Abstract：
Checking each item placed in a separate collection bin of recyclables to examine contamination is often difficult for a researcher relying on such data. This is because of the time and inconvenience involved to manually identify items. We test a proof-of-concept experiment on the ability of trail cameras to identify items placed within separate collection bins. After a pre-test of seven camera models, we selected one with the best image quality. We use this camera for lab and field trials to count the number of identifiable items based on photos compared to manual hand-counts of the items. Three lab trials of this camera resulted in an average of 82% accuracy in item identification. We then conducted a field experiment, testing photo quality to identify items in six separate collection bins across a university campus over a one-month period with a total of over 9,700 photos. Of the 1343 items placed in the separate collection bins, the trail cameras provided photographs of high enough quality such that successful identification occurred for 68.5% of the items, with poor identification for paper items and small items. We conclude that trail cameras can be useful for data collection in separate collection behavior, especially for items with the largest surface size greater than a credit card.
摘要：
对于依赖此类数据的研究人员来说，检查放置在单独的可回收物品收集箱中的每个物品以检查污染通常是困难的。这是因为手动识别项目所涉及的时间和不便。我们测试了一个概念验证实验，测试了跟踪摄像机识别放置在单独收集箱中的物品的能力。经过七个相机型号的预测试,我们选择了一个最好的图像质量。我们使用此相机进行实验室和现场试验，以根据照片计算可识别物品的数量，与手动手动计数物品相比。该相机的三次实验室试验导致物品识别的平均准确率为82%。然后我们进行了一个现场实验，在为期一个月的时间内，测试照片质量，以识别整个大学校园六个独立收集箱中的物品，总共有9,700多张照片。在放置在单独收集箱中的1343件物品中，跟踪摄像机提供了足够高质量的照片，从而成功识别了68.5%的物品，对纸质物品和小物品的识别能力较差。我们得出的结论是，跟踪摄像机可以用于单独收集行为的数据收集，特别是对于最大表面尺寸大于信用卡的物品。
