Search results

This paper aims to introduce and identify a new 3D handling operation (bin picking) for natural discrete food products using food categorisation.

The research shows a new food categorisation and the relation between food ordering processes and food categories. Bin picking in the food industry needs more flexible vision software compared to the manufacturing industry in order to decrease the degree of disarray of food products and transfer them into structure.

It has been shown that there are still manual operated ordering processes in food industry such as bin picking; it just needs new ideas of image processing algorithms such as active shape models (ASMs) on its development in order to recognise the highly varying shapes of food products.

This research was aimed at locating a new ordering process and proving a new principle, but for practical implementation this bin picking solution needs to be developed and tested further.

Identifying new ordering processes via food categorisation is unique and applying ASMs to bin picking opens a new industrial sector (food industry) for 3D handling.

This paper aims to review state of the art of techniques for dimensioning chronic wounds, and validate the possibilities of employing a new proposed optical imaging approach for general task of wound assessment.

Current techniques used for quantifying wound surface are reviewed and evaluated from various perspectives to exam their usability in wound care clinical settings. A photometric stereo (PS) approach will be identified and verified to work as an alternative solution to better satisfy practical requirements on quantifying the dimension of real and mocked wounds.

Both contact and contactless approaches provide some useful functions for wound management; however, new imaging modalities are still required for achieving good portability, affordability and applicability in assisting decision-making in clinical settings. The PS approach can work as a potential solution to provide these functionalities as well as dense geometrical and color texture information of measured areas. The experiments demonstrate that the new approach is able to conveniently produce comparable results to those from latest stereo vision-based techniques.

This work proposed and initially verified the potential of PS technique for the task of wound measurement. Substantial improvements on the prototype and more clinical trials are still required to validate the new technique before it is accepted as a tool for practical wound measurement.

This new PS approach has good potential to reliably measure the dimension of wounds as well as recover their color texture which could contain additional valuable information for predicting a healing procedure for those wound occurring deeper underneath the skin surface.

We present recent results from an EPSRC funded project VirTex (Virtual Textile Catalogues). The goal of this project is to develop graphics and image‐processing software for the capture, storage, search, retrieval and visualisation of 3D textile samples. The ultimate objective is to develop a web‐based application that allows the user to search a database for suitable textiles and to visualize selected samples using real‐time photorealistic 3D animation. The main novelty of this work is in the combined use of photometric stereo and real‐time per‐pixel‐rendering for the capture and visualisation of textile samples. Photometric stereo is a simple method that allows both bump map and colour map of a surface texture to be captured digitally. It uses a single fixed camera to obtain three images under three different illumination conditions. The colour map is the image that would be obtained under diffuse lighting. The bump map describes the small undulations of the surface relief. When imported into a standard graphics program these images can be used to texture 3D models. The appearance is particularly photorealistic, especially under changing illumination and viewpoints. The viewer can manipulate both viewpoint and lighting to gain a deeper perception of the properties of the textile sample. In addition, these images can be used with 3D models of products to provide extremely accurate visualisations for the customer. Until recently, these images could only be rendered using ray‐tracing software. However, recent consumer‐level graphics cards from companies such as Nvidia, ATI and 3Dlabs provide real‐time per‐pixel shading. We have developed software that takes advantage of the advanced rendering features of these cards to render images in real‐time. It uses photometrically acquired bump and colour maps of textiles to provide real‐time visualisation of a textile sample, under user‐controlled illumination, pose and flex.

The purpose of this paper is to describe innovative machine vision methods that have been employed for the capture and analysis of 3D skin textures; and the resulting potential for assisting with identification of suspicious lesions in the detection of skin cancer.

A machine vision approach has been employed for analysis of 3D skin textures. This involves an innovative application of photometric stereo for the capture of the textures, and a range of methods for analysing and quantifying them, including statistical methods and neural networks.

3D skin texture has been identified as a useful indicator of skin cancer. It can be used to improve realism of virtual skin reconstructions in tele‐dermatology. 3D texture features can also be combined with 2D features to obtain a more robust classifier for improving diagnostic accuracy, thereby assisting with the long‐term goal of implementing computer‐aided diagnostics for skin cancer.

The device developed for capturing 3D skin textures is known as the “Skin Analyser”, and as far as the authors know it is unique in the world in being able to recover 3D textures from pigmented lesions in vivo. There currently exist numerous methods for analysing lesions, including manual inspection (using established heuristics commonly known as ABCD rules), dermoscopy and SIAoscopy. The ability to capture and analyse 3D lesion textures complements these existing techniques and forms a valuable additional indicator for assisting with the early detection of dangerous skin cancers such as melanoma.

