For Our Future with the AI Vision

Image Processing Solution Provider

Since its foundation in 1999,

REGULUS has been providing excellent technologies

as the world leading image-processing technology provider.

Our next dream is to realize the “robot vision”,

which exceeds human sense of vision to “monitor, capture,

and analyze” the objects, anticipating the future which is realized by the AI vision,

and to become the IPS (image processing solution) provider to serve the world society.

Corporate Profile

Company Name
REGULUS Co.,Ltd
Foundation
July 2, 1999
Capital
50 million yen
President
Hirotaka Furusawa
Number of Employees
43
Headquarters
Yotsuya Orchid Building 2F, 23-3, Daikyo-cho, Shinjyuku-ku, Tokyo, Japan 160-0015
Phone
81(country code) 3-5362-1205
Parent Company
INNOTECH CORPORATION

Main Business Contents

  • Manufacturing and Sales of REGULUS-made Products
    Development, Manufacturing and Sales of Intelligent Camera Systems
  • Image Sensor-related Business
    Evaluation of CMOS Image Sensors, Development and Sales of Image Sensors and Mass-production Testers
  • Image Processing Solution Consulting Service
    Providing Optimal Image-processing Technologies Tailored to Customers’ Needs (such as by combining IPs, libraries, and image algorithms)
  • Image Inspection and Image Inspection System Business
    Development of Various DSP/CPU Farms, Development and Sales of Middleware for High-speed Processing Boards and Image Processing Boards for Cameras used for Inspection Device

REGULUS Products&Solutions

INTELLIGENT CAMERA

Summit Series

Incorporating image-processing libraries into intelligent cameras which can be combined to various image sensors and providing edge computing cameras as one-stop solutions which are tailored to various environmental needs.

McKinley

High-end model intelligent camera incorporating FPGA and high-performance CPU
OS
Windows10 IoT Enterprise / Ubuntu / CentOS
CPU
CPU Intel Atom E3825/E3845
HDD
mSATA SSD 16GB
CPU Board Memory
DDR3 2GB
Others
Ethernet /USB3.0 x1/ USB2.0 x1/
FPGA XC7A100T
FPGA On-board memory DDR3 1GB
Insulation input/output
24V supported for robot vision

Eiger

High-performance model intelligence camera incorporating MPSoC.
  • Zynq UltraScale+ MPSoC Series [XCZU4EG]
  • On-board memory DDR4 4GB
  • USB3.0
  • Gyroscope sensor
  • CAN
  • GPIO
  • Ethernet(PoE) Supported
  • Display Port
  • microSD

Eiger

Middle-class intelligent camera incorporating Zynq
  • Xilinx Zynq-7000 series   [XC7Z020-1CLG400C]
  • On-board memory DDR3 1GB  eMMC 4GB
  • USB2.0
  • GPIO
  • PoE (IF board)

Kailas

Kailas

Provides a high-performance camera system by incorporating various image-processing HWIPs and software libraries to intelligent cameras. Its applications include surveillance, FAs, drones, and automotive.

Application Solutions

  • 監視
  • FA/検査
  • ドローン
  • 自動車

Kailas HW IP

Enables high-volume data processing by SensorFusion leveraging various HW IPs

Image Clarification

Clarify low-contrast images and improve visibility as well as accuracy of image recognition.

WDR

Expand the dynamic range where the difference between brightness and darkness is significant.

Distortion Correction

Correct the distortion of lenses. Also, the camera installation location can be calibrated.

SAD(Sum of Absolute Difference)

Detect the distance of objects from stereo images.

SGBM(Semi Global Block Matching)

Detect the distance of objects from stereo images by SGBM algoithem.

ISP4K

Convert the raw data captured by sensors to YUV. Also, capable of 3A control.

  • Standard Functions
  • 12bit RAW input, 10bit YUV output
  • NR
  • Black level correction
  • Linear correction
  • Defect correction
  • Lens shading correction
  • WB gain
  • Demosaic
  • RGB2RGB color conversion
  • Gamma correction
  • YC filters
  • Color correction
  • AE
  • AWB
  • Optional Functions
  • DOL WDR (Sony sensor)
  • Sharpening
  • Single frame WDR
  • Main Features
  • Supports 4k/60p @ Ultrascale+ MPSoC
  • Supports 1080/60p @ Zynq-7000
  • 300MHz @ Ultrascale+ MPSoC
  • 150MHz @ Zynq-7000
  • Main Supported Devices
  • Xilinx Zynq 7000 Series
  • Xilinx Ultrascale+ MPSoC
  • Contact Information
  • REGULUS Co.,Ltd
  • Yotsuya Orchid Bldg 2F, 23-3 Daikyo-cho Shinjyuku-ku, Tokyo, Japan 160-0015
  • TEL: +81-3-5362-1205
  • E-mail: info_qa@reglus.co.jp

JPEG Encoder/Decoder

Capable of irreversible image compression/expansion supporting 8~12 bit input, variable speed, Gray, YUV, RGB, CMYK, and YCCK.

JPEG-LS Encoder/Decoder

Capable of reversible image compression/expansion supporting 8~12 bit input, variable speed, Gray, YUV, RGB, CMYK, and YCCK.

