Crowdsourced testing is an emerging trend in software testing, which relies on crowd workers to accomplish test tasks. Thus, who performs a test task is extremely important for detecting bugs and covering key points of test requirements in crowdsourced testing. There are a lot of candidate crowd workers who may have different testing experience but can also produce duplicate test reports for the same task due to the lack of cooperation. As crowd workers can freely participate in a test task, high quality of testing in terms of bug detection and coverage of key points of test requirements is not guaranteed. Thus, to improve bug detection and coverage of key points of test requirements, selecting an appropriate subset of workers to perform a test task is becoming an important problem. In this paper, three motivating studies are first conducted to investigate important aspects of workers in detecting bugs and covering key points of test requirements. Accordingly, the studies identify three aspects:initiative, relevance and diversity are identified, and produce a novel approach for selecting workers considering all these three aspects. This new approach is evaluated based on 46 real test tasks from Baidu CrowdTest, and the experimental results show the effectiveness of the approach.