Headnote
ABSTRACT
Objective: With the aim of developing and validating Smartergo as a technological tool capable of automating the application of the RULA methodology, this study also seeks to implement computer vision algorithms for postural detection, incorporate complementary evaluation parameters such as muscle strength and leg stability, compare the tool's performance with established systems such as SAPO and Kinovea, and verify whether the solution complies with the guidelines of NR-17 and ISO 11228-3.
Theoretical Framework: WRMSDs (Work-Related Musculoskeletal Disorders) are the most frequently reported occupational diseases in Brazil, especially among women aged 40 to 49. Biomechanical analysis and software tools such as SAPO and Kinovea help identify poor posture and joint overloads. These tools provide accurate and accessible evaluations, aiding in the prevention and treatment of postural disorders. The Ministry of Health recommends NR-17-based measures such as regular breaks, ergonomic furniture, and workplace stretching exercises.
Method: This applied technological research focused on developing and validating Smartergo, a web-based tool that automates ergonomic posture assessments using the RULA method. The system uses MediaPipe to detect body landmarks in real-time and calculates joint angles to generate RULA scores. Validation involved comparing results from Smartergo with those of established tools like SAPO and Kinovea. While Smartergo showed promising performance, limitations related to image quality and camera positioning were noted. The tool's webbased design ensures accessibility without the need for complex installations.
Results and Discussion: The Smartergo software was evaluated using a dynamic image with automated detection via MediaPipe, producing RULA scores indicating high postural risk. Compared to semi-automated methods SAPO + Spreadsheet and Kinovea + Spreadsheet, Smartergo showed similar results on the right side but underestimated the left side risk. Multi-view analysis from manual methods demonstrated higher accuracy, highlighting Smartergo's monocular vision limitations. Nonetheless, Smartergo offers greater efficiency and usability, making it promising for quick screening, though complementary validation is needed for complex postures.
Research Implications: The practical implications of this research include efficient automation of postural ergonomic assessment, enabling quick risk identification in work environments, especially for small and micro enterprises. Theoretically, it advances the application of computer vision in ergonomics, paving the way for more accurate and accessible analyses. The study impacts sectors such as occupational health, workplace safety, and ergonomic software development.
Originality/Value: This study contributes to the literature by developing an innovative software that automates postural ergonomic assessment using computer vision integrated with the RULA method, reducing manual work and increasing efficiency. Its relevance lies in the practical application potential in workplaces, especially small businesses, and in standardizing assessments, which can improve occupational injury prevention.
Keywords: Ergonomics, RULA, Postural Assessment, Computer Vision.
RESUMO
Objetivo: Com o intuito de desenvolver e validar o Smartergo como uma ferramenta tecnológica, capaz de automatizar a aplicação da metodologia RULA, este estudo também busca implementar algoritmos de visão computacional para detecção postural, incorporar parâmetros complementares de avaliação, como força e estabilidade das pernas, comparar o desempenho da ferramenta com sistemas já consolidados, como SAPO e Kinovea e, verificar se a solução está em conformidade com as diretrizes da NR-17 e da norma ISO 11.228-3.
Referencial Teórico: As LER/DORT são as doenças ocupacionais mais notificadas no Brasil, afetando principalmente mulheres de 40 a 49 anos. A análise biomecânica e o uso de softwares como SAPO e Kinovea auxiliam na identificação de posturas inadequadas e sobrecargas articulares. Essas ferramentas oferecem avaliações precisas e acessíveis, contribuindo para a prevenção e o tratamento de distúrbios posturais. O Ministério da Saúde recomenda medidas baseadas na NR-17, como: pausas, mobiliário ergonômico e ginástica laboral.
Método: Este estudo caracteriza-se como uma pesquisa tecnológica aplicada, voltada para o desenvolvimento e a validação do Smartergo, uma ferramenta web que automatiza a avaliação ergonômica postural, com base no método RULA. O sistema utiliza a biblioteca MediaPipe, para detectar pontos-chave do corpo em tempo real e calcular ângulos articulares, gerando os escores da avaliação. A validação foi feita por meio de comparação com ferramentas consolidadas, como o SAPO e o Kinovea. Apesar do bom desempenho, foram observadas limitações relacionadas à qualidade das imagens e ao posicionamento da câmera. O design web da ferramenta permite fácil acesso, sem necessidade de instalação complexa.