IMAGE PROCESSING SOLUTION

Construction Equipment, Drone and Automotive Solution

Human/automobile detection by deep learning technology

Possible to get higher recognition accuracy of human/automobile detection than conventional methodology by deep learning technology.
[Example of system configuration]
Hardware : AI Camera
Software : JPEG, human/automobile detection deep learning model, distance measurement

Stereo camera

Long-distance measurement and obstacle detection based on the information captured by the stereo camera
[Example of processing time]
8.0ms/pc (125 pcs/sec*) *Numbers may vary depending on various conditions.
[Example of system configuration]
Hardware : Stereo Eiger
Software : Distance measurement algorithm

Detection and measurement of specific wavelengths

Detection and measurement of wavelengths for specific colors of the object (rice plant, vegetables). Also, capable of NDVI calculation by simultaneous photographing utilizing multi-cameras.
[Example of system configuration]
Hardware : Multi-Wavelength Camera
Software : NDVI

Electronic mirror

[Example of system configuration]
Hardware : Eiger
Software : ISP, distortion correction, human detection,
white line detection

Surveillance Camera Solutions

Human detection by deep learning technology

Possible to detect human being in various postures.
[Example of system configuration]
Hardware : Stereo AI Camera
Software : Human detection deep learning model

Fish-eye lens & human detection

Use fish-eye lens to surveil all around, and without distortion correction, possible to detect human being in various postures with high precision by edge computing.
[Example of system configuration]
Hardware : EigerII, sensor IMX377
Software : Detection, classification

Moving object detection

Moving objects (pedestrians, animals, vehicles, and bicycles) only are detected.
[Example of system configuration]
Hardware : Eiger
Software : Moving object detection algorithm and image clarification

Object distance measurement

Object detection and distance measurement can be done by using two cameras.
[Example of system configuration]
Hardware: Stereo Eiger/Stereo AI Camera
Software: object detection algorithm + distance measurement, grouping

High sensitivity camera (in development)

With high sensitivity sensor and AIEigerII, possible to take picture in color or monochrome, in low light even only 0.003lux.
And combining with AI, also possible to detect objects.
[Example of system configuration]
Hardware: Low Light AI Camera
Software: Human detection deep learning model

Infrared camera

Thermal detection and life detection utilizing infrared cameras.
Images can be captured even if the visibility is limited such as in the darkness or in the fog.
[Example of system configuration]
Hardware : Infrared camera
Software : Thermal detection algorithm

FA Solutions

High-speed template matching

High-speed processing of OK/NG determination referring to the templates registered in the FPGA.
[Example of processing time]
8.0ms/pc (125 pcs/sec*) *Numbers may vary depending on the conditions.
[Example of system configuration]
Hardware : McKinley
Software : Template matching

High-speed foreign object (dust or a crack) detection

High-speed detection of specific objects by making the optimal algorithm as a hardware, utilizing FPGA.
[Example of system configuration]
Hardware : McKinley
Software : Various detection algorithms

High-speed positioning

High-speed calculation of the central point of a circle by making the part of processing as a hardware, utilizing FPGA.
[Example of processing time]
CPU only : 580ms→FPGA: 30ms
[Example of system configuration]
Hardware : Eiger
Software : Central point extraction algorithm

Analog gauge reading

Read analog/digital gauge and transfer to digital info. in high-precision
Text data conversion by camera to reduce transferred data
Adapt to various env. by comb. of image processing IP e.g. sharpening
Easy initial setting by MeterViewer
[Example of system configuration]
Hardware : Eiger (Sensor: OV5640)
Software (Kailas): Labelling, Binarization, Pattern Matching, Hough Transformation

IMAGE SENSOR

Various Choices

Creating diverse applications from various image sensor boards
  • AR0135
    GS 1.2M

  • IMX377
    RS 12.35M

  • XQE-0920
    RS 0.92M

  • IMX178
    RS 6.3M

  • IMX290
    RS 2.12M

  • IMX174
    GS 2.3M

Sensor Fusion

Creating diverse applications from various image sensor boards
  • Stereo

  • Stereo + Infrared Camera

  • 4 Cam

DEEP LEARNING

New Image Recognition Technologies

Nowadays various object recognition technologies for objects which were not recognizable under the traditional approaches utilizing Deep Learning are becoming available. REGULUS, mainly by the PC-free and cloud-free edge computing solution, provides a wire range of recognition technologies ranging from high-speed image analysis to real-time processing by leveraging Deep Learning and by developing recognition engines utilizing CNN which is implemented on GPUs and FPGAs. Also, REGULUS provides a learning service for various objects to be recognized.

EagleEye

Stereo camera by EigerII with two sensors connected
Features
Human and object detection with 90 degrees angle of coverage
Simultaneously detect various class of object
Distance measurement & location info. acquisition
Edge computing:GPU free, PC free, Cloud free
Dustproof/Waterproof housing (IP65 equivalent)
Mounting by magnets
Detection range: 0.8m ~ 10m
Configuration
Hardware: EigerII (Sensor: AR0135)
Software (Kailas): Stereo distance measurement, detection, classification
Primary use
Surveillance camera for construction site,
Surveillance camera for public facilities, etc.

HawkEye

Multi sensor camera by EigerII with four sensors connected
Features
High-accuracy detection with a wider coverage
Human and object detection with 180 degrees angle of coverage
Simultaneously detect various class of object
Distance measurement & location info. acquisition
Edge computing:GPU free, PC free, Cloud free
Dustproof/Waterproof housing (IP65 equivalent)
Mounting by magnets
Detection range: 0.8m ~ 10m
Configuration
Hardware: EigerII (Sensor: AR0135)
Software (Kailas): Stereo distance measurement, detection, classification
Primary use
Surveillance camera for construction site,
Surveillance camera for public facilities, etc.