Finding Field of View Overlap by Motion Analysis

Abstract: Network cameras has in the recent years become more powerful. Each camera is independent and has its own surveillance task. It is reasonable that network cameras in the future should cooperate together to increase surveillance effectiveness. There is a need to find cameras sharing the same field of view in order for an operator to switch perspective. This thesis investigates how multiple network cameras can cooperate by finding the shared field of view between cameras. With the shared field of view, we implement an additional knowledge above system of network cameras and new use cases arises. Our method consist of applying a grid of cells on each camera's video stream and study movement detection. We gather contradicting proof of connectedness between each cell in the whole network of cameras. Our method avoids problems with feature detection such as different perspectives or image quality. We found that our method works with promising results and we can find shared field of view between cameras. There is a limitation in memory of storing all cells and we can only find overlap in regions with movement. This field has not been researched so much, making evaluation hard, as many approaches focuses on feature detection.