Measuring exertion time,duty cycle and hand activity level for industrial tasks using computer vision

Abstract

Two computer vision algorithms were developed to automatically estimate exertion time, duty cycle (DC) and hand activity level (HAL) from videos of workers performing 50 industrial tasks. The average DC difference between manual frame-by-frame analysis and the computer vision DC was ?5.8% for the Decision Tree (DT) algorithm, and 1.4% for the Feature Vector Training (FVT) algorithm. The average HAL difference was 0.5 for the DT algorithm and 0.3 for the FVT algorithm. A sensitivity analysis, conducted to examine the influence that deviations in DC have on HAL, found it remained unaffected when DC error was less than 5%. Thus, a DC error less than 10% will impact HAL less than 0.5 HAL, which is negligible. Automatic computer vision HAL estimates were therefore comparable to manual frame-by-frame estimates.

Practitioner Summary: Computer vision was used to automatically estimate exertion time, duty cycle and hand activity level from videos of workers performing industrial tasks.     

The accuracy of conventional 2D video for quantifying upper limb kinematics in repetitive motion occupational tasks

Marker-less 2D video tracking was studied as a practical means to measure upper limb kinematics for ergonomics evaluations. Hand activity level (HAL) can be estimated from speed and duty cycle. Accuracy was measured using a cross-correlation template-matching algorithm for tracking a region of interest on the upper extremities. Ten participants performed a paced load transfer task while varying HAL (2, 4, and 5) and load (2.2 N, 8.9 N and 17.8 N). Speed and acceleration measured from 2D video were compared against ground truth measurements using 3D infrared motion capture. The median absolute difference between 2D video and 3D motion capture was 86.5 mm/s for speed, and 591 mm/s² for acceleration, and less than 93 mm/s for speed and 656 mm/s² for acceleration when camera pan and tilt were within ± 30 degrees. Single-camera 2D video had sufficient accuracy ( < 100 mm/s) for evaluating HAL.

Practitioner Summary: This study demonstrated that 2D video tracking had sufficient accuracy to measure HAL for ascertaining the American Conference of Government Industrial Hygienists Threshold Limit Value^® for repetitive motion when the camera is located within ± 30 degrees off the plane of motion when compared against 3D motion capture for a simulated repetitive motion task.     

The effects of cycle time on the physical demands of a repetitive pick-and-place task

This study seeks to elucidate the effects of the cycle time of a pick-and-place task on muscle activity, grip force, posture, and perception-based measures (discomfort and difficulty). Six healthy adults (3 males, 3 females) participated. A 4 x 2 repeated measures design was used with cycle time (1, 2, 5, and 10s) and grip (power and chuck) as independent variables. The task consists of repetitively picking a 0.7 kg part and placing it into a bin. A reduction in cycle time (CT) resulted in both a decrease of task time and physical rest time (p<0.001). The physiological muscle rest was much lower than the physical rest time (p<0.05). An increase in static muscle loading (p<0.01), grip force (p<0.001), and discomfort (p<0.001) were also observed. These results suggest that a pace threshold (between 2 and 5s for this task) is reached at a higher CT than that defined by the ability to perform the task.     

Effect of cycle time and duty cycle on psychophysically determined acceptable levels in a highly repetitive task

Psychophysical methodology has been used to develop guidelines for lifting and more recently similar methods have been applied to repetitive upper limb movements. While a range of cycle times are usually used, there is often no control for duty cycle. The purpose of this paper is to present psychophysically determined acceptable torques for a common upper limb task, with both cycle time and duty cycle conditions set by the researcher. Eight female participants, sitting at adjustable workstations, performed a simulated in-line screw running task. A computer-controlled torque motor applied a torque every 3, 6, 12 or 20?s with a duty cycle of 25, 50 or 83%. The participants worked with one set of conditions each day and self-selected the highest torque that they felt was acceptable without developing undue pain and discomfort. Duty cycle was found to significantly affect the amount of torque selected. With duty cycle controlled, cycle time was no longer found to have any significant effect on selected torque. Acceptable torques for 25, 50 and 83% duty cycles were 1.09, 0.9 and 0.73 Nm. Discomfort and stiffness were concentrated on the back of the hand and on the thumb web. These findings suggest that increased perception of discomfort with increased frequency (decreased cycle time) may be related to decreased rest/recovery time for muscles.     

