Under the background of the current era of smart manufacturing industry 4.0 with high-end equipment manufacturing as the core, with the deepening of the "Made in China 2025" strategy, the industrial intelligent robot industry market is showing a growing momentum, which serves as the machine vision function of the "eyes" of industrial robots. Must not be.
The security technology is constantly changing. After the industry has completed networking and high-definition, intelligent applications have become a hot technology for security manufacturers. It should be said that from 2012 onwards, the Chinese security industry has begun to conduct in-depth research and development on how to make intelligent applications of products and systems, leading to a wave of technological trends. If intelligent video analysis is the 2.0 version of smart security applications, then machine vision will open the era of China's smart security application 2.0. This is a new application space for smart security.
Accurately grasp the research content of machine vision
To explore the application of machine vision, you first need to accurately grasp what is machine vision, and distinguish it from the current hot and overlapping artificial intelligence and deep learning.
AI has the largest outreach, including machine vision, deep learning, robotics, biometrics, and natural language processing. It can be seen that artificial intelligence includes deep learning and machine vision. It studies the laws of human intelligence activities, constructs artificial systems with certain intelligence, and studies how to let computers accomplish tasks that previously required human intelligence to be competent, that is, how to study The computer software and hardware are used to simulate the basic theories, methods, and techniques of human intelligence.
Deep learning is a new field of artificial intelligence research. Its motivation lies in building and simulating the human brain to analyze and learn the neural network. It imitates the mechanism of the human brain to interpret data such as video, sound, and text. Deep learning is extremely popular in China's security industry. Face and vehicle recognition technologies are based on deep learning. The reason is that the key element of deep learning is data, and more than 60% of the total amount of large data is video surveillance data. Therefore, deep learning has been applied in all aspects of the security industry: face detection, vehicle detection, and non-machine Motor vehicle testing face recognition, vehicle brand recognition, pedestrian search, vehicle detection, human attributes, abnormal face detection, crowd behavior analysis, and tracking of various interest targets.
Machine vision is a branch of artificial intelligence that is rapidly developing. Simply put, machine vision is the use of machines instead of the human eye to make measurements and judgments. Through the CMOS or CCD image sensor will be captured target converted into an image signal, transmitted to a dedicated image processing system, get the morphological information of the target, according to pixel distribution and brightness, color and other information, into a digital signal; image system for these The signal performs various operations to extract the characteristics of the target, and further controls the equipment operation in the field according to the result of the discrimination.
Machine vision is mainly used for image recognition. Therefore, machine vision is widely used in face recognition, license plate recognition, and other aspects. Taking the intelligent transportation industry as an example, machine vision has the advantages of low cost, strong stability, high accuracy, and wide application range. It has been widely used in the traffic monitoring systems of highways and highways in various countries and is embodied in license plates. Identification, body color identification, vehicle identification, illegal identification, traffic statistics, flow control, etc.
At this time, some people may have doubts. There is so much overlap between machine vision and deep learning. Whether the two are different expressions of the same concept in the security industry. Actually, if only from the video surveillance industry, the learning algorithm is a higher level application of machine vision, because a large number of data samples are captured based on the huge sample acquisition, and machine vision is mainly in feature sensing, image preprocessing, feature extraction, and features. The screening surface is good, that is to say, machine vision is mainly in the feature recognition refining section, and deep learning is the combination of features and learning, such as the use of feature sensing and extraction to predict data.
Machine vision is the core technology of security applications
In June of this year, the authoritative organization of Business Wire announced that the global machine vision market is in rapid development and it is expected that by the end of 2025, the market value will exceed 19.2 billion US dollars. To date, machine vision technology has not only been successfully applied in many fields but also has gradually expanded its application scope. From the initial electronics manufacturing and semiconductor manufacturing companies, it has developed into security, packaging, automotive, transportation and printing industries.
Security is one of the main battlefields of machine vision applications. There are several major new technology applications that bring changes to the security industry:
1. Target recognition
Target recognition technology and stable tracking method are key technologies for the development of machine vision in security. It has been widely used in many fields, such as fingerprint identification, face recognition, iris recognition for identity verification, and license plate recognition in applications such as intelligent traffic management, vehicle detection, and parking lot management. A target recognition system should have the ability to detect, classify, and identify targets in complex backgrounds and various weather conditions so that the target can be continuously tracked.
In recent years, the target recognition technology has gradually moved from theoretical exploration to laboratory simulation to practical application. The technical method has also been identified from the classical statistical model, and it has been moving toward knowledge, model, multi-sensor information fusion and deep learning neural network recognition methods. Evolution.
