Application of video intelligent analysis technology in ATM monitoring

【Abstract】This paper describes the development trend and application characteristics of intelligent analysis technology in the field of video surveillance. Combined with the analysis of various criminal behaviors currently carried out by criminals against ATMs (bank automatic deposit and withdrawal machines), this paper discusses the role of intelligent analysis technology in financial security ATM video surveillance. At the same time, this paper describes the construction method, main technical characteristics and functions of the ATM intelligent video surveillance system in combination with actual practice.

The ATM intelligent video surveillance system described in this article has been applied in practice and has produced good social and economic benefits, making ATM surveillance a leap from traditional surveillance to intelligent video surveillance.

1 Overview

So far, the development of video surveillance systems has gone through three generations. Whether it is the first, second or third generation video surveillance system, it currently requires manual real-time monitoring of video images to capture the occurrence of abnormal events and emergencies. This has serious defects in "making appropriate and effective responses to abnormal events and emergencies in a timely manner" and cannot meet people's current requirements for security precautions. Based on the third generation of digital, networked video surveillance systems, integrating the latest research results of image processing, computer vision and artificial intelligence, using computers to realize automatic analysis, automatic capture and automatic alarm of abnormalities and emergencies in monitoring scenes, this is intelligent video surveillance.

2 Video Intelligent Analysis Technology and Its Development Trends

2.1 Intelligent Analysis Technology

Video intelligent analysis technology mainly refers to: "automatically extracting key (feature) information from video sources and performing intelligent analysis, and judging it by setting certain conditions and rules". If the front-end device (such as a camera) is regarded as a person's eyes, then video intelligent analysis technology can be regarded as a person's brain. Intelligent analysis technology uses the powerful data processing capabilities of the CPU to perform high-speed analysis on the massive data in the video screen and provide users with valuable key information.

The processing method of video intelligent analysis is mainly to analyze the movement of people in the scene. The process generally includes motion detection, target tracking, behavior recognition and understanding. Among them, the underlying processing links such as motion detection and target tracking have developed rapidly in the past few years and have been put into practical use. Behavior recognition, as the high-level processing part of human motion analysis, is a very challenging research direction and a hot topic in current academic research.

2.2 Current Development Status of Intelligent Video Surveillance

Due to its wide application prospects and potential economic value, video surveillance has attracted wide attention from researchers at home and abroad. In particular, a large number of project research has been carried out in the United States, the United Kingdom and other countries. In 1997, the Defense Advanced Research Projects Agency of the United States established the major visual surveillance project VSAM, which was led by Carnegie Mellon University (CMU) and participated by many universities such as the Massachusetts Institute of Technology (MIT). It is mainly used for automatic video understanding technology for monitoring battlefields and ordinary civilian scenes; the real-time visual monitoring system W4 can not only locate people and segment their body parts, but also achieve multi-person tracking by establishing appearance models, and can detect simple behaviors such as whether people are carrying objects; the Pfinder system developed by MIT can track a single person more accurately; the University of Reading in the United Kingdom has carried out relevant research on the tracking of vehicles and pedestrians and their interaction recognition. After the 9.11 terrorist incident in the United States, research in the field of video intelligent surveillance has received more attention, and some companies have also carried out related research. For example, companies such as ObjectVideo, NICE, IBM, Microsoft, GE and UTRC have launched intelligent monitoring products for specific occasions. These video intelligence technology products are mainly used in high-security and high-risk situations, such as national borders, docks, refineries, chemical plants, nuclear power plants, and water purification. ObjectVideo's products are based on computer vision technology and can provide detection, classification, tracking, and analysis of threatening or interesting behaviors. NICE's main products include detection of lost luggage at airports or other transportation venues, vehicle and pedestrian detection in safe areas, and crowd statistics. GE's VideoIQ product mainly realizes human detection in a variety of scenarios, and can detect and track people in a wide range of outdoor scenes, while alerting people who enter specific areas. Companies such as IBM and Microsoft are also gradually applying vision-based gesture recognition interfaces to the commercial field.

