众包模式在大规模遥感影像信息提取领域的探索

Exploration of crowdsourcing in information extraction from remote sensing images

基于地理空间数据云平台,对基于众包的大规模遥感影像信息提取模式进行了研究,提出了一套完整的流程体系,并通过多时期的青藏高原湖泊提取任务对模式的报酬发放机制、任务分配方式、任务划分方法、人才激励等领域进行了探索和完善。实验结果表明提前支付部分报酬并采用小组的方式,对提高数据质量和控制并没有很大影响,而积累人才对获取高质量的数据结果很重要。由于该模式集成了众包和机器计算能力,且对遥感影像处理是通用的,因此可用于更多的需要人工参与的海量遥感影像处理工作中。

Based on geospatial data cloud（GSCloud）, the application of crowdsourcing in large scale information extraction from satellite images was studied, and a systematic architecture of this paradigm was proposed. By performing an experiment of extracting lakes on Qinghai-Tibetan plateau from landsat images, various aspects of the paradigm like the incentive mechanism, task assignment method, task division and many others were explored. Results show that paying part of the reward in advance and assigning a task to a team instead of individuals do not help attracting more applicants and improving the quality of results. And the accumulation of talents is of critical importance to obtain high-quality task results. Since this paradigm integrates crowdsourcing and machine computing power, and it is generic, it can be applied in more massive remote sensing image processing work which requires much human intervention.