Relationships between job organisational factors,biomechanical and psychosocial exposures

The relationships between work organisational, biomechanical and psychosocial factors were studied using cross-sectional data from a pooled dataset of 1834 participants. The work organisational factors included: job rotation, overtime work, having second jobs and work pace. Task and job level biomechanical variables were obtained through sub-task data collected in the field or analysed in the laboratory. Psychosocial variables were collected based on responses to 10 questions. The results showed that job rotations had significant effects on all biomechanical and most psychosocial measures. Those with job rotations generally had higher job biomechanical stressors, and lower job satisfaction. Overtime work was associated with higher job biomechanical stressors, and possibly self-reported physical exhaustion. Those having second jobs reported getting along with co-workers well. Work pace had significant influences on all biomechanical stressors, but its impact on job biomechanical stressors and psychosocial effects are complicated.

Practitioner Summary: The findings are based on a large number of subjects collected by three research teams in diverse US workplaces. Job rotation practices used in many workplaces may not be effective in reducing job biomechanical stressors for work-related musculoskeletal disorders. Overtime work is also associated with higher biomechanical stressors.     

Variance in direct exposure measures of typing force and wrist kinematics across hours and days among office computer workers

To determine the number of direct measurements needed to obtain a representative estimate of typing force and wrist kinematics, continuous measures of keyboard reaction force and wrist joint angle were collected at the workstation of 22 office workers while they completed their own work over three days, six hours per day. Typing force and wrist kinematics during keyboard, mouse and idle activities were calculated for each hour of measurement along with variance in measurements between subjects and between day and hour within subjects. Variance in measurements between subjects was significantly greater than variance in measurements between days and hours within subjects. Therefore, we concluded a single, one-hour period of continuous measures is sufficient to identify differences in typing force and wrist kinematics between subjects. Within subjects, day and hour of measurement had a significant effect on some measures and thus should be accounted for when comparing measures within a subject. PRACTITIONER SUMMARY: The dose response relationship between exposure to computer related biomechanical risk factors and musculoskeletal disorders is poorly understood due to the difficulty and cost of direct measures. This study demonstrates a single hour of direct continuous measures is sufficient to identify differences in wrist kinematics and typing force between individuals.     

Accuracy and feasibility of using an electrogoniometer for measuring simple thumb movements

The aim of this study was to determine the accuracy and feasibility of using an electrogoniometer (Model SG 110; Biometrics, Gwent, UK) for measuring simple thumb movements. Thumb disorders have been associated with the use of hand held devices such as mobile phones and these devices have become an integral part of modern life. In 15 young subjects, the measurements of eight flexion/extension (Flex/Ext) and adduction/abduction (Ad/Ab) thumb positions were compared between a thumb-mounted electrogoniometer and manual goniometer (which was taken as the benchmark). Group mean electrogoniometric measurement errors were below 4° and 5° for Ad/Ab and Flex/Ext measurements, respectively. During mobile phone use, the electrogoniometers measured differences in maximal joint angle postures, which appeared to be related to differences in mobile phone size. High movement velocities may increase the risk of musculoskeletal injury and the results indicated that Ad/Ab movements were twice the speed of Flex/Ext movements during mobile phone use. Electrogoniometers have utility for studying thumb movements during mobile phone use and may be used to evaluate other thumb-based input devices.     