Resultados e Discussão: O software Smartergo foi avaliado com uma imagem dinâmica, usando detecção automática via MediaPipe, gerando escores RULA, que indicam alto risco postural. Comparado aos métodos semiautomatizados SAPO + Planilha e Kinovea + Planilha, o Smartergo apresentou resultados semelhantes no lado direito, mas subestimou o risco do lado esquerdo. A análise multi-vista dos métodos manuais mostrou maior precisão, destacando limitações da visão monocular do Smartergo. Apesar disso, o Smartergo oferece maior eficiência e facilidade de uso, sendo promissor para triagem rápida, embora precise de validação complementar em posturas complexas.
Implicações da Pesquisa: As implicações práticas desta pesquisa incluem a automatização eficiente da avaliação ergonômica postural, facilitando a identificação rápida de riscos em ambientes laborais, especialmente para pequenas e microempresas. Teoricamente, contribui para avanços no uso de visão computacional, aplicada à ergonomia, abrindo caminho para análises mais precisas e acessíveis. O estudo impacta setores como: saúde ocupacional, segurança do trabalho e desenvolvimento de software ergonômico.
Originalidade/Valor: Este estudo contribui para a literatura, ao desenvolver um software inovador, que automatiza a avaliação ergonômica postural, usando visão computacional integrada ao método RULA, reduzindo o trabalho manual e aumentando a eficiência. A relevância está no potencial de aplicação prática em ambientes laborais, especialmente para pequenas empresas, e na padronização das análises, o que pode melhorar a prevenção de lesões ocupacionais.
Palavras-chave: Ergonomia, RULA, Avaliação Postural, Visão Computacional.
RESUMEN
Objetivo: Con el objetivo de desarrollar y validar Smartergo como una herramienta tecnológica capaz de automatizar la aplicación de la metodología RULA, este estudio también busca implementar algoritmos de visión artificial para la detección postural, incorporar parámetros de evaluación complementarios como la fuerza muscular y la estabilidad de las piernas, comparar el rendimiento de la herramienta con sistemas consolidados como SAPO y Kinovea, y verificar si la solución cumple con las directrices de la NR-17 y la norma ISO 11228-3.
Marco teórico: Los trastornos musculoesqueléticos relacionados con el trabajo (TMRE) son las enfermedades profesionales más frecuentemente reportadas en Brasil, especialmente entre mujeres de 40 a 49 años. El análisis biomecánico y herramientas de software como SAPO y Kinovea ayudan a identificar malas posturas y sobrecargas articulares. Estas herramientas proporcionan evaluaciones precisas y accesibles, lo que facilita la prevención y el tratamiento de los trastornos posturales. El Ministerio de Salud recomienda medidas basadas en la norma NR-17, como descansos regulares, mobiliario ergonómico y ejercicios de estiramiento en el lugar de trabajo.
Método: Esta investigación tecnológica aplicada se centró en el desarrollo y la validación de Smartergo, una herramienta web que automatiza las evaluaciones posturales ergonómicas mediante el método RULA. El sistema utiliza MediaPipe para detectar puntos de referencia corporales en tiempo real y calcula los ángulos articulares para generar puntuaciones RULA. La validación consistió en comparar los resultados de Smartergo con los de herramientas consolidadas como SAPO y Kinovea. Si bien Smartergo mostró un rendimiento prometedor, se observaron limitaciones relacionadas con la calidad de la imagen y el posicionamiento de la cámara. El diseño web de la herramienta garantiza la accesibilidad sin necesidad de instalaciones complejas.
Resultados y discusión: El software Smartergo se evaluó utilizando una imagen dinámica con detección automatizada mediante MediaPipe, generando puntuaciones RULA que indican un alto riesgo postural. En comparación con los métodos semiautomatizados SAPO + Spreadsheet y Kinovea + Spreadsheet, Smartergo mostró resultados similares en el lado derecho, pero subestimó el riesgo en el lado izquierdo. El análisis multivista con métodos manuales demostró una mayor precisión, lo que pone de relieve las limitaciones de la visión monocular de Smartergo. No obstante, Smartergo ofrece mayor eficiencia y usabilidad, lo que lo hace prometedor para la detección rápida, aunque se requiere una validación complementaria para posturas complejas.
Implicaciones de la investigación: Las implicaciones prácticas de esta investigación incluyen la automatización eficiente de la evaluación ergonómica postural, lo que permite la rápida identificación de riesgos en entornos laborales, especialmente en pequeñas y microempresas. En teoría, impulsa la aplicación de la visión artificial en ergonomía, allanando el camino para análisis más precisos y accesibles. El estudio impacta en sectores como la salud ocupacional, la seguridad en el trabajo y el desarrollo de software ergonómico.
