Position detection with high precision using novel compound-eye sensor

A new compound eyes model is designed to track the fast-moving object and detect its position with high precision in complex background. The compound eyes that consist of several eyelet channels have good nature of wide fields of view and high update rate for vision systems. Adjacent eyelet fields of view have certain degree of overlapping for detecting 3D coordinates of an object. Gradient lenses, instead of fiber bundles, are used below each eyelet lens in order to transfer eyelet image onto a re-image structure in simulation. And the eyelet image is projected onto a Charge Coupled Device (CCD) detector array through the gradient lens, avoiding expensive fiber bundles and complex assembly. Besides, simulation and analyze of the compound eyes model is performed. Neural network calibration algorithm has been adapted to build the relationship between the object points and corresponding image points for each eyelet. This calibration algorithm provides a highly accurate prediction of object data points from their corresponding image points. After getting the intrinsic and extrinsic parameters of each eyelet channel, 3D coordinates of an object can be calculated from its image points. Preliminary experimental results for Neural Network calibration are presented and evaluated, showing residual errors between actual and predicted direction angles of around 10-3~10-4 rad. And errors between actual and calculated coordinates of position are within 3%.

[1]  Y. Aloimonos,et al.  Complete calibration of a multi-camera network , 2000, Proceedings IEEE Workshop on Omnidirectional Vision (Cat. No.PR00704).

[2]  P. S. Krishnaprasad,et al.  Steering laws for motion camouflage , 2005, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[3]  Isao Shimoyama,et al.  A small-sized panoramic scanning visual sensor inspired by the fly's compound eye , 1998, Proceedings. 1998 IEEE International Conference on Robotics and Automation (Cat. No.98CH36146).

[4]  Marc Levoy,et al.  High performance imaging using large camera arrays , 2005, SIGGRAPH 2005.

[5]  Richard Hornsey,et al.  Calibration techniques for object tracking using a compound eye image sensor , 2004, SPIE Security + Defence.

[6]  N. Franceschini,et al.  From insect vision to robot vision , 1992 .

[7]  P. Eggenberger,et al.  Evolving the morphology of a compound eye on a robot , 1999, 1999 Third European Workshop on Advanced Mobile Robots (Eurobot'99). Proceedings (Cat. No.99EX355).

[8]  Richard Hornsey,et al.  Electronic compound-eye image sensor: construction and calibration , 2004, IS&T/SPIE Electronic Imaging.

[9]  Keyi Wang,et al.  Fabrication of aspherical liquid lens controlled by electrostatic force , 2007, International Symposium on Advanced Optical Manufacturing and Testing Technologies (AOMATT).

[10]  M. Levoy,et al.  Light field microscopy , 2006, SIGGRAPH 2006.

[11]  J. Neumann,et al.  Multi-camera networks: eyes from eyes , 2000, Proceedings IEEE Workshop on Omnidirectional Vision (Cat. No.PR00704).