Visual-Inertial Odometry (VIO) estimate ego-motion using a camera and Inertial Measurement Unit (IMU). It shows outstanding performance for estimating ego-motion of a vehicle in absolute scale level thanks to the aid of the gyroscope and the accelerometer. However, in large-scale outdoor environments, the VIO has some difﬁculties in estimating translation because feature points along the motion direction and distant features points in the images cause degenerate situations. To solve these difﬁculties, we propose to infer the conﬁdence of feature points and to elegantly incorporate the conﬁdence to the Kalman ﬁlter based VIO. The conﬁdence is computed from motion direction and displacements of tracked feature points under our urban canyon prior, and it is applied in cases that camera is moving forward to measurement noise covariances of the Kalman ﬁlter for ego-motion estimation. Experimental results on the public KITTI dataset show that the VIO outperforms monocular and stereo visual odometries, and the proposed VIO with conﬁdence analysis achieves 1.82% translation error and 0.0018 deg/m rotation error.