Visual Odometry for Pixel Processor Arrays

Laurie Bose, Jianing Chen, Stephen J. Carey, Piotr Dudek, Walterio Mayol-Cuevas; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 4604-4612


We present an approach of estimating constrained motion of a novel Cellular Processor Array (CPA) camera, on which each pixel is capable of limited processing and data storage allowing for fast low power parallel computation to be carried out directly on the focal-plane of the device. Rather than the standard pipeline involved with traditional cameras whereby whole camera images are transferred to a general computer system for processing, our approach performs all computation upon the CPA itself, with the only information being transfered to a standard computer being the camera's estimated motion.This limited data transfer allows for high frame-rate processing at hundreds of hz while consuming less than 1.5 Watts of power.The current implementation is restricted to the estimation of the camera's rotation in yaw and pitch, along with a scaleless estimate of the camera's forward and backward translation. We describe methods of image alignment by gradient descent, edge detection, and image scaling, all of which are performed solely on the CPA device itself and which form the core components of detecting camera motion.

Related Material

[pdf] [video]
author = {Bose, Laurie and Chen, Jianing and Carey, Stephen J. and Dudek, Piotr and Mayol-Cuevas, Walterio},
title = {Visual Odometry for Pixel Processor Arrays},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2017}