NTT COMWARE English Site


Realizing Initiatives to Improve Safety in Telecommunications Facility Construction Using Multi-AI
Confirmed 60% Reduction of Safety Checking Processes

NTT Communications Corporation

NIPPON TELEGRAPH AND TELEPHONE CORPORATION (President: Jun Sawada; hereinafter NTT), NTT Communications Corporation (President: Toru Maruoka; hereinafter NTT COM), and NTT COMWARE CORPORATION (President: Masato Kuroiwa; hereinafter NTT COMWARE) conducted a field trial relating to improving safety in telecommunications facility construction (hereinafter the field trial) between June 2021 and October 2021, using multi-AI*1 that assessed workers' level of danger by combining object recognition (of stepladders, etc.) with AI that estimates worker posture. As a result of this field trial, we confirmed a 60% reduction of safety checking processes for work on stepladders, which is indispensable in telecommunications facility construction, achieving a detection rate*2 of as much as 80%. The NTT Group will keep working to create safe and secure workplaces by introducing new technologies and initiatives.

1. Background
With the increasing importance of ESG management, one of companies' most pressing tasks is to create safe and secure workplaces. The NTT Group also specifies the promotion of new work styles and the creation of a workplace in our sustainability charter and is working to build an environment where our workers can operate with peace of mind, aiming for zero personal injury accidents together with all our business partners.
Meanwhile, safety measures often rely on experience and visual inspection. As a result, the sophistication and the efficiency of safety measures have become issues. NTT COM conducts about 60,000 telecommunications facility construction jobs annually. Especially for work on stepladders requiring safety measures, we do not simply rely on on-site supervisors to caution the construction workers, but ask safety checkers to watch videos recorded at construction sites and provide guidance afterward if they notice dangerous work.
This field trial aimed to considerably reduce work that require visual inspection through the development of multi-AI that can detect certain dangerous tasks and dangerous situations using the DX solution SmartMainTechⓇ*3 that NTT COMWARE is seeking to contribute to more sustainable social infrastructure. Another aim was to shift the work of safety checkers to more detailed task analysis and appropriate safety guidance, thereby achieving safety measures that are both more sophisticated and more efficient.

2. Overview and results of the field trial
(Figure 1) is an example of danger when doing stepladder work, so we filmed actual work situations and showed the footage to an AI analytical engine, verifying if it is able to detect the dangerous situations as well as developing and checking ways to display what is detected on a viewer.
As a result of this field trial, about 80% of dangerous situations in this specific task type were detected, which suggests the possibility of reducing the work of safety checkers by about 60% when comparing the AI's detection with the conventional cases where a safety checker goes through all the videos to visually inspect dangerous situations.

(Figure 1) Example of dangerous situation in stepladder work (filmed with attention to safety)
(Figure 1) Example of dangerous situation in stepladder work (filmed with attention to safety)

(Figure 2) Danger detection AI and timeline inspection screen (image)
(Figure 2) Danger detection AI and timeline inspection screen (image)

3. Features of the detection technology
The higher accuracy and efficiency of detecting dangerous situations is made possible by mainly three features.
(1) Detecting dangerous posture through independent judgement logic
Detects dangerous situations by obtaining physical information (posture in terms of body orientation, hand location, etc.) from footage and estimating poses that are important for ensuring work safety
(2) Multi-AI that judges objects and people in combination
Comprehensively judges if situations are dangerous based on mutual relations between detected pose and other detected objects such as stepladders
(3) Viewer display with timeline
Efficiently displays the timeframe when danger was detected by the AI and issues warnings
It can display dangerous situations detected by the AI with a color-coded timeline, making it possible to efficiently skip and play segments by type as well as insert caution comments and zoom in

(Figure 3) AI detecting human physique information and stepladder location
(Figure 3) AI detecting human physique information and stepladder location

4. Each company's role
(1) NTT
Advice on NTT Group AI technology and promoting safety improvement measures
Field trial of multi-AI for detecting dangerous work and safety guidance using viewers
Providing multi-AI for detecting dangerous work and viewers

5. What's next
Even for work where after-accident confirmation was difficult due to the significant number of incidents, we will significantly broaden the scope of confirmation by using multi-AI and other approaches, which should allow us to gain a broader understanding of the potential dangers. Moreover, identifying examples of dangerous work from past videos and create example footage will make it possible to offer workers more specific safety guidance and to increase the quality of guidance.
Moreover, we will continue field trials to measure the effects of introducing these tools, increase AI accuracy, start using them in some NTT COM processes, and contribute to workplaces in other industries that have stepladder work.

*1: Combining the results of multiple AIs to achieve more sophisticated judgments and improve accuracy.
In this field trial, judgments were made by combining AIs recognizing stepladders and other objects and estimating worker posture.
*2: Calculation with visually inspected dangerous situations as the denominator and AI-detected dangerous situation as the numerator.
*3: SmartMainTechⓇ is a registered trademark of NTT COMWARE CORPORATION.

Inquiries about this matter
Public Relations Office

NTT Communications Corporation
Platform Service Division

[Inquiries about SmartMainTechⓇ]
Ito and Yamada, Business Incubation Division
[Inquiries from reporters]
Maeda and Sato, Public Relations Office

The information shown in this page is the latest as of the day of the release. It is subject to change without further notice. Thank you for your kind understanding and consideration. PageTop