Originalidad/Valor: Este estudio contribuye a la literatura mediante el desarrollo de un software innovador que automatiza la evaluación ergonómica postural mediante visión artificial integrada con el método RULA, lo que reduce el trabajo manual y aumenta la eficiencia. Su relevancia radica en su potencial de aplicación práctica en lugares de trabajo, especialmente en pequeñas empresas, y en la estandarización de las evaluaciones, lo que puede mejorar la prevención de lesiones laborales.
Palabras clave: Ergonomía, RULA, Evaluación Postural, Visión Artificial.
1 INTRODUCTION
Repetitive Strain Injuries and Work-Related Musculoskeletal Disorders, known as RSI/WMSD, are the main causes of absences and loss of productivity in the work environment in Brazil. According to data from the Ministry of Health, between 2007 and 2016, about 67,599 cases of RSI / WMSD were registered, which represents an increase of 184% in this period. In Ceará, specifically, the Observatory of Safety and Health at Work (2023) identified 4,852 cases between 2017 and 2021.
According to Silva et al. (2024), bibliometric studies indicate that the scientific production on applied ergonomics is incipient, where the most affected sectors are industry, commerce and services in general, including professions such as machine operators, cooks and cleaners, works that involve repetitive physical effort and inadequate postures.
In this context, it is essential to use efficient ergonomic methodologies to help prevent posture-related problems. A well-known and used tool is the RULA (Rapid Upper Limb Assessment), created by McAtamney and Corlett (1993). It serves to assess the posture of the arms, shoulders and neck in the work environment, helping to identify ergonomic risks linked to posture, muscle strength and applied load. With this, it is possible to make more informed decisions, about the need to make adjustments or interventions in the workplace.
However, with the limitations of these traditional evaluations, there has been an advance in the use of digital technologies in the area of ergonomics, such as computer vision and artificial intelligence. Programmes such as SAPO and Kinovea, already show the potential to make biomechanical analysis faster and more efficient. Still, these softwares are not fully integrated with normative protocols such as RULA and do not take into account some important variables, such as muscle activity or the use of force, during work (Dianat et al., 2018). This highlights an important technological gap: the absence of automated tools that, in addition to accurately estimating body postures, deliver results compatible with the logic and parameters of the RULA method.
This work presents Smartergo, an online software created to fully automate the ergonomic postural evaluation, using the RULA method. The tool allows you to analyse images or videos, made by webcam or mobile devices, automatically identifying the worker's posture and calculating the angles of important joints. In addition, the Smartergo allows the inclusion of extra data, such as: additional score for muscle activity, use of load on the arms and legs, in addition to evaluating the positioning of the legs. These resources help make the assessment more accurate (Kozak et al., 2021).
Thus, the general objective of this study is to develop and validate the Smartergo as a technological tool, focused on the automation of the RULA methodology. The specific objectives include: (i) implementing computer vision algorithms for automated postural detection; (ii) integrating complementary evaluation parameters (strength, muscle activity, leg stability); (iii) comparing the effectiveness of Smartergo against consolidated tools such as SAPO and Kinovea; and (iv) evaluating the adherence of the solution to the guidelines of Regulatory Standard NR-17 and ISO 11.228-3.
In practice, the proposal aims to offer an affordable and reliable solution, especially for micro and small companies in the industrial pole of Ceará, where, according to Sebrae (2023), 89% of businesses are small. Thus, the software seeks to facilitate the realisation of ergonomic evaluations in these environments. In the academic field, Smartergo contributes to the advancement of integration between occupational ergonomics and emerging technologies of Industry 4.0.
2 THEORETICAL FRAMEWORK
According to the Ministry of Health (2022), based on studies of Health Brazil and the Information System for Notifiable Diseases (SINAN), diseases related to RSI / WMSDs are the ones that most affect Brazilian workers. Between 2007 and 2016, 67,599 cases were registered, which represents an increase of 184% in these records. These conditions were more common among people with complete high school education, representing 32.7% of cases and among women aged 40 to 49 years, corresponding to 51% of cases, and this age group had an index of 33.6%. In the different work sectors, the highest incidence occurred among professionals in commerce, industry, transportation, domestic services and cleaning, food, as well as fixed machine operators, cooks, cleaners and workers on production lines.
RSI/WMSDs manifest themselves through various symptoms that affect the upper limbs, usually in advanced stages of the disease, including: pain, fatigue and feeling of heaviness, inflammation in the joints and tissues that cover the tendons, shoulder injuries, being responsible for most of the absences from work, representing expenses with treatments, reintegration processes to occupation and compensation payments (Lelis et al., 2012).