Representative companies in Israel include Mate, Nicevision, and Ioimage. Mate's main product, BehaveWatch, can detect six abnormal behaviors, including path detection, sudden alarm, directional motion detection, etc., and proposes a three-layer protection concept. NiceVision's video analyzer provides real-time detection functions for different threats, including intruders, vehicles, abandoned luggage and packages, perimeter and facility protection, blocking of security entrances and exits and fire routes, theft detection/prevention of security equipment being moved, counting the number of people, vehicles and other objects, displaying crowd gatherings, monitoring security rate levels, measuring traffic, and displaying attempts to follow into the entrance control point. Ioimage's products can detect five behaviors, including intrusion detection, PTZ automatic tracking, abandoned object detection, illegal detention detection, and moving object detection.

The product of French company CitiLog is an automatic video event detection system, which is mainly used in the field of intelligent transportation. Citilog's core technology is dynamic image background adaptation technology and vehicle image tracking technology, which completely eliminates the influence of light, rain, snow and dust on the system, and can timely detect traffic events in the monitoring area, collect traffic data, and assist in traffic control.

The Institute of Automation of the Chinese Academy of Sciences started early in the research of intelligent video in China, and they have also achieved certain results. They mainly study visual monitoring of dynamic scenes, focusing on motion detection, tracking, gesture, gait recognition and biometric recognition. There are also many university laboratories invested in this field. For example, the Department of Computer Science at Tsinghua University mainly focuses on face recognition, active tracking of objects and other aspects in computer vision research.

2.3 Development Trends of Video Intelligence Products

Intelligent video surveillance is a monitoring method that analyzes the video image content of the monitoring scene based on computer vision technology, extracts key information in the scene, and forms corresponding events and alarms. It is a new generation of monitoring system based on video content analysis.

Video intelligent analysis is the newest and most promising direction in the surveillance field. With the continuous increase in investment, video intelligent surveillance products will definitely have a broader room for development.

Market: From Monopoly to Diversified Competition

In the past few years, as the intelligent market was in its infancy, there were not many companies entering this field, and a few products monopolized the market. With the changes in the international situation in recent years and the development of anti-terrorism, the intelligent identification market has become increasingly large. As the market is lucrative, more and more funds are entering this field. Latecomers will transform the intelligent identification market from profiteering to profit averaging with their financial advantages and technology introduction. Competition will become increasingly fierce, and only those companies that master core technologies and product standardization will become the biggest winners of the competition.

Users: From special departments to socialized

In the past, users of smart identification products were mostly concentrated in special departments or industries such as public security, security, and banking, and rarely involved in general fields. With the development of society and the severity of public security, more and more general industries need smart identification technology and products, such as smart homes, smart toys with face recognition functions, game interactions, etc. The transformation of the smart identification field from special industry customers to general consumers will be an irresistible trend.

Application: From isolated use to networked application

Today's intelligent recognition is almost entirely limited to single recognition, and not much can be connected to the Internet. However, with the development of network transmission technology, large-capacity, high-speed transmission will no longer be a distant dream. With the maturity of 3G technology, more and more intelligent recognition data will be able to be transmitted in real time via mobile phones.

3 ATM Video Surveillance System Based on Intelligent Analysis Technology

3.1 Project Background

With the in-depth development of my country's financial electronicization and the continuous advancement of the Golden Card Project, the application scope of bank self-service equipment and self-service banks in the financial field is becoming more and more extensive, and they are generally welcomed by depositors. Due to the publicity, convenience and special environment of bank self-service equipment (ATM) and self-service banks, it is difficult for customers, banks and society to establish an effective prevention mechanism. In recent years, criminal activities against bank self-service equipment and self-service banks have continued to increase.

