Introduction to Algorithm Packages

1

Airport Event Detection (Edge)

Airport event detection

Analyzes video streams to detect and report events outside the airport, including aircraft arrival and departure, cabin door opening and closing, passenger stairs attachment and detachment, and catering truck arrival and departure.

2

Crowd Density Monitoring (Edge)

Crowd gathering detection

Estimates the density of crowds in public places to detect possible social incidents, so that the authority can take prompt measures to prevent potential dangers. If the estimated number of people in an image exceeds a preset threshold (configurable), the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

3

Outdoor people counting

4

People in factory exceed capacity

If the number of people in a specified area of a factory exceeds a preset threshold, the algorithm generates an alarm, captures images, and records the occurrence time for case reporting, ensuring factory security.

5

Individual Behavior Detection (Edge)

Detection of smoking in hazardous areas

Analyzes real-time video streams to detect smokers in hazardous areas using the pedestrian detection algorithm and human skeleton key point detection algorithm; continuously tracks and analyzes the object, confirms the specific behavior of the object, and generates an alarm for a confirmed smoker.

6

Phone calling detection

Detects people making phone calls where such behavior is forbidden, such as at gas stations. If such an event is detected, the algorithm captures onsite images and records the occurrence time for further handling.

7

Fight Detection (Edge)

Fight detection

Detects people having a fight with fists or weapons (excluding small confrontation such as clothes grabbing and verbal confrontation). If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

8

Crowd Counting Using High-Altitude Cameras (Edge)

Crowd counting using high-altitude cameras

Analyzes the crowd density of RTSP video streams output by edge cameras. The algorithm triggers alarms based on the alarm mode.

9

Speeding Detection Using High-Altitude Cameras (Edge)

Speeding detection

Estimates the driving speed of a vehicle. When the estimated speed exceeds a specified value, an event is reported.

10

Missing Manhole Cover Detection From a Dynamic Perspective (Edge)

Missing manhole cover detection using UAVs

Detects missing manhole covers from a dynamic perspective. It triggers an alarm if detecting missing manhole covers in the monitored areas.

11

Urban Mobile Trash Detection (Edge)

Mobile trash detection

Detects household wastes (such as plastic bags and leftovers) in places where they do not belong. It triggers an alarm if detecting any household wastes. Vehicle-mounted cameras can be used for such purposes.

12

Target Attribute Recognition (Edge)

Glass wearing detection

Identifies pedestrian attributes such as glasses, hat, gender, approximate age, backpack, handbag, sling bag, tops color, tops style (long or short sleeves), bottoms color, bottoms style, and face orientation in a monitored area, and generates an alarm.

13

Hat wearing detection

14

Gender detection

15

Age recognition

16

Backpack detection

17

Handbag detection

18

Sling bag detection

19

Tops color detection

20

Long/short sleeve detection

21

Bottoms color detection

22

Bottoms style detection

23

Face orientation recognition

24

Individual Action Recognition (Edge)

Climbing detection

Analyzes real-time video streams to detect climbing behavior. Specifically, the algorithm does this by analyzing the body shapes of people across multiple video frames using the human skeleton key point detection algorithm. It helps to improve human and facility safety.

25

Hand waving-for-help detection

Analyzes real-time video streams to detect hand waving-for-help. Specifically, the algorithm does this by analyzing the body shapes of people across multiple video frames using the human skeleton key point detection algorithm. It helps to accelerate emergency response and improve human safety.

26

Fall detection

Detects human falling events in camera footage, such as falling down due to slipping or a health problem. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

27

Shared Bicycle Detection (Edge)

Misparked shared bicycle detection

Detects misparked shared bicycles in a monitored area and generates alarms. The personnel in charge can be notified of parking violations in a timely manner, and public service staff can be assigned to handle problems quickly.

28

Smoke and Fire Detection (Edge)

Waste/leave burning detection

Detects visible fires (sometimes with heavy smoke), such as the burning of garbage and leaves on city streets or campuses. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

29

Fire detection

Detects visible fires (sometimes with heavy smoke). If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

30

Smoke detection

31

Construction site dust detection

Detects large-scale and strong dust on construction sites. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

32

Factory hot work environment monitoring

Detects visible fire in hot work environments. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

33

Trash Can Anomaly Detection (Edge)

Uncovered food waste detection

Detects uncovered food waste in real time. If an uncovered trash can is detected, the algorithm generates an alarm and notifies related personnel to handle the problem.

