View source for About Elphel, Inc

<!--{{Manual Header}} -->
[[Image:Elphel_andrey_oleg_olga.jpeg|250px|thumb|left]]
Elphel is a technology company doing research and development in the field of high-performance digital cameras, image processing, 3D imaging and machine learning.
Elphel imaging systems are primarily used for for scientific applications that require designs to be user-modifiable at all levels - from the hardware and FPGA code to the system and application software. Since the start of the company in 2001, Elphel was adhering to the FLOSS practice for the code and now applies CERN Open Hardware License to all electronic boards and mechanical CAD files. Elphel cameras are used in many National Laboratories, and universities in USA, European Union and other countries.

Starting with Google Books and later Google Street View projects we introduced JP4 – JPEG-based (and so compatible with the standard libraries) image compression that preserves raw Bayer data of the image sensors. Raw Bayer mosaic is a preferred format for image processing including various types of end-to-end DNN.

Elphel cameras offer unprecedented combination of high performance and complete openness at all levels – from mechanical and circuit design, FPGA RTL code up to all levels of the system and application software. All the design files are publicly available at Elphel web sites and through the popular GitHub repository. These features make Elphel products attractive for scientific research and innovative products. Elphel technology has been referenced in over a hundred scientific publications, and at least six US patent applications that use or reference Elphel products.

Since 2012 Elphel has been developing methods of precise camera calibration and designing photogrammetric multiple-view cameras with thermally compensated SFE resulting in 0.05 pix reprojection error for 5 MPix sensors. We developed modified checkerboard pattern where each edge consists of 2 arcs - this modification from traditional straight lines checkerboard makes power spectrum uniform and improves accuracy of the indirect measurement of the PSF for subsequent optical aberration correction of the acquired images.

From 2016 till present we are working on very long range 3D reconstruction and achieved 0.05 pix disparity resolution by combining several technology components developed by Elphel.

* Use of multiple image sensors (4+) instead of conventional binocular stereo.
* Advanced calibration, frequency-domain rectification, aberration correction, filtering, and 2D phase correlation.
* Training and inference of the Deep Neural Network to predict the depth map from 2D correlation results. DNN analyzes context, extracts object edges, separates background, fuses depth map with textures.

Elphel MNC393-XCAM camera is a narrow baseline 3D perception system providing 10% ranging accuracy at 2000 m, by utilizing the achieved 0.05 pix disparity resolution.


'''Contacts:'''

----

Elphel, Inc.

1455 W. 2200 S., Suite 205

West Valley City, Utah 84119

USA

email: info@elphel.com

<!--Toll-free: (888) 3 ELPHEL -->

phone: (801) 783-5555

fax: (801) 812-8267

site: [https://www.elphel.com https://www.elphel.com]



----
''Free Software and Open Hardware. Elphel, Inc., {{CURRENTYEAR}}''