METODE STEREO VISION DENGAN KAMERA CMOSUNTUK PENGUKURAN JARAK

Abstract

Excess image sensor digital camera is not just a picture taker or produce video, can be also as a measurement tool. Through the method of image processing, can be processed input image or an object taken from the camera. Image processing can be thresholding, edge detection and template matching. Required sensing with two images obtained from the stereo camera simultaneously or stereo vision. The camera used has properties or specifications are identical. OpenCV was contribute to image processing, object interpretation and also the calibration process. Calibration function to search for intrinsic and extrinsic parameters of the camera. Research uses a solid object in the form of balls of various sizes and colors. Epipolar and stereoscopic as the method to be used, because in previous studies, it is very suitable for processing images with color segmentation. As for digital image processing using the method thresholding. Research results obtained, that the object of pingpong ball motionless when lighting using sunlight has RMSE (root mean square error) of 2.64% while the value of a tennis ball RMSE is 2.53%. Research by lighting using indoor lighting, RMSE was 1.03% pingpong ball, tennis ball RMSE of 2.25%. While on the moving object, either RMSE pingpong ball or a tennis ball, its value is greater than when the motionless object. Measuring tool using a stereo camera capable of detecting targets up to a distance of 300 cm, but at a distance of 200 cm level of confidence does not the 95%. Lighting greatly affect the measurement results CMOS camera, because scalar target color easily changed.

Keyword: image processing, thresholding, OpenCV, epipolar, stereo cameras, stereoscopic