Regarding five senses of human, vision is the most developed and commonly used. We recognize objects, find a way to the destination, learn, work and so on with our eyes in our daily life. In order to make computer watch like human beings, computer vision is created, starting with optical sensors to capture an image, then transforming images into digital information that can be more easily processed by a computer and acted upon.
Computer vision is ubiquitous nowadays, we can find its implementation in all kinds of aspects of our life, such as phone camera that automatically focuses to human's face, facial recognition during an ID validation process. It is not a new term anymore because come applications from manufactory, remote sensing and high-resolution camera use this technology for several decades. It was mainly built on proprietary platforms in the past, but now computer vision is more and more combined with IP based technologies. There are new applications created, for example, the combination of computer vision and analytics engine stimulates 3D-scene recognition and analyze for a computer for virtual reality (VR) and augmented reality (AR) applications.
It is comprised of several technologies working together. For example Artificial Intelligence(AI), Signal Processing and Pattern Recognition. Computer vision engineering is an interdisciplinary field requiring cross-functional and systems expertise in a number of these technologies.
Besides image analytics, computer vision is used in the video process, though the first video analytics tools use handcrafted algorithms to identify specific features. The accuracy is satisfying in the laboratories and simulation environments, but can not meet people's requirements in case that input data such as camera views and lighting conditions deviated from design assumptions in a real-world application. It takes ages for developer and engineers to come up with new algorithms and to tune them for real-world usage. However, cameras or video recorders using these algorithms are not robust enough until deep learning was brought in.
Computer vision is revolutionized by deep learning during the recent several years, due to not only the creation of the Artificial Neural Networks(ANN) algorithm, which mimic human brain neurons but an acceleration of computational velocity and ability. From the early 2010s, graphics processing units(GPU) were developed, it is generally 10 times faster than CPU when it comes to graphic precess. Moreover, a version of the ANN called Convolutional Neural Network(CNN) demonstrated a huge leap in accuracy in 2012, based on the powerful computational capability. This drove renewed interest and excitement into computer vision engineer and graphic analytics engine.
Nowadays, deep learning algorithms even exceeded human counterparts in some applications involving image classification and facial recognition. More inconceivably, these algorithms make the computer learn and adapt different conditions just like human eyes. Computer vision is, therefore, more and more intelligent, and is used widely.