For this, as actions for the prevention of RSI / WMSDs and other pathologies, the Ministry of Health recommends employers: attention to NR-17 (Ergonomics), health education measures for workers, together with the Occupational Health Reference Centres (CEREST) of each region, and employees, stimulates effective participation in labour gymnastics in the workplace, regular performance of body movements, avoid mental overload and overtime, use ergonomic furniture, create regular break habits during working hours and, for any sign of pain you may feel, seek a specialist doctor (Maciel, 2022).
The biomechanical analysis aims to identify the changes and risks, observe the overload imposed on the body, as well as the postures that can lead to injury. For this, it identifies the joint angles and measurements present in a given activity, serving as a parameter to identify joint/muscle overloads in the occupation. The use of reliable software contributes to an effective evaluation, presenting quantitative results of postural asymmetries, if executed correctly (Araújo et al., 2021).
SAPO (Software for Postural Evaluation) is a free tool, developed to assist in the analysis of human posture, through digitised photographs. It allows the identification and measurement of postural deviations and joint angles, being widely used by health professionals, such as physiotherapists and physical educators, for diagnosis, treatment planning and monitoring of the evolution of patients (Ferreira et al., 2010). The computerised biophotogrammetry method, used by Sapo, offers more reliable data than subjective evaluations, because in addition to facilitating access to information, it has a low cost. It is especially useful for detecting asymmetries in the body, especially at angles that are difficult to estimate manually. When the diagnosis is made early, the professional can choose the best treatment or other protocols for the lesion, helping to prevent or reduce postural problems (Carletto et al., 2021).
Sapo is a versatile programme that works using the Java programming language. To begin with, the user fills in a form with personal information and clinical data, including where the patient has been feeling pain, how long it has been, as well as other notes and observations. Then the practitioner marks specific points on the patient's skin in different views: frontal, right lateral, left lateral, and posterior. Then photos are taken, which are stored in the system for analysis. The programme allows you to score free points in anatomy, measure body angles and calculate distances. After the evaluation, it is possible to generate a report or print the data obtained, helping the professional to better monitor the patient's case (Gabler et al., 2024).
The Kinovea software is a free video player for biomechanical analysis, created with a focus on professionals in the sports field, where it allows the analysis of movements, through the comparison between two videos simultaneously and the tracking to show the angles and the exact measurements of various types of movements (Softonic, 2025).
Vieira et al. (2022) describe that manually monitoring the tracking of articular points for the calculation of position and other variables such as displacement, speed and acceleration, would be very difficult and for this a system was created capable of transferring these data to the digital medium, giving more precision and reliability to the end user.
3 METHODOLOGY
This study is characterised as an applied technological research, focussing on the development of the Smartergo software, a web tool for the automation of postural ergonomic evaluation, based on the RULA method. In addition to the creation of the system, the research incorporated a comparative experimental approach, in order to evaluate the effectiveness of Smartergo in relation to consolidated tools, such as SAPO and Kinovea, used in biomechanical postural analysis (Kozak et al., 2021).
Smartergo was made as an interactive web application, using the HTML, CSS and JavaScript languages to structure, style and make the tool work. The automation of postural analysis was made possible thanks to the integration of the MediaPipe library, a technology developed by Google, which processes visual data in real time, such as videos or images. It allows you to detect key points of the body, such as: shoulders, elbows, wrists, trunk, neck and hips, in real time during the analysis. The logic of the RULA analysis was implemented in the scriptrulaFunctions.js, which maps the detected angles to the evaluation scores according to the methodology proposed by McAtamney and Corlett (1993).
To validate the software, images were obtained from an anatomical reference database, representing different human postures in work activity. The main image used contained body representation in a dynamic position, with arms and legs at different angles, allowing the analysis of multiple segments. The analyses were conducted through the developed tool and the comparative tools, with subsequent crossing of the results obtained.
The effectiveness of Smartergo was evaluated by comparing it with SAPO and Kinovea software. The SAPO was used for extraction of joint angles, based on manual markings in static image, followed by application of RULA logic, via spreadsheet. On the other hand, Kinovea was used to measure angles in video, with data insertion in RULA spreadsheet, to obtain the final scores. In both approaches, the same data from the reference image analysed by Smartergo were used.
The final RULA scores, obtained by the three methods, were compared with the analysis of the recommended level of action for each side of the body (right and left), according to the original RULA methodology.
Although the development of Smartergo is an important advance in the automation of postural assessment, using the RULA tool, it is important to remember that there are some methodological limitations. In particular, the accuracy of the analysis done automatically and the strength of the system, under different conditions of use, still need to be considered.
