• Developer Resources   
 
 

How our synchronized multi-camera system
accelerated & improved quality inspection

Camera-based inspection systems are not uncommon. Every manufacturing line has cameras at various points on the line to capture images for visual inspection and records. Traditionally, this visual inspection was fully manual. Men/Women used to physically examine manufactured parts (sometimes with the help of magnifying glasses) to determine the problem. Eventually, the magnification became digital using cameras which helped in also saving images and marking flaws for the record. Today, the manual verification part is getting automated by using Vision Algorithms which identify the flaws automatically. This case study is one such example.

Challenges of accelerating
customer expectations

The customer wanted to design a vision-based quality inspection system that would automatically capture and detect flaws in the product. The detection had to be performed in real-time. The main challenge in this requirement was that the flaws had to be detected on material surfaces that were larger in size. Apart from the size, the level of quality expected was high and hence very minor flaws on the surface had to be detected too. The cameras are placed on a robotic arm and the operations need to be performed while the block is in movement.

Six synchronous cameras

Six synchronous cameras

Good Image Quality

Good Image Quality

Edge Processing of Images

Edge Processing of Images

Proper Lighting

Proper Lighting

Selection of the camera module

The size of the material and its shape made it mandatory that multiple cameras had to be used in the solution and the movement also demanded a synchronous capture of images from these multiple cameras. This would help the algorithm as it could treat the images from all these different cameras as images taken in the exact same instance and hence process it accordingly. Since the object is moving the cameras has to capture images under a low exposure time to avoid motion blur. Apart from the imager itself, the host system had to be powerful enough to process these images on the edge and arrive at a decision.

Sensor

AR0521 with 2.2µm
pixel size

Controls

Auto white balance,
auto exposure control

High frame rate

No blur - capturing
materials on conveyer

Machine Vision

Algorithms for anomaly
detection

e-CAM55_CUMI0521_MOD is an e-con Systems™' off the shelf camera module. The large pixel size is required to capture excellent images at low exposure time and an ISP for exact color reproduction. ON Semiconductor® AR0521, the 1/2.5" Image sensor with 2.2μm pixel size along with the powerful onboard Image Signal Processor proved to be the camera for this solution. With bigger pixels and support for various image quality controls, excellent images are captured at different and challenging lighting conditions with minimal exposure time.

How e-con Systems™ delivered a
hassle-free solution

The first step e-con Systems™ took was to make sure the module and image quality was a fit for this requirement. Using the USB Camera - See3CAM_CU55 which uses the same e-CAM55_CUMI0521_MOD module images of material surfaces were captured and given to the customer so that the algorithm could be evaluated with the images. On proving that the camera was a good fit for this solution, e-con Systems™ worked on designing the solution for the customer based on this module. Six e-CAM55_CUMI0521_MOD modules were connected to a NVIDIA Jetson TX2 board over MIPI Interface. To support the long-distance cabling GMSL coaxial cables were used between the host and cameras. Being one of the early adopters of the GMSL technology, e-con Systems™ was able to achieve both synchronization and low latency on the image pipeline from the six cameras. A proper driver implementation on the NVIDIA TX2 side made sure that high-quality images were available in memory for the algorithm to process them.

Quality Inspection Flow Chart

The team from the customer side had worked on the algorithm and used the hardware and drivers provided by e-con Systems™ to complete the solution.

Industrial Tablets Case Study Demo

Get in touch

Are you looking to innovate and accelerate your product development?

Talk to Us

Cameras for quality inspection is a tricky scenario. At one end you have the expensive and high-performance machine vision cameras and at the other end, you have the low-cost web cameras. While the off the shelf webcams don’t fit the requirement, the really expensive Machine Vision Cameras are a overkill in many situations. Finding the right blend of features to suit Quality inspection requirements in an Industrial setup is important and our customization services come handy under these situations.

e-CAM55_CUMI0521_MOD


5MP MIPI Camera
Module