Traditional ATM video surveillance systems mainly record surveillance videos. After an incident occurs, the video is used for evidence collection, dispute resolution, and case solving. Such surveillance systems guarantee the safe operation of ATMs to a certain extent, but they can only provide evidence collection after the incident, which often delays the best opportunity to resolve the incident. Some incidents are solved after they occur, which is time-consuming and laborious. Even if evidence can be found, the losses may be irreversible.

Statistics show that there is a lot of redundant information in ATM surveillance videos, which not only increases the burden on recording equipment and increases the consumption of storage resources, but also makes it inconvenient for subsequent staff to trace and analyze the videos.

How to ensure the safe operation of ATMs, prevent and reduce the risks for banks and ATM users, prevent ATMs from being deliberately damaged, prevent the use of ATMs for fraud, and effectively protect the interests of banks and depositors has become an urgent problem to be solved in today's financial field.

The ATM intelligent video surveillance system is a new generation of video surveillance system developed by Beijing Shengxun Electronics Co., Ltd. to effectively solve the above difficulties in traditional video surveillance. The project development is aimed at the application scenarios of financial self-service equipment, integrating the latest achievements of current motion detection, target tracking and behavior recognition and the latest development of hardware technology, to build a comprehensive software and hardware platform for video intelligent monitoring, and realize the identification and understanding of abnormal behaviors in financial ATM business monitoring scenarios.

ATM intelligent video monitoring systems have been successfully applied in some cities in China. Before the 2008 Beijing Olympic Games, the ATMs in the Olympic Village were equipped with the above intelligent video processing products and connected to the ATM monitoring network system in Beijing. The intelligent network monitoring of these ATMs has greatly improved the original monitoring efficiency and has been well received by users.

3.2 System composition

Traditional ATM monitoring systems include: cameras, video servers/network video recorders, customer terminals, etc. The ATM intelligent video monitoring system achieves a series of functions by adding the key technology of intelligent video processing module to the traditional ATM monitoring system. It can monitor the scene in real time according to the pre-set rules. When the preset rules are violated, it will issue an alarm message in time to remind the monitoring personnel to take effective measures.

The intelligent video processing module can realize intelligent processing of video data and communicate with other hardware devices.

The intelligent video processing module is mainly composed of the following functional modules (as shown in Figure 1): program storage module, video decoding processing module, DSP video data processing module, serial port processing module, network port output processing module, and video encoding processing module.

Figure 1 Schematic diagram of the intelligent video processing module

Video decoding processing module. This module uses a video decoding chip to decode the analog signal captured by the camera, convert it into a digital signal that can be processed by the DSP module, and then transmit the data to the DSP.

Program storage module. This module uses flash memory (Flash) to store the programs that the ATM intelligent video surveillance system needs to run.

DSP processing module. This module uses an embedded intelligent video processing algorithm to realize intelligent processing of video data. In this way, the video images of the ATM scene captured by the camera are processed in real time and intelligently, and can analyze abnormal behaviors related to ATM and issue alarm information in real time.

Serial port processing module, network port output processing module, video encoding processing module. The result of video data processed by DSP can be output to other devices in the form of video port, serial port, and network port. The video data processed by DSP is encoded by the video encoding chip and sent to the video port for output; the serial data can also be converted into serial port data by the conversion chip, connected to the serial port chip respectively, and then output through the serial port; the video processing results can also be output through the network port output module using a general network protocol.

Synchronous dynamic random access memory (SDRAM). Because the DSP module needs to use a variety of intelligent video processing algorithms to realize intelligent processing of video data, it will consume a lot of storage space. The ATM intelligent video monitoring system uses SDRAM as an auxiliary device to increase the storage space and ensure the real-time and effectiveness of video processing.

The intelligent video processing module is implemented through hardware circuits. The hardware circuit board can be loaded into the hardware devices of the traditional ATM monitoring system, such as cameras, video servers/network cameras, or the hardware circuit board can be used to implement independent devices with video intelligent processing functions.