34

Littering Construction Waste Detection (Edge)

Littering construction waste detection

Detects slag remains that are piled on the streets. The algorithm analyzes videos streams to discover suspected slags on roads and streets. If slag piles are detected, the algorithm captures onsite images with timestamp and records them as evidence for case reporting.

35

Clothes Airing Violation Detection (Edge)

Hanging clothes in public

Detects hanging clothes on telegraph poles or shelves in public. The algorithm analyzes video streams to detect clothes hanging in public. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

36

Missing Manhole Cover Detection (Edge)

Damaged manhole cover detection

Identifies manholes whose covers are broken or missing on streets. The algorithm analyzes video streams. If a manhole whose cover is broken or missing appears in the video, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

37

Urban Roadside Stall Detection (Edge)

Out-of-store sales detection

Some grocery store owners may sell goods on the sidewalks outside their stores or in front of other stores. This algorithm analyzes videos to find images that include stalls, tables, or chairs at the doors of stores. The algorithm captures target images, records the time, and reports captured images along with other details to the upper-level administrative platform for case reporting.

38

Sidewalk sales detection

Analyzes videos to find images that include illegal sidewalk sales activities. Sidewalk sales include unlicensed business activities such as waste collection on sidewalks or at other public places. If such an activity is detected, the algorithm captures target images, records the time, and reports captured images along with other details to the upper-level platform for case reporting.

39

Unlicensed moving stall detection

40

Advertisement Violation Detection (Edge)

Immoderate banners detection

Detects banners in a specified monitored area in public places, such as stations, squares, and streets. If immoderate banners are detected, the algorithm generates an alarm so that the banner can be removed in a timely manner. This will help ensure a clean and tidy city outlook.

41

Exposed Bare Soil Detection (Edge)

Exposed bare soil at construction site

Detects construction sites that have a large amount of bare soil uncovered. The algorithm analyzes video streams to detect exposed bare soil. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

42

Messy Piles Detection (Edge)

Messy piles detection

Detects roads and streets for disordered materials, such as timber, sandstone, and crates. The algorithm analyzes video streams to detect improperly piled materials. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

43

Construction waste detection

Detects construction waste such as cement bags and materials at the construction site and generates alarms. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting.

44

Messy piles of building materials detection

Detects messy piles of building materials such as cement bags, wood, and mounds, and generates alarms. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting.

45

Urban Trash Detection (Edge)

Exposed trash detection

Detects packaging waste, exposed trash, and dirty green spaces in public areas. If such domestic waste is detected, the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

46

Trash overflow detection

47

Unclean roads detection

48

Green space maintenance

49

Unclean riverbank detection

If domestic waste such as garbage bags and water bottles are detected along riverbanks, the algorithm generates an alarm, captures images, and records the occurrence time for case reporting.

50

Littered food waste detection

If littered food waste is detected, the algorithm generates an alarm and notifies related personnel to handle the problem.

51

Overflowing Garbage Can Detection in Cities (Edge)

Overflowing garbage can detection

If an overflowing garbage can is detected, the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

52

Umbrella Violation Detection (Edge)

Umbrella violation detection

Detects large sunshade umbrellas that are open in public places and affect road traffic and city outlook. If such a large sunshade umbrella is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

53

Vehicle Classification (Edge)

Vehicle classification

Outputs information about vehicles in videos and accurately provides their license plate information based on the specified area. Identifies vehicle types in video footage and reports alarm events and video snapshots with timestamps in a timely manner. Supported vehicle types include cars, dump trucks, cement mixers, trailers, and dump trucks.

54

Vehicle Counting Using High-Altitude Cameras (Edge)

Vehicle counting

Counts the number of vehicles on a road. Periodically reports the number for case reporting.

55

Water Level Gauge Reading (Edge)

Urban flood gauge recognition

Reads water level gauges in urban streets to monitor changes in the volume of water, so that the authority can take prompt measures to prevent floods and waterlogging.