One of the main restrictions is the quality of the images or videos, which are used as input. The ability of the MediaPipe library algorithm to correctly detect the body points depends greatly on the sharpness, resolution and illumination of the material. If the environment is dim in light, has excessive shadows, or has little contrast between the body and the background, this can make it difficult to correctly identify landmarks, which can affect calculations of angles and thus the results.
Another limitation is related to the position of the camera when taking the pictures. It is important that the viewing angle is well aligned with the plane of the body we want to analyse. If the camera is tilted or with a different perspective, this can cause distortions in the image, which can affect the accuracy in measuring the angles of the joints and thus compromise the evaluation results.
In addition, more complex body positions, such as those in which the limbs are overlapping, the trunk is rotated, or when equipment that hides body parts is used, may make it difficult to identify landmarks. In such cases, the computer vision system may have difficulties or make errors when locating these anatomical points, which impairs the correct application of the RULA method rules.
According to Laperuta et al. (2017), there are several ergonomic methods used for the evaluation of postural risks and workstation, such as: REBA (Rapid Entire Body Assessment), ROSA (Rapid Office Strains Assessment), RULA (Rapid Upper Limb Assessment), Suzzane Rodgers and ERGO / IBV. But in this work, we will focus on the RULA tool.
For the application of the RULA tool, the position of the collaborator regarding the shoulder, elbow, wrist, neck, trunk and legs should be observed, in addition to the load exerted and the muscular effort. In Figure 1, the data are filled in after each observation. In the spaces of the scores, there are formulated tables, to be replaced when the data are collected. The final score is obtained after finalising the sums. The conclusion of this score shows whether it is necessary or not, the change in the workplace. Level one demonstrates that the posture is acceptable and does not demand change. While level four needs immediate changes in the workplace.
The software developed by the team for the automation of RULA analysis was implemented as an interactive web application, using standard front-end technologies: HTML for structure (index.html), CSS for visual styling (style.css) and JavaScript for functional logic and interactivity (script.js, config.js). This approach allows the tool to be accessed and executed directly in a web browser, without the need for complex installation, requiring only access to a camera (webcam).
The processing flow of the application starts with the capture of video, in real time, from the user's webcam. Each frame of the video is then processed by a computer vision module, integrated into the application. This module specifically uses the MediaPipe library, developed by Google, to perform pose estimation in real time. The MediaPipe detects and locates multiple key points (landmarks) in the body of the individual visible in the video (whose configuration and integration would be detailed in config.js or script.js). The management and manipulation of the coordinates of these key points seem to be treated in the landmarks.js file.
Once the 2D or 3D coordinates of these key points (such as shoulders, elbows, wrists, hips, neck) are obtained, the software proceeds to the geometric calculation of the angles between the body segments relevant to the RULA analysis. These angular calculations are the basis for automated postural assessment.
The specific logic of the RULA assessment is primarily encapsulated in the rulaFunctions.js file. This script receives the calculated angles and maps them to the RULA scores corresponding to each body part (Group A: arm, forearm, wrist; Group B: neck, trunk, legs). To perform this mapping, the system queries data structures representing the RULA scoring tables, as suggested by the trunkPostureScoreDic.js (trunk scoring dictionary) and wristPostureScoreDic.js (wrist scoring dictionary) files. Additional factors, such as muscle use and force/load scores (RULA standard elements), can be manually entered by the user or estimated/configured within the rulaFunctions.js logic.
Finally, the partial scores of Groups A and B are combined using tables C of the RULA method, whose combination logic and mapping to the final score probably resides in rulaFunctions.js with support of the data structure in finalScoreDic.js. The result of the analysis is the final RULA score and the corresponding action level, which is dynamically displayed in the application's web interface (index.html manipulated by script.js), often accompanied by a visual overlay of the detected skeleton on the user's image, providing immediate feedback on the evaluated posture.
4 RESULTS AND DISCUSSIONS
For the evaluation, three views of an image were chosen, which shows a representation of the human body without skin, exposing the muscles. Figure 2 is in a dynamic pose, with one arm raised and flexed, and the other extended to the side. The legs are in different positions, suggesting movement or balance.
The application of the team's software (Smartergo), which uses automated pose detection via MediaPipe on the reference image, generated the following final RULA scores, according to the report extracted from the application (Figure 3-A, 3-B and 3-C):
* Right side: Final Score RULA = 7 (Action Level 4: Investigate and change immediately).