2. Target tracking
Motion target tracking is the process of determining the position of the same object in different frames of an image sequence. Its main working method is to select good target features and adopt appropriate search methods. According to the matching principle, the existing tracking methods are divided into tracking based on models, regions, features, and active contours, which are the capabilities of machine vision.
3. Binocular technology
The core purpose of binocular stereo technology is to improve the accuracy of recognition. Because the three-dimensional geometric information of the object is contained in the field of view formed by the stereoscopic vision technology, the detection rule can be effectively set, and interference factors such as light and shadow can be eliminated, and the accuracy of the intelligent analysis can be greatly improved. When China's leading security companies detected PTZ cameras at high altitudes, they discovered that they were all based on machine vision binocular technology. Their recognition rates for humans and objects were greatly improved, and they used dual cameras or multiple cameras to space within the visual field. The three-dimensional position coordinates and attitude of the free moving body are measured with high precision, the centroid position of the moving target is determined, and the moving target is highly accurately tracked according to the calibration result.
4. Multi-PTZ speed dome camera machine linkage tracking technology
Multi-ball machine linkage tracking technology is based on single ball machine intelligent tracking technology. From the perspective of application, it is possible to upgrade the single-point monitoring of ordinary trackball cameras to the seamless relay tracking of a single target in the system. In conjunction with the use of electronic maps, it is easy to implement high security level areas. Seamless tracking, and achieve high-level security requirements such as target trajectory description and criminal behavior warning. The realization of multi-camera linkage tracking technology requires multi-target recognition and tracking technology. In the application, a billiard machine is usually set as an initiation point to analyze the intelligent behavior of the target in a wide area, and the multiple targets monitored at the same time are sorted according to a predetermined strategy, and the intelligent tracking ball is commanded according to the sequence. The machines track the monitoring targets one by one. Compared with single-target tracking, the key point of multi-target tracking technology is the data correlation problem. That is, a unified coordinate system is established so that the originating dome camera can transmit the target coordinate information to the tracking dome camera and realize linkage tracking.
5. Video summary and video retrieval for hindsight applications
With the popularization of monitoring probes, massive video data is stored in the monitoring system. In the current manual viewing mode, traditional methods need to be played from beginning to end. It often takes several times longer than the original video to complete the review. A large number of people need to work overtime for several weeks to review the video. In order to avoid omissions and errors, we must increase the method of human input. How to reduce the timeliness, cost, and fatigue caused by manual viewing and playback, and how to accurately identify the information that needs to be obtained in videos with different resolutions and resolutions. Based on the above requirements, security manufacturers Research and development of video abstracts, video retrieval and other technical means, which is one of the research content of machine vision.
Machine vision opens smart security applications
From the above introduction and analysis, we can probably understand that machine vision has a certain understanding, but what it can do in the security industry and what changes it brings to the industry deserve attention. Let's take a look at the value of machine vision applications first:
On July 16, 2017, in the Wal-Mart Supermarket in Xixiang Golden Harbor, Bao'an District, Shenzhen, a man was injured by a kitchen knife and killed 2 people and injured 9 others. Today, surveillance can be found everywhere in large and small shopping malls, becoming almost essential for every mall store. Obviously, the use of monitoring is very useful for the daily management and security of shops. However, in reality, current human-powered monitoring equipment has a huge security loophole, which makes security work to a certain extent useless. The intelligent security system based on machine vision technology can identify images from many video materials, search for information such as suspicious characters, identify and screen high-risk individuals, and automatically identify suspicious objects from multiple monitoring devices, such as kitchen knives. And so on, automatic alarms are issued to remind security personnel.
In addition to the preventive and deterrent functions in advance, the video summary and retrieval of machine vision is an intelligent technology for ex-post applications. Among them, the video summary technology enables 24-hour video to be made into brief videos of a few minutes to become a reality, which will greatly improve the efficiency of mass video surveillance video analysis. Video retrieval technology mainly relies on the video algorithm to preprocess the video. After the video content is structured, the effective information in the video content is extracted, and after marking or related processing, the human can perform description through various attributes. Quick search.
It can be said that if the above cases have intelligent security products for machine vision, such tragedies will certainly not occur.
How does machine vision apply the new era of China's security?
