Computer Vision allows machines and computers to ‘see’ the physical world. It can extract visual information much beyond what a naked human eye can see. The computer vision market is on the verge of explosive growth; it is pegged at USD 10.9 billion and slated to grow to USD 17.4 billion by 2024-growing at a CAGR of 7.8% during the forecast period.
Computer Vision allows machines & computers to ‘see’ the physical world.
The benefits of computer vision include time-efficiency (works 24/7), high accuracy, repeatability and reduced cost (cheaper than labor cost). Computer vision can be used in various use cases across industries. Some of the examples include measurement inspections of machine parts, anomaly detection for any physical parts/components etc., defect identification, packaging Inspection, product composition check, and many more. Some cameras that can help industry 4.0 in automation and inspection needs are explored in this blog.
- High Precision 2D Cameras
These cameras range from a simple VGA camera to a high resolution up to 47 megapixel. Most of the manufacturers package an imager (camera) and a software (can help in measurement & anomaly identification). These cameras are typically very fast, small in size, and cover a larger inspection area, to measure multiple products on a moving conveyor belt. These cameras operate in challenging environments and hence have a high threshold for temperature, humidity, and weather conditions. Some applications for such cameras are as follows:
- 2D measurement inspection (up to µm)
- Location and orientation inspection
- Print inspection
- Pattern matching
- Component location/inspection
- Connectors inspection
- 3D Profiler Cameras
These cameras are based on laser technology to identify even the slightest variations on the surface of the inspected product. 3D profiler cameras create a 3D profile of an ideal product. This profile will be used to inspect products at the time of manufacturing. Generally, these cameras come with bundled software to help in anomaly detection or measurements. These cameras have the capability to weed out even a small variation like an embossed logo on a phone. The 3D profiler cameras score over VGA cameras because they can work on 3D model and even difficult surfaces such as shiny, reflective, translucent and dark. Some applications are as follows:
- 3D measurements
- Anomaly detection in 3D
- Hyperspectral Imaging
This is where computer vision gets very interesting. There are only a limited range of UV range that a human eye can see and interpret. There is far more information available in other UR & IR ranges that might be useful to make decisions. Hyperspectral imaging were initially developed to help in mining and geology to identify the quality of ore and minerals. For example, the human eye will not be able to identify the composition of a medicine by just looking at it, whereas a hyperspectral camera can identify if there is a variation in the salt composition. Sugar, salt, and citric acid might look the same to a human eye, but under a hyperspectral camera, they are visible in a completely different color.
Some applications include the following:
- Mining (to identify a mining site based on the composition of soil and rock)
- Medical (identify an anomaly in the composition of salt in medicine, eye care)
- Agriculture (pest inspection in crops, food processing)
- Quality categorization of mineral, ore, oil, agriculture produce, or any product
- Surveillance, chemical imaging, environment
- Thermographic Cameras
It is also called as an infrared camera or thermal imager. The infrared camera creates an image using IR radiations which is sensitive to wavelengths from 1µm to 14µm. Again, these wavelengths are not visible to the human eye and this information is very useful in a lot of industrial and medical scenarios. For example, the thermographic images below highlight the portion in red (or orange) to alert the user on raised temperature which might result in fire.
These thermal imager cameras can be handheld or deployed over a moving robotic arm to keep monitoring multiple machines in an industrial setup. Some applications include:
- Industrial: Fire prevention caused by short circuit or loose connection.
- Automotive: Aid the driver in automotive night vision
- Medical: identify temperature in human and animals
- Firefighting, building inspections, refrigeration leaks, surveillance, and flame detection
- Neuromorphic Cameras
It is also called an event camera, and is unlike a regular VGA camera which captures the image using a shutter. The event camera camera captures each pixel information independent of each other asynchronously and just report changes in brightness if they occur, else they stay silent. This kind of camera ensures that the data storage space needed is low as this stores only delta. This smaller size image is useful when transferring data. This can work in a dark environment as well. 
Some Applications:
- Industrial: Vibration monitoring of machines
- Surveillance (with minimal data storage and transfer)
- Domain agnostic: Any application where data storage and transfer is expensive
