Specific industry applications of image recognition technology from perceptual cameras

Image recognition is an important branch of artificial intelligence

In the past one or two years, the field of artificial intelligence has received unprecedented attention from the media, industry and academia. Everyone agrees that the era of intelligence is coming, and machines are increasingly replacing the unique advantages and skills of human beings. An important possibility is image recognition technology.

Image recognition is a technique in which a computer processes, analyzes, and understands images to identify targets and objects in various modes. To put it simply, let the machine read the content of the image like humans by processing the sensory information instead of just seeing the pixels. At present, with the picture becoming the main information carrier in the Internet, the problem arises. When the information is recorded by words, we can easily find the required content through the keyword search and edit it arbitrarily. When the information is recorded by the picture, we can't retrieve the content in the picture, which affects us from the picture. Find the efficiency of key content. The picture gives us a quick way to record and share information, but it reduces the efficiency of our information retrieval. In this environment, computer image recognition technology is particularly important.

Computer vision has a wide range of applications, including medical imaging analysis to improve disease prediction, diagnosis, and treatment; face recognition is used by Facebook to automatically identify people in photos; it is used to identify suspects in security and surveillance; In terms of shopping, consumers can now take a product with a smartphone to get more purchase options.

Our research in the field of image recognition has made many breakthroughs. The convolutional neural network invented by YannLeCun, the head of Facebook artificial intelligence, has promoted the rapid development of the entire artificial intelligence field in the near future, and its most important applications are image recognition and speech recognition. In 2012, a team led by Wu Enda showed an analysis of millions of YouTube video images from an unsupervised learning machine. The machine learned to classify the common objects it has seen, including human faces and cats (for netizens), including various movements that can be seen everywhere on the Internet: sleeping, jumping, and skateboarding. Humans did not indicate the words "faces" or "cats" on these videos. Instead, the machine simply concludes by looking at the countless examples of each object, and the statistical patterns they exhibit are already universal enough to classify these objects. Papers by Stanford University Andrej Karpathy and Li Feifei describe a computer vision system that can mark a particular part of a given image. For example, give it a breakfast table that recognizes the fork, banana chips, a cup of coffee, and the flowers on the table as well as the table itself. It can even be described in natural English in the scene - although the technology is not particularly perfect.

Application scenarios determine the popularity of image recognition technology

MIT cosmologist Max Tegmark said that the operation of artificial intelligence is already out of the laboratory and into the society. We do see and even use many artificial intelligence services and products, such as better search engine services, voice assistants, and so on. In the segmentation related to image recognition technology, there are many such services and products, such as image search, image comparison, face recognition, automatic image classification, and so on. But although we have seen so many products or features, we have not found any application that generalizes image recognition. Many domestic and foreign startups, even technology giants, have not found the most explosive and promising application direction in the field of image recognition. The reason for this is the lack of application scenarios.

The development and maturity of artificial intelligence depends on three elements, algorithms, big data and application scenarios. Whether it is a startup or a technology giant, they will pay enough attention to the algorithm and spend a lot of manpower and financial resources on algorithms and models and R&D. Second, thanks to the Internet, social media, mobile devices and cheap sensors, the amount of data generated by the world has increased dramatically. As the value of these data continues to be recognized, new technologies for managing and analyzing data have also evolved. Big data is a booster for the development of artificial intelligence because some artificial intelligence techniques use statistical models to make probability estimates of data, such as images, text, or speech, by exposing them to the ocean of data. Optimization, or "training" - now such conditions are everywhere.

Baidu scientist Wu Enda once compared the algorithm and data to the rocket's engine and fuel. Only the two complement each other, and the artificial intelligence rocket can take off. This is also the two major aspects of the current focus of companies in the field of artificial intelligence, but it is easy to ignore the third factor that plays a decisive role in artificial intelligence - application scenarios. The main reason is that our ultimate goal for artificial intelligence is to create a machine that can match humans in terms of comprehensive intelligence. But such a slightly sci-fi goal is difficult to guide our specific work, and may even affect it. Healthy development of the field. When returning to the specific application of artificial intelligence, we should forget the ultimate goal, respect a step-by-step development process, pay attention to the progressive progress of artificial intelligence technology and the segmentation application of various industries. The current technology giants are mainly based on the general-purpose products for the public in the Internet era, such as search engines, or operating systems, and so on. Therefore, they lack the accumulation and experience of certain specific industries to a certain extent, and it is difficult to find out the potential needs of specific industries and the specific application of artificial intelligence technology in this field. At the same time, compared with the industry solutions hidden behind, the application of artificial intelligence technology to popular civilian products can play a better promotion effect and educational significance.