A secure framework exploiting content guided and automated algorithms for real time video searching

This paper presents an architecture that allows End Users, via the services of Search Engines, to search, in a secure and efficient way, the video content belonging to Content Providers. The search can be accomplished with any searching scheme that the Search Engines wish to provide, as long as certain security constraints are met. However we propose specific algorithms that demonstrate an efficient way to search video data without sacrificing security effectiveness of the system. The search is completed without the End Users or Search Engines needing to purchase the premium content beforehand, and without the Content Providers needing to purchase the search technology. The business motivation of this technique is to assist End Users to purchase content best suiting their requirements—they are offered search results only, not actual content. The objective is to face the problem caused by the current segregation between content ownership and video processing technology ownership. To face this segregation, we present an architecture that guarantees security of Content Provider’s data and Search Engine’s technology and we also present two innovative algorithms that make real time video searching a feasible process. Particularly these algorithms (a) organize video content into a graph based hierarchical structure and (b) perform content guided, non interactive and real time search by exploiting the graph based video structures. The proposed algorithms are incorporated in the presented architecture under the given security constraints. Experimental results and comparisons with conventional techniques are presented to demonstrate the outperformance of the proposed algorithms.

Risk assessment of cheese processing tasks using the Strain Index and OCRA Checklist

The purpose of this study was to conduct and compare two ergonomic risk assessment methods often used in occupational health research and practice: the Strain Index (SI) and Occupational Repetitive Actions (OCRA) Checklist. Seven raters used the SI and OCRA Checklist to assess task-level physical exposures to the upper extremity of workers performing 21 cheese-manufacturing tasks. Of the total task exposures assessed with both methods, nearly half (49.1%) were classified as hazardous using the OCRA Checklist while 60.2% were classified as hazardous using the SI. Although the underlying injury risk characterization constructs of the SI and OCRA Checklist differ, the results indicated that the SI and OCRA Checklist often classified job tasks into similar risk categories. The differences in risk classifications determined by the SI and OCRA Checklist for job tasks were likely related to the definition of variables measured by these assessment methods as well as the complexity of tasks evaluated. By design, the SI is specific to the distal upper extremity while the OCRA Checklist accounts for the entire upper extremity including the shoulder. When conducting risk assessments of industrial work tasks, the choice of analysis tools should be based on the purpose of the assessment and the complexity of task functions. Both the SI and OCRA Checklist yield risk assessment ratings that are similar for cheese processing tasks.     

The predictive validity of the RSI QuickScan questionnaire with respect to arm,shoulder and neck symptoms in computer workers

The aim of this study was to determine whether results from the RSI QuickScan questionnaire on risk factors for arm, shoulder and neck symptoms can predict future arm, neck and shoulder symptoms in a population of computer workers. For this prospective cohort study, with a follow-up of 24 months, 3383 workers who regularly worked with a computer were approached. Generalised estimating equations (GEE) with 6, 12, 18 and 24 months time lags were used to determine whether high exposure was related to symptoms at follow-up. The results showed that high scores on 9 out of 13 scales, including previous symptoms, were significantly related to arm, shoulder and neck symptoms at follow-up. These results provide support for the predictive validity of the RSI QuickScan questionnaire.

Practitioner Summary: The results showed that high scores on 9 out of 13 scales, including previous symptoms, were significantly related to arm, shoulder and neck symptoms at follow-up. The RSI QuickScan questionnaire may be recommended as a tool in the identification of computer workers who should be targeted with interventions aimed at prevention of future symptoms.     

Psychophysical studies of repetitive wrist flexion and extension

Abstract

The purpose of this experiment was to investigate the feasibility of using psychophysical methods to determine maximum acceptable forces for various types and frequencies of repetitive wrist motion. Four adjustable work stations were built to simulate repetitive wrist flexion with a power grip, wrist flexion with a pinch grip, and wrist extension with a power grip. The study consisted of two separate experiments. Subjects worked for two days per week during the first experiment, and five days per week during the second experiment. Fifteen women completed the first experiment, working seven hours each day, two days per week, for 20 days. Repetition rates of 2, 5, 10, 15 and 20 motions per minute were used with each flexion and extension task. Maximum acceptable torques were determined for the various motions, grips, and repetition rates without dramatic changes in wrist strength, tactile sensitivity, or number of symptoms. Fourteen different women completed the second experiment, performing a wrist flexion motion (power grip) fifteen times per minute, seven hours per day, five days per week, for 23 days. There were no significant differences in maximum acceptable torque from day to day. However, the average maximum acceptable torque for a five days per week exposure was 36-3% lower than for the same task performed two days per week. Assuming that maximum acceptable torques decrease 36-3% for other repetition rates and motions, tables of maximum acceptable force were developed for female wrist flexion (power grip), female wrist flexion (pinch grip), and female wrist extension (power grip).     