Images of the water level readings can be provided for further confirmation by humans. An alarm is generated when the water level exceeds a preset threshold. The water level readings and relevant evidence are saved in a database, and the water level processing module can quickly access such data and generate near real-time water level information at each monitoring point. Such information can be used to support decision-making.

56

River water gauge reading

Reads water level gauges in reservoirs, lakes, and rivers to monitor changes in the volume of water, so that the authority can take prompt measures to prevent floods and waterlogging.

Images of the water level readings can be provided for further confirmation by humans. An alarm is generated when the water level exceeds a preset threshold. The water level readings and relevant evidence are saved in a database, and the water level processing module can quickly access such data and generate near real-time water level information at each monitoring point. Such information can be used to support decision-making.

57

Drainage Outlet Monitoring (Edge)

Drainage outlet monitoring

Monitors the flow and discharge content at various drainage or sewage outlets in the city or water systems, for purposes like flood warning, detection of illegal discharging for pollution prevention, and generates alarms based on preset rules.

58

Floating Debris Detection (Edge)

Floating debris detection

Detects floating debris, such as plastic bags and water bottles, on rivers. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

59

Vessel Intrusion Detection (Edge)

Vessel intrusion detection

Detects vessels in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

60

Waterlogging Detection (Edge)

Standing water detection

Detects standing water on roads in camera-captured images. Standing water is dangerous, so it is important to detect and get rid of standing water as soon as possible. If standing water is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

61

Tunnel waterlogging detection

Detects standing water in tunnels. Standing water is dangerous, so it is important to detect and get rid of standing water as soon as possible. If standing water is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

1

Abandoned Object Detection (Edge)

Abandoned object detection

Recognizes objects (for example, packages and luggage trolleys) left or abandoned in public places such as transport stations and public campuses. Supported object types include packages and trolley cases. If such objects are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

2

Passageway Obstruction Detection (Edge)

Detection of fire access route blockage by vehicles

Detects fire passage blockages and sends alarms with license plates immediately so that the administrative team can handle them in a timely manner. This helps to keep the fire passage clear at all times for fire safety.

3

Detection of fire escape blockages

4

People Flow Counting (Edge)

People flow counting

Counts the number of people in (or entering and exiting) a certain area of interest during specified periods of time. This algorithm is typically used in campuses or stores to identify the most visited areas, providing insights to better serve both internal and external customers.

5

E-bike Rider Helmet Detection (Edge)

E-bike rider helmet detection

Detects e-bike riders not wearing helmet in streets. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

6

Intrusion Detection (Edge)

People intrusion detection

Detects people and vehicle intrusion in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

7

Vehicle intrusion detection

8

Intrusion in ROIs

Detects intrusions in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

9

Lawn trampling detection

By analyzing video, the algorithm can help check whether there are people who are walking across grass in a specified area. The grass area needs to be configured. If the algorithm identifies lawn trampling behavior, it captures images, records the occurrence time, and produces captured images as evidence.

10

Factory intrusion detection

Detects intrusion events in key areas and dangerous areas in a factory. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

11

Non-motorized Vehicle Detection (Edge)

Detection of misparked non-motorized vehicle

Detects non-motorized vehicles parking outside the designated areas in camera footage. If such an event is detected, the algorithm further checks the length of time of the parking violation, records images and the occurrence time, generates an alarm, and sends the recorded information to the upper-level platform for further handling.

12

Detection of non-motorized vehicle

Detects non-motorized vehicles that appear in places where non-motorized vehicles are not allowed. The algorithm selects areas where non-motorized vehicles should not be parked. If non-motorized vehicles are detected in those areas, this algorithm further records images and the time, generates an alarm, and reports the recorded information to the upper-level platform so that the platform can generate case with received evidence.

13

E-bike-in-building detection

Detects e-bikes in staircases or elevators. The algorithm selects areas where e-bikes should not be parked. If e-bikes are detected in those areas, this algorithm further records images and the time, generates an alarm, and reports the recorded information to the upper-level platform so that the platform can generate case with received evidence.