The tipping point of image recognition technology lies in the specific industry solutions.

As mentioned above, the neglect of application scenarios by most companies affects the popularity of artificial intelligence technology in various fields, especially for image recognition technology, which requires a specific application environment as a cognitive computing technology. As a support, we hope that machines can understand the outside world like human beings, instead of making decisions, which is closely related to the specific environment in which the machine is located. Therefore, we have accumulated rich experience in specific industries to understand the needs of the industry, and then Re-use image recognition technology to address these needs and use advanced technology as part of a total solution to truly extend the reach of image recognition technology to truly solve our specific problems, rather than just as a fleeting gimmick.

Whether in the field of artificial intelligence or in the field of segmented image recognition, there are two paths in the process of transforming from technology to practical application. The first is to take the generalized route, that is, this technology can meet various industries. The needs of users, for example, the Watson Open Plan launched by IBM, have now applied this smart computer to finance, medical and customer management. Many technology giants and startups in the field of image recognition are also aiming to bring image recognition technology to general-purpose applications. This is a top-down application path for artificial intelligence. This trend cannot be changed. Any machine and smart device in the future needs "vision", but the problem is that the current image recognition technology may not have reached such a "singularity". This is one of the reasons why most image recognition technology companies have not found the best application direction. This leads to the second path, which is based on the existing image recognition technology level, combined with the application scenarios of specific industries, to solve the needs of the industry, to achieve the best combination of demand and technology. For example, autonomous vehicles, robotic kitchens, face payment, remote face authentication for banking and securities, and so on.

In this respect, some companies with deep accumulation in specific industries have certain advantages. For example, Kodak in Suzhou, which entered the security surveillance field more than a decade ago, may not be as familiar as the Internet-based Internet technology company, and it doesn't look so cool. But the company has been planning and developing image recognition technology since 2006. And the reason they cut into the field of image recognition technology is the increasingly intelligent demand from their customers in the security field. It is this industry accumulation and company genes that determine their ability to stand at the forefront of specific industries and then apply image recognition technology to the specific needs of users.

At the end of 2014, Kodak introduced a new camera category, the Perceptual Camera. Through their product case, we can understand the importance of combining image recognition technology with specific industry needs.

The emergence of imaging technology has greatly mentioned the efficiency of collecting information and storing information, but at the same time it has seriously affected the efficiency of our analysis of information. When we can't extract valuable things from massive data, we lost our original collection. The meaning of the data. The emergence of image recognition technology is to resolve this contradiction. The same is true for the security surveillance field. We have deployed more and more cameras to collect information, but in the end we found that although we seem to have obtained a huge amount of data, but the data processing capability, we have found valuable from the massive data. The ability to inform, but still depends on the human vision behind the monitor screen, and this contradiction has spawned the emergence of video analysis and intelligent monitoring. Due to cost reasons, the intelligent analysis technology for video has gradually migrated from the server to the camera side. This is called a smart camera. At present, the smart cameras on the market are mainly located in the warning applications such as the warning line and the area watching defense, and the intelligent IPC can identify the content in the monitoring screen based on the intelligent analysis of the video. Semantic description and best image capture, while deeper data mining based on the back-end big data platform.

The following will illustrate the application scenario through three specific smart cameras:

1) Feature Analysis Camera

It is mainly for the recognition and capture of moving objects in scenes of people, cars and objects in a scene with a large field of view. The urban roads and intersections under China's national conditions are complex environments of people, motor vehicles and non-motorized hybrids, and at the same time they are the focus of public safety. The feature analysis camera is designed for this scene. It can comprehensively identify the basic characteristics of people's car classification, color, direction, etc., and then carry out professional application of image recognition. The most typical is to provide this information to the large database platform. The car or the person's map and analysis and judgment to further lock similar suspects and vehicles.

