Obstacle Avoidance Sensor Research Based On Infrared Binocular Vision

The obstacle avoidance sensor is one of the important subsystems in intelligent security equipment.With the miniaturization and production of thermal infrared cameras,related technologies for obstacle avoidance sensors based on infrared binocular vision have been developed.It makes up for the deficiency that visible light cameras are hard to work effectively in special environments such as night.In this paper,the image preprocessing,camera calibration,epipolar correction,and stereo matching of infrared binocular vision are studied,and a set of obstacle avoidance sensor systems is implemented.The preprocessing of infrared image in binocular stereo vision is studied.For the non-uniformity of the original image,two-point correction is used to make the image clear;for the high dynamic range of the original image,bilateral filtering and dynamic range compression are used to effectively retain the image details while compressing the amount of image data.The calibration of infrared binocular system is studied.According to the different reflectance of different materials,a circular lattice target is designed which is suitable for the infrared camera calibration.It is compared with the checkerboard target.Experiments have found that both effects can meet the calibration requirements.The circular lattice target has slightly better effect.The camera model of the lens during focusing and focusing is analyzed,and calibration experiments are performed during the lens focusing process,which proves that the camera of this type of lens needs to be calibrated in multiple stages to obtain its complete camera model.The epipolar correction of infrared binocular images is studied.Aiming at the situation that the real scene image does not meet the epipolar constraint,the Bouguet algorithm is used to correct the image based on the result of camera calibration.Taking a corner point as the observation point,it can be seen that the difference in the row direction decreases from 18 pixels to 1 or 2 pixels,and the correction effect is excellent.The stereo matching algorithm of infrared binocular image is studied.The semi-global matching algorithm which is used in visible binocular vision gives consideration to both effects and efficiency.However,due to the weak texture and high noise of the infrared image,the disparity map generated by SGM is rough.The edges are not clear,and there are a lot of holes inside.In this paper,the disparity map repair algorithm based on fast bilateral space is used to repair the rough disparity map.It is proved by experiments that a higher quality infrared disparity map can be obtained.In this paper,an obstacle avoidance sensor system based on infrared binocular vision is built.Disparities are obtained based on the results of binocular camera calibration and stereo matching.An experimental study of passive ranging of targets is carried out.An experimental study of passive ranging of targets is carried out.The actual ranging results prove that within a short distance of about 1m,the ranging error of the system is controlled within 10% and has a good resolution.It can meet the needs of practical applications and realize the effect evaluation of obstacle avoidance function.