14

Non-motorized vehicles lane violation

Detects non-motorized vehicles occupying motor vehicle lanes in a monitored area. If such events are detected, an alarm is generated and relevant departments can be notified in a timely manner to handle problems quickly.

15

Wrong-Way Driving Detection (Edge)

Wrong-way driving detection

Detects wrong-way driving events. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

16

Safety Suite Detection (Edge)

No-workwear detection

Detects workers who are not wearing safety helmets or reflective clothing when working in potentially hazardous positions and generates warnings when such events are detected. This helps to improve worker safety in factories, construction sites, and many other places. It also helps to improve workers' safety awareness.

17

No-safety helmet detection

18

Staff On-Duty Detection (Edge)

Employee absence detection

The algorithm does this by counting the number of people in specific areas of interest captured in camera footage. If the number of people detected is less than the preset number (configurable), this algorithm determines that some employees working at key positions are off duty when they are not supposed to. When such an event is detected, this algorithm automatically records original images and the occurrence time, and generates an alarm.

19

Abnormal Parking Detection (Edge)

Motor vehicle parking violation

Detects motor vehicles parking outside the designated areas for a time that exceeds the allowed period. If such an event is detected, the algorithm further checks the length of time of the parking violation, records images and the occurrence time, generates an alarm, and sends the recorded information to the upper-level platform for further handling.

20

Vehicle Detection (Edge)

License plate detection

Outputs information about vehicles in videos and accurately provides their license plate information based on the specified ROI.

21

Mask Wearing Detection (Edge)

Mask wearing detection

Detects people who do not wear masks in a monitored area and generates alarms. If such a person is detected, the algorithm generates an alarm and this person will be denied entry. Alternatively, related personnel will be sent to handle the problem.

1

Airport Event Detection (Cloud)

Airport event detection

Analyzes video streams to detect and report events outside the airport, including aircraft arrival and departure, cabin door opening and closing, passenger stairs attachment and detachment, and catering truck arrival and departure.

2

Crowd Density Monitoring (Cloud)

Crowd gathering detection

Estimates the density of crowds in public places to detect possible social incidents, so that the authority can take prompt measures to prevent potential dangers. If the estimated number of people in an image exceeds a preset threshold (configurable), the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

3

Outdoor people counting

4

People in factory exceed capacity

If the number of people in a specified area of a factory exceeds a preset threshold, the algorithm generates an alarm, captures images, and records the occurrence time for case reporting, ensuring factory security.

5

Individual Behavior Detection (Cloud)

Detection of smoking in hazardous areas

Analyzes real-time video streams to detect smokers in hazardous areas using the pedestrian detection algorithm and human skeleton key point detection algorithm; continuously tracks and analyzes the object, confirms the specific behavior of the object, and generates an alarm for a confirmed smoker.

6

Phone calling detection

Detects people making phone calls where such behavior is forbidden, such as at gas stations. If such an event is detected, the algorithm captures onsite images and records the occurrence time for further handling.

7

Crowd Counting Using High-Altitude Cameras (Cloud)

Crowd counting using high-altitude cameras

Analyzes the crowd density of RTSP video streams output by edge cameras. The algorithm triggers alarms based on the alarm mode.

8

Fight Detection (Cloud)

Fight detection

Detects people having a fight with fists or weapons (excluding small confrontation such as clothes grabbing and verbal confrontation). If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

9

Crowd Counting Using High-Altitude Cameras (Cloud)

Crowd counting using high-altitude cameras

Analyzes the crowd density of RTSP video streams output by edge cameras. The algorithm triggers alarms based on the alarm mode.

10

Speeding Detection Using High-Altitude Cameras (Cloud)

Speeding detection

Estimates the driving speed of a vehicle. When the estimated speed exceeds a specified value, an event is reported.

11

Wrong-Way Driving Detection Using High-Altitude Cameras (Cloud)

Wrong-way driving detection

Detects wrong-way driving events. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

12

Missing Manhole Cover Detection From a Dynamic Perspective (Edge)

Missing manhole cover detection from a dynamic perspective

Detects missing manhole covers from a dynamic perspective. It triggers an alarm if detecting missing manhole covers in the monitored areas.