2) Personnel bayonet camera

Identify people and details, including face and body (front and back), gender, age, clothing, walking direction, color. The application scenario is: the suspect has been locked and determined to be hiding in a certain community. The traditional means of investigation of the public security department is to send a number of police forces to be strictly guarded in the community, and to identify and judge each entry and exit person as a suspect. Now, the personnel bayonet camera can completely replace the police manual - it automatically recognizes each person's face and body information and captures the best photos submitted to the platform, the platform can automatically perform the comparison analysis in real time, and then the percentage of similarity The suspects were ranked and issued a warning, and the suspect information was transmitted to the on-site standby police in real time. After the site was further clarified, the arrest was carried out.

3) Vehicle bayonet camera

Identify vehicle details including license plates, models, logos, body colors, driving directions, and speed. Typical applications are: 30 serial thefts, using different fake license plates in different locations. The vehicle bayonet camera records all the vehicle details of each case and captures the best photos, and then provides the textual description class structured data and video and photo unstructured data to the big data platform. The platform will compare these millions or even tens of millions of structured data and will collide with all vehicles with similar appearances in 30 cases, providing detailed information on these vehicles and correlating the corresponding photos and videos.

As mentioned above, the development of artificial intelligence requires the common support of algorithms, big data and application scenarios. In addition to the sensory camera with image recognition technology, Kodak also has a big data analysis platform on the back end. For smart cities closely related to security monitoring, massive video data is the most important industry feature in the field of public safety and intelligent transportation. Therefore, big data has become the most urgently needed technology for video applications in these two industries. Through the integration with the smart city big data platform, the intelligent IPC has achieved numerous applications in smart cities, including real-time monitoring, semantic-based intelligent search, high-risk personnel comparison, face photo search. The whole body is like a search, a multi-point collision of a portrait, a map of a vehicle, a multi-point collision of a vehicle, and the like.

The intelligent IPC with perceptual capability is equivalent to one visual sensor in the Internet of Things. The massive amount of information perceived by the camera enters the big data and cloud computing platform, so that we can not only judge the content from a single camera, but also from mass. In the monitoring data, in-depth analysis and mining have a profound impact on social management. The Kodak-aware camera is used in conjunction with the back-end big data platform: the perceptual camera collects, analyzes, identifies, and submits valid data to the back-end. The big data platform stores the data in the cloud, twice. In-depth analysis and prediction of judgment results. At this point, a complete closed loop of video data acquisition, recognition, perception, thinking, and action is formed.

As Ke Wei, general manager of Keda, said, perceptual cameras are the key to smart city big data applications. In the era of big data, perceptual cameras are the future of video surveillance.

Kodak's sensory cameras may be far from our average users, and it seems that there are no products and functions that technology giants and startups have to do with artificial intelligence technologies such as image recognition, but this is the image recognition technology. The best application. And Kodak company is deeply rooted in an industry, and then from the specific needs of the industry, the application of image recognition technology to the industry, and the implementation of artificial intelligence technology to solve specific problems in the industry also provides a valuable value for other artificial intelligence companies. Reference path.

450W LED Grow Light

Ultra Plantâ„¢ Grow Light offers One Chip Technology aimed to meet your indoor growing expectation such as improve plants' quality, increase yield, or better the margin, etc., all for helping you realize a higher return on your crops.


Ultra Plantâ„¢ Grow Light is combined our advanced All-In-One technology with patented optical design and customized light full spectrum supported from our experienced LED engineers, plant specialists and other partners working on horticulture.


From Ultra Plantâ„¢ APP, you are able to schedule the growing process including photoperiod, brightness and spectral in advance. The lighting system will help you grow smarter, easier and better.


Ultra Plantâ„¢ is the most versatile horticultural grow lighting fixture for indoor plants with flexible full spectrum, brightness control and uniform, wider light distribution, suitable for top lighting of all types of crops. No matter it applies to anywhere for any crop, Ultra Plantâ„¢ can do perfect work for you.

450W Led Grow Light,Best Grow Light,Indoor Grow Lights,Indoor Grow Light

Feton Corporation , https://www.fetonledlight.com