School of Science

Research

Wide-ranging applications in machine vision and AI

Humans have eyes to see and brains to think and make decisions. Intelligent machines, such as robots and driverless vehicles, use cameras to see. Machine vision is about developing algorithms and technology to detect various objects and human from video and images, so that intelligent machines can understand what is going on through their “camera eyes” and make right decision. Dr Baihua Li is involved in a number of projects with exciting real-world impact.

Automated Football Player Performance Analysis – funded by Innovate UK

Football coaches and clubs need detailed performance data for each player (e.g. kicking/passing the ball, using left or right foot, successful or not, and their position and interaction with the ball and other players). Currently watching video back to collect and analyse such data is very time consuming.

Funded by the Innovate UK, Dr Li and her colleagues are developing computer vison and deep learning-based technology to automatically track players, detect body limbs and poses, and recognise various action events in football match videos.  Such machine vision technology allows automatic video processing for player performance analysis.

In collaboration with the  industrial project partner, some outcomes from this project are already undergoing commercial trials with football clubs.

Embedded AI for Food Production Automation - funded by Innovate UK

Automation in food industry is lagging behind on state-of-the-art technology to a large extent. The standard sandwich production line demands many workers but the industry faces a workforce shortage and a low production rate. 

High-tech automation machines (e.g. Millitec's 'Delta' robots) can locate bread slices using shape information obtained from expensive laser cameras but, so far, can only perform simple tasks such as picking up breads.

Working with Millitec, the Loughborough team is applying their expertise to the development of low-cost, reliable, embedded machine vision solutions for the robots, allowing a range of new automation functions. Real-time AI algorithms can teach the robot to recognise a variety of food items, effectively allowing the robot to carry out the complex tasks of identifying and picking up ingredients, assembling the sandwich and packaging it.

Autonomous Underwater Vehicles – EPSRC CDT

Autonomous underwater vehicles are in high demand for underwater environment protection, scientific research on marine organisms, search and rescue, inspection and many industry and commercial applications.

Dr Li and her colleagues are working with an Italian company on an autonomous underwater vehicle (AUV) project funded by EPSRC CDT.

Currently, the robot vehicle is controlled manually through a cable. The aim of the project is to develop autonomy technology for underwater robot to enable autonomous navigation and exploration.

We hear a lot these days about driverless vehicles on land, but the underwater environment presents additional challenges - no GPS, poor conditions for cameras and other sensors, and an environment that is largely unstructured and dynamic.

There are many rarely seen or unknown species and plants to explore which makes this development project very challenging and valuable. In addition, due to the on-board computation capacity and power consumption limitation for AUVs, the developed AI algorithms and embedded system need to be portable and robust at low energy consumption.  Loughborough University’s research on curiosity-driven exploration behaviour in unknown environment will take the technology to the next level.

Agri-Robots for Precision Spraying - funded by Innovate UK

Across the world there is an urgent need to increase farming productivity and associated yields while, at the same time, reducing pesticide use.  As such, the development of agricultural robots utilising intelligent crop-spraying is widely seen as the answer to improving farming productivity as well as cost and pollution reduction.

The Loughborough University team is working on a hybrid multi-layer robot control architecture, sensing, avoidance, and navigation capabilities, as well as sensor-based situation understanding and high-level reasoning in real agriculture environment.  These state-of-the-art Agri-Robots will be able to navigate orchards, recognising fruit trees, and spraying farm chemicals precisely to reduce chemical waste and pollution in soil and air.  

Though still in the early development phase, this technology will bring a new low-cost light weight alternative product and service to the market, where significant growth of demand on such product is expected.