13

Target Attribute Recognition (Cloud)

Glass wearing detection

Identifies pedestrian attributes such as glasses, hat, gender, approximate age, backpack, handbag, sling bag, tops color, tops style (long or short sleeves), bottoms color, bottoms style, and face orientation in a monitored area, and generates an alarm.

14

Hat wearing detection

15

Gender detection

16

Age recognition

17

Backpack detection

18

Handbag detection

19

Sling bag detection

20

Tops color detection

21

Long/short sleeve detection

22

Bottoms color detection

23

Bottoms style detection

24

Face orientation recognition

25

Individual Action Recognition (Cloud)

Climbing detection

Analyzes real-time video streams to detect climbing behavior. Specifically, the algorithm does this by analyzing the body shapes of people across multiple video frames using the human skeleton key point detection algorithm. It helps to improve human and facility safety.

26

Hand waving-for-help detection

Analyzes real-time video streams to detect hand waving-for-help. Specifically, the algorithm does this by analyzing the body shapes of people across multiple video frames using the human skeleton key point detection algorithm. It helps to accelerate emergency response and improve human safety.

27

Fall detection

Detects human falling events in camera footage, such as falling down due to slipping or a health problem. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

28

Shared Bicycle Detection (Cloud)

Misparked shared bicycle detection

Detects misparked shared bicycles in a monitored area and generates alarms. The personnel in charge can be notified of parking violations in a timely manner, and public service staff can be assigned to handle problems quickly.

29

Smoke and Fire Detection (Cloud)

Waste/leave burning detection

Detects visible fires (sometimes with heavy smoke), such as the burning of garbage and leaves on city streets or campuses. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

30

Fire detection

Detects visible fires (sometimes with heavy smoke). If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

31

Smoke detection

32

Construction site dust detection

Detects large-scale and strong dust on construction sites. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

33

Factory hot work monitoring

Detects visible fire during hot work. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

34

Trash Can Anomaly Detection (Cloud)

Uncovered food waste detection

Detects uncovered food waste in real time. If an uncovered trash can is detected, the algorithm generates an alarm and notifies related personnel to handle the problem.

35

Littering Construction Waste Detection (Cloud)

Littering construction waste detection

Detects slag remains that are piled on the streets. The algorithm analyzes videos streams to discover suspected slags on roads and streets. If slag piles are detected, the algorithm captures onsite images with timestamp and records them as evidence for case reporting.

36

Clothes Airing Violation Detection (Cloud)

Hanging clothes in public

Detects hanging clothes on telegraph poles or shelves in public. The algorithm analyzes video streams to detect clothes hanging in public. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

37

Missing Manhole Cover Detection (Edge)

Damaged manhole cover detection

Identifies manholes whose covers are broken or missing on streets. The algorithm analyzes video streams. If a manhole whose cover is broken or missing appears in the video, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

38

Urban Roadside Stall Detection (Cloud)

Out-store sales detection

Some grocery store owners may sell goods on the sidewalks outside their stores or in front of other stores. This algorithm analyzes videos to find images that include stalls, tables, or chairs at the doors of stores. The algorithm captures target images, records the time, and reports captured images along with other details to the upper-level administrative platform for case reporting.

39

Sidewalk sales detection

Analyzes videos to find images that include illegal sidewalk sales activities. Sidewalk sales include unlicensed business activities such as waste collection on sidewalks or at other public places. If such an activity is detected, the algorithm captures target images, records the time, and reports captured images along with other details to the upper-level platform for case reporting.

40

Unlicensed moving stall detection

41

Advertisement Violation Detection (Cloud)

Immoderate banners detection

Detects banners in a specified monitored area in public places, such as stations, squares, and streets. If immoderate banners are detected, the algorithm generates an alarm so that the banner can be removed in a timely manner. This will help ensure a clean and tidy city outlook.

42

Exposed Bare Soil Detection (Cloud)

Exposed bare soil at construction site

Detects construction sites that have a large amount of bare soil uncovered. The algorithm analyzes video streams to detect exposed bare soil. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

43

Messy Piles Detection (Cloud)

Messy piles detection

Detects roads and streets for disordered materials, such as timber, sandstone, and crates. The algorithm analyzes video streams to detect improperly piled materials. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

44

Construction waste detection

Detects construction waste such as cement bags and materials at the construction site and generates alarms. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting.

