Comparing established visitor monitoring approaches with triggered trail camera images and machine learning based computer vision

Abstract

The management of protected areas and other recreational landscapes is subject to a variety of challenges. One aspect hereof, visitor monitoring, is crucial for many management and valuation tasks of ecosystem services. Its core data are visitor numbers which are costly to estimate in absence of entry fees for protected areas or recreational landscapes. Camera-based approaches have the potential to be both, accurate and deliver comprehensive data about visitor numbers, types and activities. So far, camera-based visitor monitoring is, however, costly due to time consuming manual image evaluation. To overcome this limitation, we deployed a convolutional neural network and compared its hourly counts against existing visitor counting methods such as manual in-situ counting, a pressure sensor, and manual camera image evaluations. Our study is the first one to implement, and explicitly assess the performance of a computer vision approach for visitor-monitoring. The results showed that the convolutional neural network derived comparable visitor numbers to the other visitor counting approaches regarding visitation patterns and numbers of visits. Further, our approach also allowed for counting dogs and recreational equipment such as backpacks and bicycles in automatic manner. We thus conclude that it is a fast and reliable method that could be used in protected areas as well as in a much wider array of visitor counting settings in other recreational landscapes. Managers of protected and recreational areas could benefit from our comparisons of convolutional neural network camera image evaluations with existing visitor counting approaches as: • Time-consuming manual image evaluation can be replaced by computer vision approaches based on convolutional neural networks (40 h to manually analyze more than 13,000 images by one expert vs. 10 h to do it automatically in the background). • In contrast to pressure sensors, this approach also allows to differentiate visitor types and activities (dog-walking, cycling, etc.) at comparably low-costs. • Future efforts should concentrate on training specific convolutional neural networks dedicated to visitor monitoring in recreational settings which could process imagery at real-time in the field using single-board computers. • Nevertheless, this approach is prone to the usual disadvantages of camera-based visitor monitoring (risks of theft, vandalism, malfunctioning; data security issues), which need to be considered when setting up the device.