利用OpenCV的运动物体智能视频图象处理监控系统设计

目前,无论是家庭还是社会安全性一直都备受关注。随着计算机视觉、视频智能识别等技术的高速发展,实现监控系统的实时分析报警已成为了可能,本文从运动物体的检测、运动跟踪及行为判断三大技术出发,利用OpenCV设计了一个运动物体智能视屏图象处理监控系统,该系统不但有效的实现对所监控区域的实时监控与报警,而且也对传统视频监控系统进行了革新,重要场所的安全运行提供了强有力的保障。     

Process and implementation of participatory ergonomic interventions: a systematic review

Participatory ergonomic (PE) interventions may vary in implementation. A systematic review was done to determine the evidence regarding context, barriers and facilitators to the implementation of participatory ergonomic interventions in workplaces. In total, 17 electronic databases were searched. Data on PE process and implementation were extracted from documents meeting content and quality criteria and synthesised. The search yielded 2151 references. Of these, 190 documents were relevant and 52 met content and quality criteria. Different ergonomic teams were described in the documents as were the type, duration and content of ergonomic training. PE interventions tended to focus on physical and work process changes and report positive impacts. Resources, programme support, ergonomic training, organisational training and communication were the most often noted facilitators or barriers. Successful PE interventions require the right people to be involved, appropriate ergonomic training and clear responsibilities. Addressing key facilitators and barriers such as programme support, resources, and communication is paramount.

Statement of Relevance: A recent systematic review has suggested that PE has some effect on reducing symptoms, lost days of work and claims. Systematic reviews of effectiveness provide practitioners with the desire to implement but do not provide clear information about how. This article reviews the literature on process and implementation of PE.     

Research on risk assessment model of subjective and objective integration in manual handling work

The main hazards in the process of manual handling work are triggered by human factors and ergonomics. This study is intended to implement a valid approach to quantitatively evaluate the risk level during manual handling work. A risk assessment model for manual handling workers was proposed based on subjective and objective correlation. A simulation experiment of manual handling process was developed and an ergonomics evaluation method was carried out. The 33-point human joint model of BlazePose neural network and the Rapid Upper Limb Analysis (RULA) method were utilized to determine the risk posture and risk index of the manual handling workers. This study brought together hemodynamic parameters and the score of Borg Rating of Perceived Exertion (RPE) scale and Physical Resources Scale (PRS) to obtain the final comprehensive index of work risk. The results showed that the risk indexes from the three stages of the experiment obtained by RULA method were 3, 4, and 7. And the scores of comprehensive indexes obtained by the risk assessment model were 1.841, 1.900, and 1.987, suggesting that the evaluation model based on subjective and objective correlation had the same ability to determine the risk levels of different handling tasks. Therefore, the risk assessment model proposed in this study verified the effectiveness of the comprehensive evaluation index integrating hemodynamic parameters and subjective evaluation scores.     

Design of assembly lines with the concurrent consideration of productivity and upper extremity musculoskeletal disorders using linear models

Work-related musculoskeletal disorders (WMSDs) are common occupational diseases among assembly workers due to repetitive motions or heavy workloads. The conventional approaches to decreasing WMSDs in assembly workers usually focus on individual assembly work at the station level. These approaches, however, do not pay enough attention to work allocation at the whole assembly line level such as balancing ergonomic burdens among workers by proper work assignment. This paper presents a methodology that can be used to integrate ergonomic measures of upper extremities into assembly line design problems. Linear models are developed to link work-worker assignment to the upper extremity ergonomic measures based on a guideline from American Conference of Governmental Industrial Hygienists. These linear models allow ergonomic and productivity measures to be integrated as a mixed-integer programming model. The case studies of this paper show the new model can effectively balance and control exposure levels in the upper extremity while not significantly decreasing line efficiency. This research shows the potential to reduce the need of numerous task adjustments for ergonomic improvement after initial assembly line design in conventional trial-and-error based assembly task adjustment. Furthermore, these linearization methods can be generalized in order to incorporate other ergonomic measures in tabulated forms into assembly line design problems.     