1. Establish a face capture database with public security agencies
In areas where people are crowded in public places, such as Tiananmen Square, railway stations and other sensitive public places, relevant machine vision products are continuously refined with the development of technology. The product can establish a face recognition comparison system with a public security agency, establish a face capture database, archive face information, and establish relationships with personal identities. Face search, blacklist deployment, and stranger identification are used in the actual application process. A number of intelligent analysis capabilities greatly increase the effectiveness of video surveillance and prevent criminals from becoming invisible.
2. Identification and Control of Vehicles in Intelligent Transportation
In the transportation industry, compared with other identification technologies, machine vision has the advantages of low cost, strong stability, high accuracy, and a wide range of applications. It has been widely used in traffic monitoring systems for highways and highways at home and abroad. , such as license plate recognition, body color identification, vehicle identification, illegal identification, traffic statistics, flow control.
3. Face deployment
Face recognition is based on human facial feature information for identification, mainly through the camera to capture the face or video stream containing the face, and then in the picture or video stream for face detection and capture, facial feature extraction, in order to capture The face is compared with the face in the database. At present, many leading security companies in China, such as Sysvideo, have introduced a machine vision face placement control system that can be adapted to various traffic arteries, accommodation sites, smart buildings, large chain shopping malls, safe cities and other occasions, with face deployment The core business functions such as graph search and user management.
Some companies are working hard to change security with machine vision. For example 2017 China Shanghai International Machine Vision Exhibition VisionChina held in Shanghai, Hikvision robot technology and design are derived from machine vision; Zhejiang Dahua and Intel jointly push the machine vision business. On June 29 this year, Corel Technology and Intel released a series of machine vision products such as the 9MP/12MP high-performance small area array camera and the 50MP CoaX Press large area array camera. In addition, the general-purpose three-dimensional real-time imaging technology developed by Beijing Qingying Machine Vision Technology Co., Ltd. has the common characteristics and "two major characteristics", filling the gap in the field of Chinese machine vision; Sysvideo has introduced a license plate recognition camera and face recognition camera based on deep learning algorithms.
Conclusion
China's security industry has such a view that machine vision has formed a closed loop of intelligent video analysis. From feature extraction to the application, it has greatly enhanced the application of smart security industry. It is Smart Security 2.0. For this reason, a smart security storm triggered by machine vision has already blew the world and is rapidly spreading, changing the smart security landscape.
The security technology is constantly changing. After the industry has completed networking and high-definition, intelligent applications have become a hot technology for security manufacturers. It should be said that from 2012 onwards, the Chinese security industry has begun to conduct in-depth research and development on how to make intelligent applications of products and systems, leading to a wave of technological trends. If intelligent video analysis is the 2.0 version of smart security applications, then machine vision will open the era of China's smart security application 2.0. This is a new application space for smart security.
Accurately grasp the research content of machine vision
To explore the application of machine vision, you first need to accurately grasp what is machine vision, and distinguish it from the current hot and overlapping artificial intelligence and deep learning.
AI has the largest outreach, including machine vision, deep learning, robotics, biometrics, and natural language processing. It can be seen that artificial intelligence includes deep learning and machine vision. It studies the laws of human intelligence activities, constructs artificial systems with certain intelligence, and studies how to let computers accomplish tasks that previously required human intelligence to be competent, that is, how to study The computer software and hardware are used to simulate the basic theories, methods, and techniques of human intelligence.
Deep learning is a new field of artificial intelligence research. Its motivation lies in building and simulating the human brain to analyze and learn the neural network. It imitates the mechanism of the human brain to interpret data such as video, sound, and text. Deep learning is extremely popular in China's security industry. Face and vehicle recognition technologies are based on deep learning. The reason is that the key element of deep learning is data, and more than 60% of the total amount of large data is video surveillance data. Therefore, deep learning has been applied in all aspects of the security industry: face detection, vehicle detection, and non-machine Motor vehicle testing face recognition, vehicle brand recognition, pedestrian search, vehicle detection, human attributes, abnormal face detection, crowd behavior analysis, and tracking of various interest targets.
Machine vision is a branch of artificial intelligence that is rapidly developing. Simply put, machine vision is the use of machines instead of the human eye to make measurements and judgments. Through the CMOS or CCD image sensor will be captured target converted into an image signal, transmitted to a dedicated image processing system, get the morphological information of the target, according to pixel distribution and brightness, color and other information, into a digital signal; image system for these The signal performs various operations to extract the characteristics of the target, and further controls the equipment operation in the field according to the result of the discrimination.