There are three common ways to build an intelligent video monitoring system.

Method 1: By loading the intelligent video processing module on the camera in the ATM monitoring system, an intelligent video processing camera is formed. In this way, without changing the connection method of the traditional ATM monitoring system, the intelligent video processing camera can perform intelligent analysis and processing on the video data it captures.

Method 2: By loading the intelligent video processing module on the video server/recorder and other devices in the ATM video monitoring system, an ATM intelligent video monitoring system is formed (as shown in Figure 2). In this way, without changing the connection method of the traditional ATM monitoring system, the intelligent video server/recorder has the intelligent video analysis and processing function.

Figure 2 ATM intelligent video surveillance system

In the figure, the intelligent video server/recorder is connected to the camera through a video cable to collect analog video signals. The intelligent video server/recorder can automatically analyze the video captured by the camera and detect abnormal behavior around the ATM machine scene. At the same time, the intelligent video server/recorder has all the functions of the traditional video server/recorder and can encode and transmit the video (the recorder can store it). The intelligent video server can transmit the ATM video and the analysis results obtained through intelligent video analysis and processing - alarm information to the client terminal of the monitoring center through the network port. The client terminal is the back end of the entire ATM intelligent video monitoring system. The client terminal of the monitoring center is installed with a management program with video management function. The client management program can receive the information transmitted by the intelligent video server/recorder, decode and play the video information in time, and also provide playback, retrieval and other functions.

Method 3: By adding a hardware device with intelligent video processing function to the ATM video monitoring system. The device is embedded with the above-mentioned intelligent video processing module. The device can communicate with other hardware devices in the traditional intelligent video monitoring system through the network port, serial port, video port, etc.

3.3 Main technical features

The intelligent processing process of the ATM intelligent video surveillance system is as follows: 1. The camera captures data; 2. The area of ​​the moving target is obtained through the motion detection algorithm; 3. Based on the detection results, the target tracking algorithm will obtain a tracking result for each target; 4. The behavior analysis results and abnormal behavior detection results of the target in the scene are obtained through the tracking results and the feature extraction of the original video.

The main technologies for realizing intelligent processing of video data include: motion detection, motion tracking, abnormal behavior detection, etc.

1. Motion detection

Motion detection is to extract the changing area from the background image in the video sequence image. Motion detection is the first and key step in the entire intelligent video surveillance processing. Whether the moving target can be effectively detected is very important for post-processing such as target tracking, classification and abnormal behavior detection, and is a reference for subsequent processing processes (the subsequent processing process only considers the pixels corresponding to the moving area in the image).

2. Target tracking

Target tracking plays an important role in the field of video intelligent monitoring. It is between the underlying motion detection and the high-level abnormal behavior recognition. Video tracking is the process of establishing a corresponding relationship between the targets to be tracked in continuous video frames. Although in some cases, tracking does not involve motion detection, in most cases it is based on motion detection and provides a basis for high-level behavior recognition.

In the ATM intelligent video surveillance system, two aspects of tracking are mainly achieved:

1) Tracking people in the scene as a whole target. When a person enters the monitored ATM scene, the system can automatically track the person and record the location of the target at each moment. This information will provide important information for high-level abnormal behavior recognition.

2) Tracking part of the body of the person in the scene as a target. Tracking part of the body of the person in the scene mainly includes the hands and head. This is necessary for the ATM intelligent video surveillance system. Because many abnormal behaviors occur on the ATM, such as keyboard film, card port sealing, ATM knocking, etc., the movement of the hands and head can provide a lot of information. For the tracking of fast and repetitive actions such as violently hitting the ATM, wielding blunt objects, and fighting, due to the violent movement of the limbs, a more complex motion model and tracking algorithm are adopted.