45

Messy piles of building materials detection

Detects messy piles of building materials such as cement bags, wood, and mounds, and generates alarms. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting.

46

Urban Trash Detection (Cloud)

Exposed trash detection

Detects packaging waste, exposed trash, and dirty green spaces in public areas. If such domestic waste is detected, the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

47

Trash overflow detection

48

Unclean roads detection

49

Green space maintenance

50

Unclean riverbank detection

If domestic waste such as garbage bags and water bottles are detected along riverbanks, the algorithm generates an alarm, captures images, and records the occurrence time for case reporting.

51

Littered food waste detection

If littered food waste is detected, the algorithm generates an alarm and notifies related personnel to handle the problem.

52

Overflowing Garbage Can Detection in Cities (Edge)

Overflowing garbage can detection

If an overflowing garbage can is detected, the algorithm captures images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

53

Umbrella Violation Detection (Cloud)

Umbrella violation detection

Detects large sunshade umbrellas that are open in public places and affect road traffic and city outlook. If such a large sunshade umbrella is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

54

Vehicle Classification (Cloud)

Vehicle classification

Outputs information about vehicles in videos and accurately provides their license plate information based on the specified area. Identifies vehicle types in video footage and reports alarm events and video snapshots with timestamps in a timely manner. Supported vehicle types include cars, dump trucks, cement mixers, trailers, and dump trucks.

55

Vehicle Counting Using High-Altitude Cameras (Cloud)

Vehicle counting

Counts the number of vehicles on a road. Periodically reports the number for case reporting.

56

Water Level Gauge Reading (Cloud)

Urban flood gauge recognition

Reads water level gauges in urban streets to monitor changes in the volume of water, so that the authority can take prompt measures to prevent floods and waterlogging.

Images of the water level readings can be provided for further confirmation by humans. An alarm is generated when the water level exceeds a preset threshold. The water level readings and relevant evidence are saved in a database, and the water level processing module can quickly access such data and generate near real-time water level information at each monitoring point. Such information can be used to support decision-making.

57

River water gauge reading

Reads water level gauges in reservoirs, lakes, and rivers to monitor changes in the volume of water, so that the authority can take prompt measures to prevent floods and waterlogging.

Images of the water level readings can be provided for further confirmation by humans. An alarm is generated when the water level exceeds a preset threshold. The water level readings and relevant evidence are saved in a database, and the water level processing module can quickly access such data and generate near real-time water level information at each monitoring point. Such information can be used to support decision-making.

58

Drainage Outlet Monitoring (Cloud)

Drainage outlet monitoring

Monitors the flow and discharge content at various drainage or sewage outlets in the city or water systems, for purposes like flood warning, detection of illegal discharging for pollution prevention, and generates alarms based on preset rules.

59

Floating Debris Detection (Cloud)

Floating debris detection

Detects floating debris, such as plastic bags and water bottles, on rivers. If such events are detected, the algorithm captures onsite images and the occurrence time for case reporting.

60

Vessel Intrusion Detection (Cloud)

Vessel intrusion detection

Detects vessels in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

61

Waterlogging Detection (Cloud)

Standing water detection

Detects standing water on roads in camera-captured images. Standing water is dangerous, so it is important to detect and get rid of standing water as soon as possible. If standing water is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

62

Tunnel waterlogging detection

Detects standing water in tunnels. Standing water is dangerous, so it is important to detect and get rid of standing water as soon as possible. If standing water is detected, this algorithm captures the images, records the occurrence time, and sends the captured images along with other details to the upper-level platform for case reporting.

1

Abandoned Object Detection (Cloud)

Abandoned object detection

Recognizes objects (for example, packages and luggage trolleys) left or abandoned in public places such as transport stations and public campuses. Supported object types include packages and trolley cases. If such objects are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

2

Fire Passage Obstruction Detection (Cloud)

Detection of fire access route blockage by vehicles

Detects fire passage blockages and sends alarms with license plates immediately so that the administrative team can handle them in a timely manner. This helps to keep the fire passage clear at all times for fire safety.

