AI vision processing recognition for reduction of musculoskeletal injury risk in industry

工作场所的受伤和疾病是司空见惯的。41%的此类事故可归因于肌肉骨骼损伤(仅去年英国就有49.8万人)，导致英国物流、建筑、航空和农业等行业每年损失690万个工作日。这是一项必须由雇主承担的成本，仅在欧洲和美国，每年就超过100B GB。雇主可以实施的最有效的控制措施是消除或减少工人甚至需要做出使他们面临肌肉骨骼损伤风险的动作。然而，它们往往是最昂贵的，因此必须进行分析，以确保设计和实施正确的风险控制，为企业提供投资回报。目前，这种分析主要是通过安全专业人员进行的评估完成的。这些评估通常是通过“眼睛”完成的，专业人士观察一个人承担一项任务，衡量(或猜测)他所承担的角度和位置，然后创建一份报告，就如何改进这项特定任务提出建议。常见的解决方案包括引入新工具、自动执行风险较高的任务，或者重新设计任务流程以降低风险。然而，依靠人来进行这些测量和判断，往往会得出错误的结论，很难在大型工作场所进行衡量。作为之前InnovateUK可行性研究的一部分，Soter Analytics开发了一款视觉识别软件，可以识别一个人的动作并对每个动作的风险进行分类。虽然开发SOTER是为了解决不同的内部挑战，但SOTER的客户群要求使用该软件来识别任务风险。目前，还没有能够准确和可扩展地识别和衡量固有工作场所和任务风险的解决方案。风险评估是由安全专业人员“亲眼”完成的。这导致了主观评估，缺乏可扩展性导致进行了少量评估，降低了工作场所改进举措的可靠性和准确性。SoterTask将是在该项目期间开发的新产品，旨在创建基于软件的视觉处理人工智能解决方案来解决这一问题。Soter的管理团队和董事会已经批准了该项目，前提是可以获得资金，并估计在商业化的前5年内投资回报率为18倍。

Injuries and illness in the workplace are commonplace. 41% of such incidents can be attributed to musculoskeletal injury (498,000 in the UK alone last year), and results in 6.9m workdays being lost in the UK every year across sectors including logistics, construction, aviation and agriculture. This is a cost that has to be shouldered by the employer and, in across Europe and US alone exceeds £100B per annum.The most effective control that employers can implement are eliminating or reducing the need for workers to even have to make movements that put them at risk of a musculoskeletal injury. However, they often can be the most expensive and thus analysis must be done to ensure that the correct risk control is designed and implemented that will provide the business with an ROI.Currently this analysis is primarily done through assessments undertaken by safety professionals. These assessments are usually done 'by eye', with the professional watching a person undertake a task, measure (or guess) the angles and positions the person undertakes, and then creates a report with recommendations on how that particular task could be improved. Common solutions include introducing new tooling, automating higher-risk tasks, or redesigning how the task flows to reduce risk. However, by relying on the person to make these measurements and judgements, the wrong conclusions are often made and it's very difficult to scale this across large workplaces.As part of a previous InnovateUK feasibility study, Soter Analytics developed a vision recognition software that identifies the movements a person makes and categorises the risk of each movement. While developed to solve a different, internal challenge, Soter's customer base have requested the software to be used to identify task risk.Currently, there are no solutions able to accurately and scalable identify and measure the inherent workplace and task risk. Risk assessments are done 'by eye' by safety professionals. This leads to subjective assessments and the lack of scalability leads to a small amount of assessments being done, reducing the reliability and accuracy of workplace improvement initiatives. SoterTask will be the new product developed during this project, to create a software based, vision processing AI solution to solve this problem.Soter's management team and board have approved the project, providing funding can be secured, and estimate an 18X ROI over the first 5-years of commercialisation.