which can sometimes be inferred from the existing flight track records, but more often it is based on geographic knowledge and clues in photos without any track records. Location. If the images are taken in a long time or the landscape has changed greatly, it is necessary to use more historical images or corroborating data to search and make inferences. "We need to interpret the photos one by one and establish the attribute data.
For example, the traditional cameras in the early days did phone database not have a positioning function, and we often saw that there was only a large mountain ridgeline in the photos. At this time, we must carefully compare the maps and satellite images, and find a way. Find it as close to the correct as possible." Chen Xuanying said. "Creating keywords" may seem like a relatively easy task, but in fact, just deciding which keywords to use is a lot of work. The first step is to identify the scene in the image. For example, what kind of crops are planted in a farmland, other data must be collected to assist in the judgment; secondly, due to the different scales of the aerial photos, the scene that occupies the scene needs to set up keywords. It is also necessary to set rules after discussion;
finally, it is also necessary to think from the user's point of view and set up keywords according to the search habits of ordinary people. Therefore, in the process of establishing the attribute data, the See-Led Zeppelin Foundation also consulted a number of experts to jointly study and plan the keyword construction logic suitable for Taiwan aerial images, and based on this basis, the classification, correction and inspection were carried out to ensure that Images are properly summarized and used.