Text this: KE-RSIC: Remote Sensing Image Captioning Based on Knowledge Embedding