If you have a production line or a quality control procedure the chances are you have heard of Smart Cameras. But what do they offer and what are they really capable of?
Multipix Imaging reviews the extremes of this technology by first summarising the most popular smart camera format, followed by discussing the very latest in high-tech Deep Learning cameras which provide more powerful ‘AI’ based solutions.
First, lets start with the new HIKrobot SC2000 Pro. This is an all-in-one solution of lens-lighting-camera-processor-IO; just what one would expect of a smart device for many of today’s basic vision task which compromises of inspection tools such as Pattern Matching, Measuring, Counting, Location and Brightness/Contrast measurement. These are fundamental machine vision tasks and cover many production line quality requirements. The user requires no programming experience, they simply configure and monitor the SC2000 Pro using the intuitive web-based interface. Supporting communication protocols, including TCP, UDP, Serial, IO, Modbus, PROFINET, Ethernet/IP, FTP, which means it is an easy step to interface the camera to the production line system.
At the other extreme of Smart Camera technology, we have the new ADLINK NEON-2000 JT2-X. This is built on the NVIDIA Jetson TX2 GPU, meaning it has been designed with Deep Learning AI based applications in mind. There are a choice of sensor resolutions and colour/mono options. The lighting and lensing are not included, Multipix Imaging provide these separately, but there is extensive I/O as part of the design to ease the direct interface to PLC’s and so on. Unlike the SC2000 Pro, here you require additional inspection software to be installed/developed for this camera, such as MVTec’s HALCON. The open-platform design requires the developer to have software skills, capable of programming and creating a vision solution.
Although this adds a layer of complexity in using the NEON-2000 JT2-X, it provides a very powerful AI edge-based solution for inspection tasks typical of Deep Learning, for example, sorting and classification… think food grading & sorting, surface inspection such as scratches & defects, grading of material such as wood, textiles and so on. The tasks that are usually impossible with standard machine vision tools.