* Left Side: Final Score RULA = 6 (Action Level 3: Investigate and Change Logo).
These scores were calculated automatically by the software, after the detection of the landmarks and the insertion of additional data by the user via the form (in this case: Static posture > 1 min for arms and legs; Burden < 2kg intermittent for arms and legs; Legs and feet supported and balanced).
In parallel, the same image was analysed using the combined method "SAPO + RULA Spreadsheet". This flow involved the analysis of the image in SAPO to obtain postural data (definition of points of interest, location of angles, as defined in the SAPO application methodology in this study) and the subsequent manual insertion of these data into a spreadsheet, which applies the RULA scoring logic. The results obtained by this method, visualised in the sample worksheet (Figure 4), were:
Right side: Final Score RULA = 7 (Action Level 4: Investigate and change immediately).
Left Side: Final Score RULA = 7 (Action Level 4: Investigate and change immediately).
Similarly, Figure 2 was analysed using the combined method "Kinovea + RULA Spreadsheet". As in the previous analysis, the Kinovea software was used to obtain postural data (definition of points of interest, location of angles, as defined in the Kinovea application methodology in this study) and the subsequent manual insertion of these data into a spreadsheet, which applies the RULA scoring logic. The results obtained by this method, visualised in the sample worksheet (Figure 5), were:
Right side: Final Score RULA = 7 (Action Level 4: Investigate and change immediately).
Left Side: Final Score RULA = 7 (Action Level 4: Investigate and change immediately).
Table 1 summarises the final RULA scores and corresponding action levels obtained for the reference image by the two methods effectively applied to it.
The divergent result of this point comparative analysis is in the final RULA score for the left side of the body and, consequently, in the recommended action levels, when comparing the fully automated evaluation (Team Software) with the semi-automated approach, combined with spreadsheet (SAPO + RULA Spreadsheet), for the same input image.
The introduction of multi-view analysis (front, side and top) for the SAPO and Kinovea based methods, brought new elements for comparison with the Smartergo software. A first relevant point is that the multi-view analysis raised the RULA score on the right side from 6 to 7, in the SAPO + Spreadsheet and Kinovea + Spreadsheet methods, matching it to the score on the left side and reinforcing the maximum risk classification (Level 4) for both sides, in this more complete approach. This highlights the importance of multiple perspectives to adequately capture the three-dimensionality of posture, in manual or semi-automated analyses, as recommended in the ergonomic literature and in the protocols of tools such as SAPO.
With the new data, the direct comparison with the Smartergo (which operated with a single front-view), reveals an interesting scenario: there was agreement in the final score for the right side (Score 7), but the divergence persists to the left side (Smartergo = 6 vs. SAPO / Kinovea Multi-Vista = 7). The agreement on the right side suggests that, for certain aspects of posture, captured in that specific view, the automated monocular Smartergo analysis can achieve a result similar to that of a more detailed multi-view analysis.
However, the discrepancy on the left side (Smartergo, underestimating the risk at one point, compared to the other two methods), reinforces the limitations inherent to the analysis, based on a single camera, as discussed previously. It is likely that the left-sided pose in the reference image contained elements (possibly specific wrist, forearm or neck rotations or lateral deviations) that could only be adequately quantified or interpreted, with the combined information of the frontal and lateral views used in the manual/semi-manual analysis. The Smartergo, limited to 2D (or estimated 3D) information from a single perspective, may not have captured or may have misinterpreted these critical components for the left-hand RULA score, resulting in a slightly lower final score. The accuracy of landmark detection by MediaPipe, in that specific region and in that view, may also have influenced the result.
Despite this point difference on the left side, it is important to note that all methods, in their final versions (Smartergo 1-vista and SAPO/Kinovea multi-vista), classified the posture as high risk (Level 3 or 4), indicating the need for intervention. This suggests that the Smartergo, even with the limitations of monocular vision, was able to identify the general problematic nature of posture. Its main advantage lies in the complete automation of the process, eliminating the laborious steps of manual marking (SAPO) or angular measurement (Kinovea) and external calculation in spreadsheets, offering a substantial gain in efficiency and standardisation, as well as reducing inter-evaluator variability.
The incorporation of additional variables (muscle activity, strength/load) directly into the Smartergo interface also makes it a more complete tool for the application of RULA, than a simple video analysis, focused only on angles. Its web nature facilitates access and dissemination, being particularly promising for micro and small companies that need practical and low-cost ergonomic solutions to comply with regulations such as NR-17 and prevent RSI / WMSD.
