Output details
11 - Computer Science and Informatics
Kingston University
An object-based comparative methodology for motion detection based on the F-Measure
<23> Performance evaluation of computer vision based visual surveillance algorithms has hitherto been restricted to pixel-based approaches as the concept of ROC "false negatives" is meaningless in object-based approaches (where objects are a more natural unit of evaluation). This paper introduced the F-measure and demonstrated how more realistic object-based performance analysis could be implemented. The paper also established the inevitable presence of "evaluation parameters" and showed how the parameterisation of the evaluation procedure needed to be cognisant of the final application. Both observations have started to make an impact on the development of evaluation methodologies as evidenced by citations.