Image vision and computer vision services
Image vision
DAB-Embedded BV has experience with the integration of next image sensors:- Linear CMOS
- CCD sensors (Hamamatsu, Sony)
- Event-based sensors EVS (Sony)
- TDI sensors (Gpixel, Hamamatsu)
- Monochrome and Color (Bayer) sensors with YUV, COMP8, and RAW8/RAW10/RAW12/RAW14 output formats (plan and compressed)
(Gpixel, Sony, Omnivision, Onsemi) - 3D Time of Flight sensors (Sony, ADI, Teledyne)
- SLVS-EC (FPGA IP core)
- MIPI CSI-2 C- or DPHY
- LVDS
- Parallel interfaces (DVP)
Modules can be in mobile and industrial formats. Mobile camera modules can be equipped with an AutoFocus system (with VCM chip) and 5-6 small lenses to achieve the best quality or fixed focus lens system. Customers can select a field of view (FoV) angle normally from 33 deg to 180deg with a pinhole or fisheye lens (depends on module construction – if you need specific parameters – ask us, and we will help to calculate).
For example, DAB-Embedded has Belgium stock of the next sensor modules:
· Sony IMX298 (16Mpix) - read more here
· Sony IMX219 (8Mpix)
· Sony 64Mpix - read more here
· Omnivision OV8865 (8Mpix)
· Omnivision OV13865 (13Mpix)
· Onsemi AR1335 (13 Mpix) - read more here
Image sensors Embedded Linux drivers (V4L2, GStreamer ready) available for next platforms:
- i.MX8MPlus (with ISP) up to 12Mpix camera support
- i.MX8Mini quad
- i.MX8M Mini
- i.MX8QuadMax
- i.MX8M Nano
- NVidia AGX (with OpenCL plugin)
- NVidia TX2 (with OpenCL plugin)
- NVidia Jetson Nano (with OpenCL plugin)
- Microchip SAMA5D2 (with ISP) (Up to 5Mpix camera support). We have our own product – Hanesom (the smallest Linux SOM module with image ISP on board - HaneSOM module)
- Microchip SAM7G45 (with ISP) (8Mpix sensor support)
- Texas Instruments Sitara family and Jacinto Automotive family
- Qualcomm Snapdragon 888 (with ISP)
- Ingenic T40 (with ISP) (8Mpix camera support + H.264/H.265 4K video encoder)
- Amlogic ARM-based SoC (including 4K ISP, H.265 4K@60 video encoder, computer vision algorithms accelerator, NPU).
We have experience with:
- Autoexposure algorithms (HF and PDAF)
- IQ tuning (on/off-chip lens shadow correction LSC, black level compensation BLC, bad pixel masking, Gamma correction, CCM, CFA configuration, AWB configuration, CLM)
- RAW10/RAW12/RAW14 conversion to YUV/RGB using GPU OpenCL plugins and ARM SIMD/OpenMP plugins
- Building Gstreamer pipeline with color conversion (on ISP), compression using hardware codecs, such as H.264, and H.265, and making a network streaming interface (RTSP).
- Making a video overlay application on top of camera video streaming.
- Adding Region of Interest (ROI) and scale modes to the drivers (possible to change online during streaming/offline – eg digital zoom).
- Preparation of video/imaging data for post-processing using Machine Learning / Deep learning algorithms and computer vision algorithms.
- We have ready the next pipelines using ML
- Building full image processing pipeline including computer vision pipeline.
- Using an accelerometer for image stabilization and compensation.
- Using internal temperature sensors for image quality compensation.
- Building DSP, OpenCL, and SIMD processing modules (you can read them here).
- Adding AI to customers’ video applications, and training models and optimizing it for the hardware platform.
We have our own developed solutions based on FPGAs:
- Frame grabber: Intel Cyclone 10 GX FPGA (video ISP, DMA, PCIe core) + eFinix FPGA (MIPI CSI-2 RX 4-lanes 1.5Gbps each)
- High-speed frame grabber: Xilinx Artix Ultrascale+ PCIe and MIPI CSI-2 RX 4 lanes 2.5Gbps each
DAB-Embedded offers the full set of services for building end products:
- Concept design/review
- Electronic design
- Enclosure design
- Prototyping
- Mass production setup (local, worldwide, Asia)
- Post-production support