However, the results reinforce that Smartergo, in its current implementation, based on monocular vision, should be used with awareness of its limitations. It proves to be an excellent tool for rapid screening and identification of obvious risks, but in complex, ambiguous postures or with scores close to the limits of action levels, validation by a professional or complementary analysis with multiple views (either manually or by future versions of the software) may be necessary to ensure maximum accuracy, especially in distinguishing between action levels 3 and 4.
Future directions should therefore include extensive validation of Smartergo in real scenarios and with a broader dataset, but also explore the implementation of techniques that use multiple views (combining different front and side view angles) or more sophisticated 3D pose estimation algorithms, to overcome the limitations of monocular vision and improve accuracy, in the evaluation of complex rotations and postures.
5 CONCLUSION
This study developed and validated Smartergo, a web software that automates the ergonomic postural assessment by the RULA method, integrating variables such as muscle activity and load. Smartergo has demonstrated practical feasibility, increased automation and accuracy compared to hand tools, aligning with NR-17 and ISO 11228-3 standards. It contributes academically by combining computer vision and normative protocols, and socially by enhancing the prevention of RSI / WMSDs and promoting healthier work environments. Limitations include dependence on image quality and restricted basis for validation. It is recommended to expand data, integrate IoT sensors and test in various sectors to improve the tool. Ergonomic automation by Smartergo can reduce absences and improve the quality of working life, especially in emerging industrial regions.
References
REFERENCES
Araújo, E. O., & Duarte, A. D. (2021). Análise de confiabilidade de software na análise biomecânica: revisão de literatura. Revista Ação Ergonômica, 15 (2), 1-11. http://www.abergo.periodikos.com.br/article/6276c8eca9539525df01aa53/pdf/abergo-16- 1-e202202.pdf
Brasil, Ministério da Saúde. (2022). LER e DORT são as doenças que mais acometem os trabalhadores, aponta estudo.<https://www.gov.br/saude/ptbr/assuntos/noticias/2019/abril/ler-e-dort-sao-as-doencas-que-mais-acometem-os>- trabalhadores-aponta-estudo
Campcursos. (2023). Avaliação rápida de membro superior (RULA). https://campcursos.com.br/avaliacao-rapida-de-membro-superior-rulair-para-navegacaopesquisa/
Carbonera & Tomazini Advogados. (2022). Guia completo do auxílio-doença do bancário.<https://www.carboneraetomazini.com.br/blog/guia-completo-do-auxilio-doenca-dobancario/
Carletto, A., & Matta, M. (2021). Utilização do protocolo SAPO como recurso para avaliação postural. http://www.repositorioguairaca.com.br/jspui/bitstream/23102004/337/1/Adriane%20Carl etto.pdf
Charmant, J. (2004). Kinovea: A microscope for your videos. https://www.kinovea.org/
Dianat, I., Karimi, M. A., & Asgharimoghaddam, H. (2018). Ergonomic evaluation of working postures using RULA and observational methods: A review and comparison. International Journal of Industrial Ergonomics, 66, 69-79.
Dias, C. P. (2017). Análise ergonômica do trabalho: aplicação da ferramenta RULA em atividade industrial (Dissertação de mestrado, Universidade Tecnológica Federal do Paraná).
Ferreira, E. A. G., et al. (2010). Postural assessment software (PAS/SAPO): validation and reliability. Clinics, 65, 675-681. https://bmclab.pesquisa.ufabc.edu.br/sapo/
Franceschi, A. de. (2013). Ergonomia. Secretaria de Educação Profissional e Tecnológica. https://www.ufsm.br/app/uploads/sites/413/2018/11/12_ergonomia.pdf
Gabler, N. C., Menezes, M. F., & Favero, M. T. (2024). Avaliação postural com o software Sapo. Revista Saberes da Unijipa, 21 (6), 99-109.<https://estacio.periodicoscientificos.com.br/index.php/revista-saberesunijipa/article/view/2928
Grilo, A. (2017). Precisamos projetar uma ergonomia de concepção, e não de correção: Relembrando aprendizados sobre ergonomia durante o 16 Ergodesign/USIHC & CINAHPA. https://andregrilo.medium.com/precisamos-projetar-uma-ergonomia-deconcepção-e-não-de-correção-78bad0d5896f
