Computer Vision, as we all know it, is a study of acquiring, processing, analyzing and understanding images, video, live stream data from the real world and extract information in the form of decisions, thereby automating what human visual systems can do. Some of the applications include surveillance, machine vision, medical imaging, mixed reality and automation of image and video processing. Surveillance has become faster and efficient with the advent of computer vision algorithms. Automation has become a push factor in the heavy industrial sector and we are finding robotic arms helping with the location and positioning of parts in the assembly lines. Huge amounts of videos and images are analyzed today and decisions are made in a blink of a second due to this technology. The extent of computer vision’s capability is undebatable and therefore the significance of decision making in the design of such products.
Decisions are always hard to make, especially when the factors to consider are many. In a position like a project manager or a product designer, it becomes even more challenging as the decisions taken at that time will shape a product in the market and drive it in the right direction. First primary thing is the camera type because it is crucial to select a camera that can be used to capture the image. In Today’s world, many multi-featured cameras are available and making a choice has become almost impossible. Also, it becomes important as to what type of media are we handling with our product. Let us now discuss the various options available for the camera and media types and see how they influence the decisions for our computer vision product.
This is the most traditional camera type widely used in webcams, mobile phones and CCTV. They are the three-channel (RGB) camera which produces two-dimensional output images. They produce 30 frames per second and are available in full HD. They are used to analyze normal surveillance footage or even worn as body cameras by policemen for collecting evidence.
This is a camera type that analyzes the depth of the object in addition to the two dimensions. It essentially generates a three-dimensional data and produces 60 frames per second. These cameras can detect how close or far the object is. Typical business use cases are proximity detection, gesture detection. Realsense and Kinect as such examples.
These are low light sensor cameras which senses the IR light emitted and convert them into a single frequency signal or channel data. They convert it into a single channel, two-dimensional data. They can be used for night vision cameras or in places where light conditions are very low. They have a low resolution and produce 30 frames per second.
The industrial cameras are the higher end cameras that produce 200 frames per second. They are used in places where there is a lot of electromagnetic interference, vibrations or where durability is of the essence. In heavy industrial sectors with assembly lines, such cameras prove very useful because the detection and analysis need to be performed in a few milliseconds.
Media is available in the form of stored media or live media. Stored media can be stored images or videos. In the medical field, we find stored images in the form of CT scan images, X-ray images. Satellite images are again a form of stored media and are high-resolution images. Stored video examples include surveillance videos and videos of sporting events. They produce 30 frames per second and in full HD format.
Live Media are used in surveillance streams and AR, VR, XR applications. They are of very high resolution as lower rates can cause fatigue in an individual when using the applications. They are full HD and are of 60 frames per second. In our product development discussion, the type of media handled has to be analyzed before taking a step towards formulating a design.
Home surveillance applications make the use of RBG cameras for alerting the house owners for possible intruders or events that they may want to keep track of. In this type of applications, both stored and live media is used depending on what type of event the surveillance is tracking. For example, if we are tracking an everyday event like a courier drop or monitoring some deliveries that were dropped, then we are analyzing stored images or videos from our home cameras. For situations like intruder alert or some security breach, immediate attention is required and we require live media monitoring for the same. In essence, we need to think about where our product is going to be used and in what conditions to appropriately determine the camera or the media type to be used for collecting the data. To bring out the maximum efficiency and effectiveness of our product, choosing an appropriate camera and media type is essential.