I've got a question regarding the topic of SLAM. My composition consists of a mobile robot and a top down looking 2D camera with wide-angle lens. It is comparable to a survaillance camera mounted on the ceiling and is able to capture the ground beneath. Although the robot has encoders and a ultrasonic sensor, I was asking myself, if SLAM is really needed? Because I do not need to scale the map by using the camera if I'm right. I take a picture of the environment and by image processing I can create a grid map as well as track the robot. A KalmanFilter (vision data and Odometry) Localization will keep track of the robot in it. So..Is the SLAM obsolet or do I need to do it with the use of sonar? I've read papers about people sketching map by themself which leads to scaling. In this cases the environment gets directle snapped and processed.
I would like to get further and create a map updating for a dynamic environment, which means the map needs to take changes in it and update itself for the robot to use it.
Cheers, best regards and thanks!