In short, the tracking in the ATM intelligent video surveillance system can not only correctly track people as a whole target, but also track important limbs of people, including people's hands and heads. The dual tracking mode provides sufficient information for abnormal behavior tracking.

3. Abnormal behavior detection

Abnormal behavior detection is the core content of the ATM intelligent video surveillance system. Abnormal behavior recognition is carried out through flexible and effective behavior recognition algorithms. The ATM intelligent video surveillance system has established different abnormal behavior recognition algorithms according to different abnormal behaviors. From the perspective of perceptual psychology, we know that behavior obeys the hierarchical structural characteristics, and the abnormal recognition algorithm is implemented based on this theory. Behavior is similar to sentences in language processing, which can be classified and organized and placed on different abstract layers. At the same time, we divide the behavior again, that is, divide the behavior into sub-behaviors, and further divide the sub-behaviors into primitive behaviors (i.e., indivisible behaviors). There is a temporal constraint relationship between the primitive sequence and the sub-behavior sequence of the behavior. The behavior in the real scene is composed of multiple behavior sequences that meet the above hierarchical characteristics. Therefore, the recognition algorithm used in the ATM intelligent video surveillance system can not only express the internal structural characteristics of the behavior, but also capture its temporal constraint relationship.

3.4 System Functions

The ATM intelligent video surveillance system provides smarter monitoring, more efficient alarms and safer protection for ATM applications. The specific functions that the system can achieve are as follows:

1. Detection of normal and abnormal behaviors of people in the scene

The ATM intelligent video surveillance system can analyze the operations occurring on the ATM in real time through monitoring equipment, and determine whether the human behavior is normal or abnormal through intelligent video processing algorithms. The system can detect the following abnormal behaviors:

1) Abnormal behaviors of keyboard film and fake keyboard installation

Keyboard film or fake keyboard installation refers to artificially adding a film to the keyboard or adding a thin fake keyboard to the real keyboard, which can illegally intercept the password when the withdrawer enters the password. Usually this criminal behavior is combined with fake card swallowing, first obtaining the withdrawer's password and then swallowing the withdrawer's card, thereby achieving the purpose of the crime.

The ATM intelligent video surveillance system can accurately identify abnormal behaviors such as adding a film to the ATM keyboard or installing a fake keyboard. When abnormal behaviors are detected, an alarm message will be prompted, and the alarm message will be transmitted to the client terminal of the monitoring center through the network. The client terminal can obtain the alarm message in real time to prevent the password of the cashier from being illegally intercepted after the cashier enters the password.

2) Abnormal behaviors

of adding a small device to the card slot There are two common abnormal behaviors of adding a small device to the card insertion point. One is to fix a fake socket on the original position of the real socket on the ATM machine. The cashier inserts the bank card into the fake socket without knowing it. There is a card information reading and writing device inside the fake card slot, which can copy all the information on the card.

Normal card slot	Added card slot

Another method is to add a hook to the card slot. When the cashier inserts the card into the ATM, the card cannot be inserted or pulled out. Usually, this situation is matched with a fake announcement. The fake announcement will prompt "Contact XX by phone after the card is swallowed". The phone pretends to be a bank customer service staff, obtains the card number and password of the cashier, and then tells the cashier to retrieve the card at XX Bank the next day. When the cashier leaves, the criminal takes out the card and takes the cash according to the password information obtained.

The ATM intelligent video monitoring system can identify abnormal behavior of artificially installing small devices at the card insertion point. When abnormal behavior is detected, an alarm message will be prompted, and the alarm information can be transmitted to the client terminal of the monitoring center through the network. The client terminal can obtain the alarm information in real time and deal with it in time, thereby preventing the cashier from illegally obtaining the card number or causing false card spitting when inserting the card.

3) Abnormal behavior of cash outlet sealing

Common abnormal behavior of cash outlet is to seal the cash outlet with plastic cards, reminder signs (such as warm reminders) or glue. After the cashier leaves, the seal is scraped off and the cash is taken away.