Machine vision is mainly used for image recognition. Therefore, machine vision is widely used in face recognition, license plate recognition, and other aspects. Taking the intelligent transportation industry as an example, machine vision has the advantages of low cost, strong stability, high accuracy, and wide application range. It has been widely used in the traffic monitoring systems of highways and highways in various countries and is embodied in license plates. Identification, body color identification, vehicle identification, illegal identification, traffic statistics, flow control, etc.
At this time, some people may have doubts. There is so much overlap between machine vision and deep learning. Whether the two are different expressions of the same concept in the security industry. Actually, if only from the video surveillance industry, the learning algorithm is a higher level application of machine vision, because a large number of data samples are captured based on the huge sample acquisition, and machine vision is mainly in feature sensing, image preprocessing, feature extraction, and features. The screening surface is good, that is to say, machine vision is mainly in the feature recognition refining section, and deep learning is the combination of features and learning, such as the use of feature sensing and extraction to predict data.
Machine vision is the core technology of security applications
In June of this year, the authoritative organization of Business Wire announced that the global machine vision market is in rapid development and it is expected that by the end of 2025, the market value will exceed 19.2 billion US dollars. To date, machine vision technology has not only been successfully applied in many fields but also has gradually expanded its application scope. From the initial electronics manufacturing and semiconductor manufacturing companies, it has developed into security, packaging, automotive, transportation and printing industries.
Security is one of the main battlefields of machine vision applications. There are several major new technology applications that bring changes to the security industry:
1. Target recognition
Target recognition technology and stable tracking method are key technologies for the development of machine vision in security. It has been widely used in many fields, such as fingerprint identification, face recognition, iris recognition for identity verification, and license plate recognition in applications such as intelligent traffic management, vehicle detection, and parking lot management. A target recognition system should have the ability to detect, classify, and identify targets in complex backgrounds and various weather conditions so that the target can be continuously tracked.
In recent years, the target recognition technology has gradually moved from theoretical exploration to laboratory simulation to practical application. The technical method has also been identified from the classical statistical model, and it has been moving toward knowledge, model, multi-sensor information fusion and deep learning neural network recognition methods. Evolution.
2. Target tracking
Motion target tracking is the process of determining the position of the same object in different frames of an image sequence. Its main working method is to select good target features and adopt appropriate search methods. According to the matching principle, the existing tracking methods are divided into tracking based on models, regions, features, and active contours, which are the capabilities of machine vision.
3. Binocular technology
The core purpose of binocular stereo technology is to improve the accuracy of recognition. Because the three-dimensional geometric information of the object is contained in the field of view formed by the stereoscopic vision technology, the detection rule can be effectively set, and interference factors such as light and shadow can be eliminated, and the accuracy of the intelligent analysis can be greatly improved. When China's leading security companies detected PTZ cameras at high altitudes, they discovered that they were all based on machine vision binocular technology. Their recognition rates for humans and objects were greatly improved, and they used dual cameras or multiple cameras to space within the visual field. The three-dimensional position coordinates and attitude of the free moving body are measured with high precision, the centroid position of the moving target is determined, and the moving target is highly accurately tracked according to the calibration result.
4. Multi-PTZ speed dome camera machine linkage tracking technology
Multi-ball machine linkage tracking technology is based on single ball machine intelligent tracking technology. From the perspective of application, it is possible to upgrade the single-point monitoring of ordinary trackball cameras to the seamless relay tracking of a single target in the system. In conjunction with the use of electronic maps, it is easy to implement high security level areas. Seamless tracking, and achieve high-level security requirements such as target trajectory description and criminal behavior warning. The realization of multi-camera linkage tracking technology requires multi-target recognition and tracking technology. In the application, a billiard machine is usually set as an initiation point to analyze the intelligent behavior of the target in a wide area, and the multiple targets monitored at the same time are sorted according to a predetermined strategy, and the intelligent tracking ball is commanded according to the sequence. The machines track the monitoring targets one by one. Compared with single-target tracking, the key point of multi-target tracking technology is the data correlation problem. That is, a unified coordinate system is established so that the originating dome camera can transmit the target coordinate information to the tracking dome camera and realize linkage tracking.
5. Video summary and video retrieval for hindsight applications
With the popularization of monitoring probes, massive video data is stored in the monitoring system. In the current manual viewing mode, traditional methods need to be played from beginning to end. It often takes several times longer than the original video to complete the review. A large number of people need to work overtime for several weeks to review the video. In order to avoid omissions and errors, we must increase the method of human input. How to reduce the timeliness, cost, and fatigue caused by manual viewing and playback, and how to accurately identify the information that needs to be obtained in videos with different resolutions and resolutions. Based on the above requirements, security manufacturers Research and development of video abstracts, video retrieval and other technical means, which is one of the research content of machine vision.