The paper presents a new approach to texture analysis. The need for a more formal definition of the term surface texture is first identified, and an appropriate texture taxonomy proposed. A method of analysis is described, synthesising innovative elements of machine vision and computer graphics to achieve an object‐centred inspection technique, which is both robust and flexible in application. A selection of experimental results is presented in the paper.

A three-dimensional (3D) printed custom-fit respirator mask has been proposed as a promising solution to alleviate mask-related injuries and supply shortages during COVID-19. However, creating a custom-fit computer-aided design (CAD) model for each mask is currently a manual process and thereby not scalable for a pandemic crisis. This paper aims to develop a novel design process to reduce overall design cost and time, thus enabling the mass customisation of 3D printed respirator masks.

Four data acquisition methods were used to collect 3D facial data from five volunteers. Geometric accuracy, equipment cost and acquisition time of each method were evaluated to identify the most suitable acquisition method for a pandemic crisis. Subsequently, a novel three-step design process was developed and scripted to generate respirator mask CAD models for each volunteer. Computational time was evaluated and geometric accuracy of the masks was evaluated via one-sided Hausdorff distance.

Respirator masks were successfully generated from all meshes, taking <2 min/mask for meshes of 50,000∼100,000 vertices and <4 min for meshes of ∼500,000 vertices. The average geometric accuracy of the mask ranged from 0.3 mm to 1.35 mm, depending on the acquisition method. The average geometric accuracy of mesh obtained from different acquisition methods ranged from 0.56 mm to 1.35 mm. A smartphone with a depth sensor was found to be the most appropriate acquisition method.

A novel and scalable mass customisation design process was presented, which can automatically generate CAD models of custom-fit respirator masks in a few minutes from a raw 3D facial mesh. Four acquisition methods, including the use of a statistical shape model, a smartphone with a depth sensor, a light stage and a structured light scanner were compared; one method was recommended for use in a pandemic crisis considering equipment cost, acquisition time and geometric accuracy.

The application of robots to industrial problems often requires grasping and manipulation of the work piece. The robot is able to perform a task adequately only when it is assigned proper tooling and adequate methods of grasping and handling work pieces. The design of such a task requires an in‐depth knowledge of several interrelated subjects including: gripper design, force, position, stiffness and compliance control and grasp configurations. In this paper, we review the research finding on these subjects in order to present in a concise manner, which can be easily accessed by the designers of robot task, the information reported by the researchers, and identify based on the review, future research directions in these areas.

Automatic defect detection is a fundamental and vital topic in the research field of industrial intelligence. In this work, the authors develop a more flexible deep learning method for the industrial defect detection.

The authors propose a unified framework for detecting defects in industrial products or planar surfaces based on an end-to-end learning strategy. A lightweight deep learning architecture for blade defect detection is specifically demonstrated. In addition, a blade defect data set is collected with the dual-arm image collection system.

Numerous experiments are conducted on the collected data set, and experimental results demonstrate that the proposed system can achieve satisfactory performance over other methods. Furthermore, the data equalization operation helps for a better defect detection result.

An end-to-end learning framework is established for defect detection. Although the adopted fully convolutional network has been extensively used for semantic segmentation in images, to the best knowledge of the authors, it has not been used for industrial defect detection. To remedy the difficulties of blade defect detection which has been analyzed above, the authors develop a new network architecture which integrates the residue learning to perform the efficient defect detection. A dual-arm data collection platform is constructed and extensive experimental validation are conducted.

Describes current research work into the design of a 3‐D vision sensor for use in the field of robot navigation and autonomous vehicles. Outlines the development of a stereo vision system which uses fast data processing to extract feature points in the stereo images and a new fast stereo matching algorithm. Gives results of experiments performed using this system and concludes that the applications require fast, self‐adaptive algorithms which can be processed by parallel processors. This was obtained by means of a special configuration and a highly parallelizable stereo vision process based on the declivity feature matched by dynamic programming.

To realize stable and precise localization in the dynamic environments, the authors propose a fast and robust visual odometry (VO) approach with a low-cost Inertial Measurement Unit (IMU) in this study.

The proposed VO incorporates the direct method with the indirect method to track the features and to optimize the camera pose. It initializes the positions of tracked pixels with the IMU information. Besides, the tracked pixels are refined by minimizing the photometric errors. Due to the small convergence radius of the indirect method, the dynamic pixels are rejected. Subsequently, the camera pose is optimized by minimizing the reprojection errors. The frames with little dynamic information are selected to create keyframes. Finally, the local bundle adjustment is performed to refine the poses of the keyframes and the positions of 3-D points.

The proposed VO approach is evaluated experimentally in dynamic environments with various motion types, suggesting that the proposed approach achieves more accurate and stable location than the conventional approach. Moreover, the proposed VO approach works well in the environments with the motion blur.

The proposed approach fuses the indirect method and the direct method with the IMU information, which improves the localization in dynamic environments significantly.