Use "warm reminder" to seal the cash outlet	Use glue to block the cash outlet

The ATM intelligent video monitoring system can identify abnormal behaviors such as the cash outlet being sealed and detect such abnormal behaviors. When such behaviors are detected, an alarm message will be prompted, and the alarm message will be transmitted to the monitoring center client terminal through the network. The client terminal can obtain the alarm message in real time, thereby preventing the withdrawal of money from being illegally obtained by vandals due to illegal sealing of the cash outlet.

4) Abnormal behaviors of smashing ATMs

The ATM intelligent video monitoring equipment can identify abnormal behaviors such as kicking, hitting, smashing, using tools to hit the ATM violently, or using tools to pry open the ATM body in an attempt to illegally obtain cash in the ATM cash box. When such abnormal behaviors are detected, an alarm message will be prompted, and the alarm message will be transmitted to the monitoring center client terminal through the network. The client terminal can obtain the alarm message in real time, thereby preventing the occurrence of ATM vandalism incidents and ensuring the safety of ATMs.

5) Abnormal behaviors of repetitive rapid actions

The ATM intelligent video monitoring system can identify abnormal behaviors such as violent impacts on ATMs, wielding blunt instruments, and fighting. ATM intelligent video surveillance equipment can accurately detect abnormal behaviors such as violently hitting ATMs, wielding blunt instruments, and fighting at ATMs by identifying the rapid movement of the whole body or parts of the human body, thereby preventing the occurrence of criminal incidents. When such abnormal behaviors are detected, an alarm message will be prompted, and the alarm message will be transmitted to the client terminal of the monitoring center through the network, and the client terminal can obtain the alarm message in real time. In this way, the occurrence of illegal behaviors can be terminated in advance through local warnings and deterrence at the ATM, and the police can be notified to deal with the situation on the scene in a timely manner to reduce the loss of people, money, and materials.

6) Abnormal behavior of posting or installing advertising boxes on ATMs

The ATM intelligent video surveillance system can identify abnormal posting or installation of advertising boxes on ATMs. When such abnormal behaviors are detected, an alarm message will be prompted, and the alarm message will be transmitted to the client terminal of the monitoring center through the network, and the client terminal can obtain the alarm message in real time, thereby preventing the withdrawer from being confused and induced by false notices and false advertisements such as "transfer to a designated account" that are illegally posted, causing unnecessary economic losses.

7) Abnormal packet loss behavior

The ATM intelligent video surveillance system can identify the operator who places the items he carries on the ATM or somewhere near the ATM, and forgets to take them away after the transaction, resulting in the frequent loss of the left-behind items. When such abnormal behavior is detected, an alarm message will be issued in real time, and the ATM user will be automatically prompted by voice that the item has been forgotten, or the alarm message will be transmitted to the monitoring center, and the monitoring center will give a prompt through intercom. This effectively solves the unnecessary economic losses caused to ATM users due to forgotten items.

2. Real-time trigger recording

The ATM intelligent video surveillance system can trigger the start and stop of ATM monitoring recording by real-time monitoring of human behavior. Trigger storage only saves the information of intelligent monitoring under normal circumstances. When someone enters the monitoring scene of the ATM, the recording device is triggered to start recording. When the person leaves the monitoring scene of the ATM after the operation, the recording device is triggered to stop recording, which can save storage space on the hard disk and improve the efficiency of ATM monitoring.

4 The role of intelligent analysis in ATM video surveillance

The above-mentioned ATM intelligent video surveillance system is based on ordinary network video surveillance and applies intelligent recognition and analysis technology to ATM video surveillance. In addition to the well-known advantages of network video surveillance, the intelligent video surveillance system can also intelligently analyze and process the real-time images of ATM machines. It provides the financial system with an effective protection measure against ATM crimes, and can also assist relevant departments in bringing criminals to justice.

