IP3 2019-eCommerce 1
This lot is generally related to object identification and classification. Techniques are disclosed to differentiate objects detected by object identification system using an artificial intelligence or computer vision algorithms by building a histogram of pixel intensity values in an image’s region of interest, locate maxima and minima in histogram and generate a binary image. The techniques further assign a contour score to each pixel intensity value, segment region of interest using two contour scores and classify the object. Few techniques perform edge analysis on detected object regions or analyze depth images to classify the object. Also disclosed are techniques to detect small objects located at far distances from a camera having a limited field of view using planar or nearly planar surfaces, vehicle collision avoidance systems that use only a vision based stereo configuration to reduce overall cost and improve performance in vehicle detection and tracking. The technology may be implemented in object identification systems, object tracking or monitoring systems, collision detection systems, autonomous vehicle software, artificial intelligence-based object detectors, etc.