This thesis focuses on precision agriculture of livestock farming. Precision Livestock Farming is a modern farming development that emphasises on deploying advanced information and communication technology in physical farms to optimise the contributions of individual animals. Relevant techniques such as the Internet of Things (IoT), machine learning and 5G communication are all needed for improving the level of automation, intelligence and efficiency towards smart farming.With the support of more powerful computing resources, we are capable of handling massive volumes of data, where highly automated processes can be applied for data capturing, analysis and improved decision making. In addition to the economic benefits, Precision Livestock Farming also meets several social goals, such as high-quality and safe food products, efficient and sustainable livestock farming, better animal welfare and low footprint to the environment.Manual observation of animals is a tedious job, especially over a long time, and it can often be affected by the observer’s bias. As a result, IoT enabled AI machine learning-based automatic monitoring and management of livestock with image processing and analysis can offer massive potential for producing the unbiased status report in a more efficient and effective way. By utilizing the real-time monitoring technology, and the corresponding management system can boost productivity whilst reducing the cost and environmental emissions in livestock farms.To tackle this emerging issue and needs, a Precision Livestock Farming platform has been designed and implemented in this thesis, with the support of various sensors and corresponding analytics technologies. By continuously recording and analysing live data, it can not only recognise the welfare and health status of the animals but also for evidence-based smart decision-making with the support of the massive volumes of data. For implementing the proposed system, various techniques have been introduced to address the challenging issues, especially real-time multi-camera video streaming and transmission in an on-demand manner for improving the efficiency and efficacy in mobile/cloud-based environments.As a case study, the developed generic platform has been applied for tracking and behaviour recognition of pigs. These include background detection, object detection, object classification and target selection, followed by object tracking and behaviour recognition. The successful application has not only validated the efficacy of the proposed system but also demonstrated the flexibility and great potential of the proposed system in a wide range of application areas.Finally, some future directions are also provided after the summary of the contribution points, which are expected to benefit the further development of the corresponding fields and the automation and intelligence of the livestock farming.
|Date of Award||15 Sep 2020|
- University Of Strathclyde
|Sponsors||University of Strathclyde|
|Supervisor||Jinchang Ren (Supervisor) & James Windmill (Supervisor)|