The outstanding advantages of intelligent analysis technology applied to ATM monitoring are mainly reflected in the following aspects.

4.1 Change post-event investigation to pre-event warning

This is the most valuable embodiment of intelligent identification.

Currently, the camera monitoring widely used in banks, shopping malls and other occasions usually only records videos for post-event evidence collection. It loses the basic value of the image (a dynamic, real-time medium), just like turning live broadcast into video recording, which is different from people's real demand for video systems. Countless facts have proved that post-event investigation is necessary, but pre-event warning is more important. It can suppress some major event risks in the bud and avoid making up for the loss. Therefore, an important research direction of the intelligent monitoring system is to shift from post-event analysis to pre-event prevention, monitor potential security threats, and improve early warning capabilities. For example, an object of unknown origin is found to appear in the security management area for a long time, or a suspicious person is found to be wandering in the security management area, or a person is found to have special attention or abnormal behavior towards the protection target. For such events with potential threats, the video surveillance system should be able to capture them in time and make intelligent judgments on whether to intervene in the event.

4.2 Reliable all-weather monitoring

It is easy for people to get tired when staring at the screen, and it is generally impossible to display the TV wall one-to-one, so there will inevitably be many pictures that are not observed for a long time. Intelligent recognition analyzes the content of the picture 24 hours a day, completely replacing the boring labor of staring at the screen, and truly achieving all-weather monitoring. The intelligent video surveillance system can effectively suppress a series of problems caused by people as the subject of behavior, realize true all-weather monitoring, improve the accuracy of alarms, reduce the phenomenon of missed analysis or misanalysis, and shorten the response time, increase the response speed, and form more effective on-site data. Before the security threat occurs, it can prompt security personnel to pay attention to the relevant monitoring pictures and take precautions against potential threats.

4.3 Improve monitoring quality

When the monitoring image is blurred, blocked, or shifted, intelligent analysis can accurately detect the problem. Through fault reports, administrators can quickly find the fault and avoid the situation where the image is unclear or the video is blurred for a long time.

4.4 Improvement of the overall quality of video recording

In the absence of intelligent analysis, the activation of video recording is generally divided into manual, timed automatic, alarm linkage, motion detection and other methods, which require long recording time and occupy a large amount of hard disk space. According to statistics, more than 90% of surveillance videos are garbage resources and do not need to be stored at all, but the machine cannot know whether they are valuable, so they have to be stored for later investigation. The activation of intelligent recognition can apply the machine's interpretation of image content as a condition for starting video recording. For example, VMD in intelligent recognition has a higher level of recognition and accuracy than ordinary motion detection. Intelligent analysis can store images that are interpreted as valuable, thereby reducing garbage videos and greatly saving hard disk space.

The improvement of the overall quality of video recording by intelligent analysis is also reflected in the subsequent retrieval. For videos of the same time period, it is generally necessary to manually watch all or play quickly or drag and play to discover the events that occurred during the period, while intelligent analysis can quickly locate the relative position of the event for playback.

5 Conclusion

The ATM intelligent video surveillance system provides an ATM security solution that can provide real-time alarms and advance warnings by effectively combining video intelligent analysis technology with ATM traditional monitoring technology. It replaces the traditional ATM security solution of post-event evidence collection and advertising safety prompts, making ATM monitoring a leap from traditional monitoring to intelligent video monitoring.

The promotion and application of video intelligent analysis technology can drive the exploration of advanced applications for ATM monitoring in China, and can also drive the further research of intelligent recognition technology by relevant domestic units. With the continuous improvement of intelligent analysis technology, intelligent video monitoring will show a broader development space. In addition to being well applied in ATM monitoring, intelligent video analysis technology can also be promoted to transportation, museums, monuments, sports venues, aviation venues, parking lots, shopping malls and other places, thereby improving the application field of intelligent monitoring, making our living environment safer and more harmonious, and truly achieving long-term peace and stability for the country and society.