focus mainly on tracking and imitating human motion to control the robot’s end-effector. However, in complex environments, these methods that lack direct control over other robotic joint configurations may encounter obstacle avoidance issues. Therefore, we proposed a method for generating human-like joint configurations in vision-based non-anthropomorphic robotic teleoperation. This method captures human motion data using only a monocular RGB camera and translates it into robotic motion. Specifically, a novel inverse kinematics solver is developed to calculate the robot’s human-like joint configuration. Through simulation experiments, the human-likeness of the robot’s posture and the tracking performance of its end-effector were analyzed, validating the effectiveness of the proposed method.