3

Detection of fire escape blockages

4

People Flow Counting (Cloud)

People flow counting

Counts the number of people in (or entering and exiting) a certain area of interest during specified periods of time. This algorithm is typically used in campuses or stores to identify the most visited areas, providing insights to better serve both internal and external customers.

5

E-bike Rider Helmet Detection (Cloud)

E-bike rider helmet detection

Detects e-bike riders not wearing helmet in streets. If such an event is detected, the algorithm captures onsite images with timestamps and records them as evidence for case reporting.

6

Intrusion Detection (Cloud)

People intrusion detection

Detects people and vehicle intrusion in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

7

Vehicle intrusion detection

8

People intrusion in ROIs

Detects people intrusion in specific areas of interest (configurable) in camera footage. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

9

Lawn trampling detection

By analyzing video, the algorithm can help check whether there are people who are walking across grass in a specified area. The grass area needs to be configured. If the algorithm identifies lawn trampling behavior, it captures images, records the occurrence time, and produces captured images as evidence.

10

Factory intrusion detection

Detects intrusion events in key areas and dangerous areas in a factory. If such events are detected, the algorithm generates an alarm, captures onsite images, and records the occurrence time for case reporting, ensuring factory security.

11

Non-motorized Vehicle Detection (Cloud)

Detection of misparked non-motorized vehicle

Detects non-motorized vehicles parking outside the designated areas in camera footage. If such an event is detected, the algorithm further checks the length of time of the parking violation, records images and the occurrence time, generates an alarm, and sends the recorded information to the upper-level platform for further handling.

12

Detection of non-motorized vehicle

Detects non-motorized vehicles that appear in places where non-motorized vehicles are not allowed. The algorithm selects areas where non-motorized vehicles should not be parked. If non-motorized vehicles are detected in those areas, this algorithm further records images and the time, generates an alarm, and reports the recorded information to the upper-level platform so that the platform can generate case with received evidence.

13

E-bike-in-building detection

Detects e-bikes in staircases or elevators. The algorithm selects areas where e-bikes should not be parked. If e-bikes are detected in those areas, this algorithm further records images and the time, generates an alarm, and reports the recorded information to the upper-level platform so that the platform can generate case with received evidence.

14

Non-motorized vehicles lane violation

Detects non-motorized vehicles occupying motor vehicle lanes in a monitored area. If such events are detected, an alarm is generated and relevant departments can be notified in a timely manner to handle problems quickly.

15

Wrong-Way Driving Detection (Edge)

Wrong-way driving detection

Detects wrong-way driving events. If such events are detected, the algorithm captures onsite images and records the occurrence time for case reporting.

16

Safety Suite Detection (Cloud)

No-workwear detection

Detects workers who are not wearing safety helmets or reflective clothing when working in potentially hazardous positions and generates warnings when such events are detected. This helps to improve worker safety in factories, construction sites, and many other places. It also helps to improve workers' safety awareness.

17

No-safety helmet detection

18

Staff On-Duty Detection (Cloud)

Employee absence detection

The algorithm does this by counting the number of people in specific areas of interest captured in camera footage. If the number of people detected is less than the preset number (configurable), this algorithm determines that some employees working at key positions are off duty when they are not supposed to. When such an event is detected, this algorithm automatically records original images and the occurrence time, and generates an alarm.

19

Abnormal Parking Detection (Cloud)

Motor vehicle parking violation

Detects motor vehicles parking outside the designated areas for a time that exceeds the allowed period. If such an event is detected, the algorithm further checks the length of time of the parking violation, records images and the occurrence time, generates an alarm, and sends the recorded information to the upper-level platform for further handling.

20

Vehicle Detection (Cloud)

Vehicle detection on the cloud

Analyzes RTSP video streams from edge cameras, outputs information about vehicles in videos, and accurately provides their license plate information based on the specified ROI.

21

Mask Wearing Detection (Cloud)

Mask wearing detection

Detects people who do not wear masks in a monitored area and generates alarms. If such a person is detected, the algorithm generates an alarm and this person will be denied entry. Alternatively, related personnel will be sent to handle the problem.