A Monocular Vision Sensor-Based Obstacle Detection Algorithm for Autonomous Robots

Abstract

This paper presents a monocular vision sensor-based obstacle detection algorithm for autonomous robots. Each individual image pixel at the bottom region of interest is labeled as belonging either to an obstacle or the floor. While conventional methods depend on point tracking for geometric cues for obstacle detection, the proposed algorithm uses the inverse perspective mapping (IPM) method. This method is much more advantageous when the camera is not high off the floor, which makes point tracking near the floor difficult. Markov random field-based obstacle segmentation is then performed using the IPM results and a floor appearance model. Next, the shortest distance between the robot and the obstacle is calculated. The algorithm is tested by applying it to 70 datasets, 20 of which include nonobstacle images where considerable changes in floor appearance occur. The obstacle segmentation accuracies and the distance estimation error are quantitatively analyzed. For obstacle datasets, the segmentation precision and the average distance estimation error of the proposed method are 81.4% and 1.6 cm, respectively, whereas those for a conventional method are 57.5% and 9.9 cm, respectively. For nonobstacle datasets, the proposed method gives 0.0% false positive rates, while the conventional method gives 17.6%.

Fast Obstacle Detection for Monocular Autonomous Mobile Robots

“…Although this method achieves high segmentation accuracy, further improvement is required for more challenging cases. As reported in [12], their algorithm produces good results when the camera is located at about 60 mm or higher. However, for much lower camera settings, it produces relatively unstable results due to stronger perspective distortion of the warped image.…”

“…This detect-then-segment scheme is inspired by the state-ofthe-art work of Lee et al [12]. Compared to the method of Lee et al, our algorithm achieves much faster computation with comparable detection accuracy.…”

“…In this paper, we use the same assumptions as [12]. We assume that the ground is relatively flat and there are no overhanging obstacles in the environment.…”

“…The camera height is 34.5 mm above the ground. This is a more challenging camera setting than in the state-of-the-art method [12]. Wheel encoders and a gyroscope are also installed for acquiring wheel odometry.…”

“…This setting reduces the amount of disparities and makes robust point tracking difficult. Lee et al [12] proposed a monocular vision based approach for small robots using IPM based coarse detection and MRF based fine segmentation. To the best of our knowledge, this is the first algorithm designed for "thin" robots like robotic vacuum cleaners.…”