International Organization for Standardization. (2007). ISO 11228-3: Ergonomics - Manual handling - Part 3: Handling of low loads at high frequency. https://www.iso.org/standard/26522.html
Kozak, J., Haghighi, M., & Kuo, P.-C. (2021). Deep learning-enabled RULA ergonomic risk assessment using monocular images. Computers in Industry, 131, 103500. https://doi.org/10.1016/j.compind.2021.103500
Laperuta, D. G. P., et al. (2017). Revisão de ferramentas para avaliação ergonômica. Revista Produção Online, 18 (2), 665-690. https://www.producaoonline.org.br/rpo/article/view/2925
Leite, A. A., et al. (2022). Ergonomia e saúde: Uma abordagem voltada para o profissional de estética. Revista Estética em Movimento, 1 (2), 162-180. https://revista.fumec.br/index.php/esteticaemmovimento/article/view/7352
Lelis, C. M., Battaus, M. R. B., Freitas, F. C. T. de, Rocha, F. L. R., Marziale, M. H. P., & Robazzi, M. L. de C. C. (2012). Distúrbios osteomusculares relacionados ao trabalho em profissionais de enfermagem: revisão integrativa da literatura. Acta Paulista de Enfermagem, 25(3), 477-482. https://doi.org/10.1590/S0103-21002012000300022
Maciel, V. (2022). LER e DORT são as doenças que mais acometem os trabalhadores, aponta estudo. Ministério da Saúde.<https://www.gov.br/saude/ptbr/assuntos/noticias/2019/abril/ler-e-dort-sao-as-doencas-que-mais-acometem-os>- trabalhadores-aponta-estudo
McAtamney, L., & Corlett, E. N. (1993). RULA: a survey method for the investigation of workrelated upper limb disorders. Applied Ergonomics, 24 (2), 91-99.
Ministério do Trabalho e Emprego. (2024). Norma Regulamentadora n 17 (NR-17).<https://www.gov.br/trabalho-e-emprego/pt-br/acesso-a-informacao/participacaosocial/conselhos-e-orgaos-colegiados/comissao-tripartite-partitaria-permanente/normasregulamentadora/normas-regulamentadoras-vigentes/norma-regulamentadora-no-17-nr>- 17
Ministério do Trabalho e Emprego. (2022). Tabela de classificação e tipificação de NRs e anexos.<https://www.gov.br/trabalho-e-emprego/pt-br/acesso-a-informacao/participacaosocial/conselhos-e-orgaos-colegiados/comissao-tripartite-partitariapermanente/arquivos/normas-regulamentadoras/tabela-de-classificacao-tipificacao-de>- nrs-e-anexos-2022_12_14.pdf
Ministério do Trabalho e Emprego. (s.d.). Normas regulamentadoras.<https://www.gov.br/trabalho-e-emprego/pt-br/acesso-a-informacao/participacaosocial/conselhos-e-orgaos-colegiados/comissao-tripartite-partitaria-permanente/normas>- regulamentadora
Neto, I. R., et al. (2018). Estudo de diferentes métodos ergonômicos e suas aplicações. Revista Ação Ergonômica, 13 (1), 222-229. https://app.periodikos.com.br/article/10.4322/rae.v13e201816/pdf/abergo-13-1- v13e201816.pdf
Observatório de Segurança e Saúde no Trabalho. (2023). Panorama estadual - Ceará (2017- 2021). https://smartlabbr.org/observatoriosst
Rosa, M. A. G., & Quirino, R. (2017). Ergonomia, saúde e segurança no trabalho: interseccionalidade com as relações de gênero. Revista Cientec, 9 (3), 51-66. https://revistas.ifpe.edu.br/cientec/article/view/156/42
Rosa, V. I., Merino, E. A. D., & Figueiredo, L. F. G. (2014). Ergonomia e design: aplicação da ferramenta RULA no ambiente da Cooperativa Colimar. Revista Ação Ergonômica, 9 (1), 96-105. http://www.abergo.periodikos.com.br/article/627
Silva, J., & Pereira, A. (2021). Visão computacional na análise da postura humana: um estudo de caso. Revista de Engenharia Biomédica, 30 (2), 123-130. Rev. Gest. Soc. Ambient. | Miami | v.19.n.7 | p.1-21 | e012801 | 2025. https://dspace.mackenzie.br/bitstream/handle/10899/31075/896-Artigo%20Final-4060-1- 4-20211209.pdf?sequence=1
Silva, W. N. da, Lima, E. H. B., Oliveira, F. A. P. M., Melo, M. F. S., Santos, M. W. C. dos, Paulo, W. M. A., & Lima, M. S. (2024). Análise da produção científica em ergonomia e segurança do trabalho nos últimos 20 anos: Um enfoque bibliométrico utilizando a plataforma Scopus. Revista de Gestão Social e Ambiental, 18(12), 1-19. https://doi.org/10.24857/rgsa.v18n12-187