Ergonomic risk factors in glass artware industries and prevalence of musculoskeletal disorder

Glass artware manufacturing activities are highly repetitive, labor-intensive, and demand prolonged standing during the working shift. This leads to the development of musculoskeletal disorders (MSDs) symptoms among the workers. However, there is a scarce of study on the prevalence of MSDs symptoms and their associated risk factors among the glass artware workers. To address this research gap, this study aims to evaluate the prevalence of MSDs and working conditions among Indian glass artware workers. Discomfort data is collected using Cornell Musculoskeletal Discomfort Questionnaire and working posture data is collected through direct observations using Rapid Entire Body Assessment (REBA). Multiple logistic regression model is developed for each body region to study the relationship between the associated risk factors and MSDs symptoms with the help of collected data. The findings of this study are first, MSDs symptoms are highly prevalent and severe, particularly in the lower back, thigh, and foot. Second, the average REBA grand score is 7.22, which reveals the high-risk level of worker's posture while performing the task. Third, Job characteristics and individual factors are associated with MSD symptoms in multiple logistic regression models. The results highlight the prevalence and severity level of MSDs and illustrate the demand for ergonomic interventions to alleviate musculoskeletal symptoms among these working groups.Relevance to industryThis study draws attention to work-related and other associated risk factors, which would help the researchers and managers to understand the working environment of glass artware industry. These risk factors should be considered for designing and implementing ergonomic interventions to prevent and alleviate musculoskeletal discomfort among the glass artware workers.     

Minimizing total earliness and tardiness through unrelated parallel machine scheduling using distributed release time control

Using unrelated parallel machine scheduling to minimize the total earliness and tardiness of jobs with distinct due dates is a nondeterministic polynomial-hard problem. Delayed customer orders may result in penalties and reduce customer satisfaction. On the other hand, early completion creates inventory storage costs, which increase the total cost. Although parallel machines can increase productivity, machine assignments also increase the complexity of production. Therefore, the challenge in parallel machine scheduling is to dynamically adjust the machine assignment to complete the job within the shortest possible time. In this paper, we address an unrelated parallel machine scheduling problem for jobs with distinct due dates and dedicated machines. The objective is to dynamically allocate jobs to unrelated parallel machines in order to minimize the total earliness and tardiness time. We formulate the problem as a mixed integer linear programming (MILP) model and develop a modified genetic algorithm (GA) with a distributed release time control (GARTC) mechanism to obtain the near-optimal solution. A preliminary computational study indicates that the developed GARTC not only provides good quality solutions within a reasonable amount of time, but also outperforms the MILP model, a classic GA and heuristic approaches described in the literature.     

Simulating and modelling the DAX index and the USO Etf financial time series by using a simple agent-based learning architecture

This work presents an extensive case study on modelling the DAX (Deutscher Aktienindex) index and United States Oil Fund (USO) exchange-traded fund (Etf) time series with the financial agent-based system learning financial agent-based simulator (L-FABS) that exploits simulated annealing as a learning method. The USO Etf time series is highly correlated with oil price behaviour, and the DAX index is based on the weighted and accumulated behaviour of the share prices of some of the largest companies traded on the Frankfurt Stock Exchange. These two time series are driven by completely different economic factors and thus provide two diverse empirical settings to evaluate the effectiveness of our methodology. Our experimentation shows that a relatively simple computational representation of real financial markets is effective in capturing the overall behaviour of the time series with varying approximation levels while the prediction target is moved into the future. The reported experimental investigation of L-FABS shows that it is robust notwithstanding the learning method used and the data sets exploited. L-FABS indeed produced a relatively low approximation error in several settings even when evaluated with respect to other modelling approaches, for example, 0.88% and 1.61% errors on average for 1 day ahead experiments in, respectively, DAX index and USO Etf.