Machine vision opens smart security applications
From the above introduction and analysis, we can probably understand that machine vision has a certain understanding, but what it can do in the security industry and what changes it brings to the industry deserve attention. Let's take a look at the value of machine vision applications first:
On July 16, 2017, in the Wal-Mart Supermarket in Xixiang Golden Harbor, Bao'an District, Shenzhen, a man was injured by a kitchen knife and killed 2 people and injured 9 others. Today, surveillance can be found everywhere in large and small shopping malls, becoming almost essential for every mall store. Obviously, the use of monitoring is very useful for the daily management and security of shops. However, in reality, current human-powered monitoring equipment has a huge security loophole, which makes security work to a certain extent useless. The intelligent security system based on machine vision technology can identify images from many video materials, search for information such as suspicious characters, identify and screen high-risk individuals, and automatically identify suspicious objects from multiple monitoring devices, such as kitchen knives. And so on, automatic alarms are issued to remind security personnel.
In addition to the preventive and deterrent functions in advance, the video summary and retrieval of machine vision is an intelligent technology for ex-post applications. Among them, the video summary technology enables 24-hour video to be made into brief videos of a few minutes to become a reality, which will greatly improve the efficiency of mass video surveillance video analysis. Video retrieval technology mainly relies on the video algorithm to preprocess the video. After the video content is structured, the effective information in the video content is extracted, and after marking or related processing, the human can perform description through various attributes. Quick search.
It can be said that if the above cases have intelligent security products for machine vision, such tragedies will certainly not occur.
How does machine vision apply the new era of China's security?
1. Establish a face capture database with public security agencies
In areas where people are crowded in public places, such as Tiananmen Square, railway stations and other sensitive public places, relevant machine vision products are continuously refined with the development of technology. The product can establish a face recognition comparison system with a public security agency, establish a face capture database, archive face information, and establish relationships with personal identities. Face search, blacklist deployment, and stranger identification are used in the actual application process. A number of intelligent analysis capabilities greatly increase the effectiveness of video surveillance and prevent criminals from becoming invisible.
2. Identification and Control of Vehicles in Intelligent Transportation
In the transportation industry, compared with other identification technologies, machine vision has the advantages of low cost, strong stability, high accuracy, and a wide range of applications. It has been widely used in traffic monitoring systems for highways and highways at home and abroad. , such as license plate recognition, body color identification, vehicle identification, illegal identification, traffic statistics, flow control.
3. Face deployment
Face recognition is based on human facial feature information for identification, mainly through the camera to capture the face or video stream containing the face, and then in the picture or video stream for face detection and capture, facial feature extraction, in order to capture The face is compared with the face in the database. At present, many leading security companies in China, such as Sysvideo, have introduced a machine vision face placement control system that can be adapted to various traffic arteries, accommodation sites, smart buildings, large chain shopping malls, safe cities and other occasions, with face deployment The core business functions such as graph search and user management.
Some companies are working hard to change security with machine vision. For example 2017 China Shanghai International Machine Vision Exhibition VisionChina held in Shanghai, Hikvision robot technology and design are derived from machine vision; Zhejiang Dahua and Intel jointly push the machine vision business. On June 29 this year, Corel Technology and Intel released a series of machine vision products such as the 9MP/12MP high-performance small area array camera and the 50MP CoaX Press large area array camera. In addition, the general-purpose three-dimensional real-time imaging technology developed by Beijing Qingying Machine Vision Technology Co., Ltd. has the common characteristics and "two major characteristics", filling the gap in the field of Chinese machine vision; Sysvideo has introduced a license plate recognition camera and face recognition camera based on deep learning algorithms.
Conclusion
China's security industry has such a view that machine vision has formed a closed loop of intelligent video analysis. From feature extraction to the application, it has greatly enhanced the application of smart security industry. It is Smart Security 2.0. For this reason, a smart security storm triggered by machine vision has already blew the world and is rapidly spreading, changing the smart security landscape.
Setiap IP camera biasanya mempunyai software IP installer yang digunakan untuk mendeteksi, dan menyeting IP address walaupun berbeda segment dengan PC/laptop yang terhubung dengannya. https://www.completealarms.com.au/for-home/cctv-for-home/
ReplyDelete