Pedestrian motion centroid model based on robot dynamics

In recent years, with the rapid development of computer vision technology, image-based human body research has become an important task, such as pedestrian target detection, trajectory tracking, posture estimation and behaviour recognition. The centre of mass is one of the important characteristics that can reflect the phenomenon of pedestrian movement. This paper first introduces the biped robot model in robotics, starting from forward and inverse kinematics, to find the mapping relationship between the position of each joint and the pose of the end effector. Then, corresponding to the skeleton model of the human joint points, the characteristics of the bone posture and joint angle are determined. The moment of inertia factor is introduced, and the motion superposition of different joint points is considered to establish a pedestrian motion centroid model. By calculating the equivalent dynamic centroid, the pedestrian kinematics law can be explored and the pedestrian movement mechanism can be more deeply recognized.

A Study of Athlete Pose Estimation Techniques in Sports Game Videos Combining Multiresidual Module Convolutional Neural Networks

In this paper, we propose a multiresidual module convolutional neural network-based method for athlete pose estimation in sports game videos. The network firstly designs an improved residual module based on the traditional residual module. Firstly, a large perceptual field residual module is designed to learn the correlation between the athlete components in the sports game video within a large perceptual field. A multiscale residual module is designed in the paper to better solve the inaccuracy of the pose estimation due to the problem of scale change of the athlete components in the sports game video. Secondly, these three residual modules are used as the building blocks of the convolutional neural network. When the resolution is high, the large perceptual field residual module and the multiscale residual module are used to capture information in a larger range as well as at each scale, and when the resolution is low, only the improved residual module is used. Finally, four multiresidual module convolutional neural networks are used to form the final multiresidual module stacked convolutional neural network. The neural network model proposed in this paper achieves high accuracy of 89.5% and 88.2% on the upper arm and lower arm, respectively, so the method in this paper reduces the influence of occlusion on the athlete’s posture estimation to a certain extent. Through the experiments, it can be seen that the proposed multiresidual module stacked convolutional neural network-based method for athlete pose estimation in sports game videos further improves the accuracy of athlete pose estimation in sports game videos.

Continuous ergonomic risk perception for manual assembly operations using wearable multi-sensor posture estimation

Purpose This paper aims to perform the real-time and accurate ergonomics analysis for the operator in the manual assembly, with the purpose of identifying potential ergonomic injuries when encountering labor-excessive and unreasonable assembly operations. Design/methodology/approach Instead of acquiring body data for ergonomic evaluation by arranging many observers around, this paper proposes a multi-sensor based wearable system to track worker’s posture for a continuous ergonomic assessment. Moreover, given the accurate neck postural data from the shop floor by the proposed wearable system, a continuous rapid upper limb assessment method with robustness to occasional posture changes, is proposed to evaluate the neck and upper back risk during the manual assembly operations. Findings The proposed method can retrieve human activity data during manual assembly operations, and experimental results illustrate that the proposed work is flexible and accurate for continuous ergonomic assessments in manual assembly operations. Originality/value Based on the proposed multi-sensor based wearable system for posture acquisition, a real-time and high-precision ergonomics analysis is achieved with the postural data arrived continuously, it can provide a more objective indicator to assess the ergonomics during manual assembly.

A Laser Vision System for Relative 3-D Posture Estimation of an Underwater Vehicle with Hemispherical Optics

This paper describes the development and experimental validation of algorithms for a novel laser vision system (LVS), suitable for measuring the relative posture from both solid and mesh-like targets in underwater environments. The system was developed in the framework of the AQUABOT project, a research project dedicated to the development of an underwater robotic system for inspection of offshore aquaculture installations. In particular, an analytical model for three-medium refraction that takes into account the nonlinear hemispherical optics for image rectification has been developed. The analytical nature of the model allows the online estimation of the refractive index of the external medium. The proposed LVS consists of three line-lasers within the field of view of the underwater robot camera. The algorithms that have been developed in this work provide appropriately filtered point-cloud datasets from each laser, as well as high-level information such as distance and relative orientation of the target with respect to the ROV. In addition, an automatic calibration procedure, along with the accompanying hardware for the underwater laser vision system has been developed to reduce the calibration overhead required by regular maintenance operations for underwater robots operating in seawater. Furthermore, a spatial image filter was developed for discriminating between mesh and non-mesh-like targets in the LVS measurements. Finally, a set of experiments was carried out in a controlled laboratory environment, as well as in real conditions at offshore aquaculture installations demonstrating the performance of the system.

An Efficient Approach Using Knowledge Distillation Methods to Stabilize Performance in a Lightweight Top-Down Posture Estimation Network

Multi-person pose estimation has been gaining considerable interest due to its use in several real-world applications, such as activity recognition, motion capture, and augmented reality. Although the improvement of the accuracy and speed of multi-person pose estimation techniques has been recently studied, limitations still exist in balancing these two aspects. In this paper, a novel knowledge distilled lightweight top-down pose network (KDLPN) is proposed that balances computational complexity and accuracy. For the first time in multi-person pose estimation, a network that reduces computational complexity by applying a “Pelee” structure and shuffles pixels in the dense upsampling convolution layer to reduce the number of channels is presented. Furthermore, to prevent performance degradation because of the reduced computational complexity, knowledge distillation is applied to establish the pose estimation network as a teacher network. The method performance is evaluated on the MSCOCO dataset. Experimental results demonstrate that our KDLPN network significantly reduces 95% of the parameters required by state-of-the-art methods with minimal performance degradation. Moreover, our method is compared with other pose estimation methods to substantiate the importance of computational complexity reduction and its effectiveness.