The NormalizedPosition(...) method inputs the player's character position and the target GameObject's position. It has the goal of outputting the relative position of the target to the player, returning a Vector3 object (actually, a C# struct – but we can think of it as a simple object) with a triplet of X, Y, and Z values. Note that since the radar is only 2D, we ignore the Y-value of the target GameObjects, so the Y-value of the Vector3 object that's returned by this method will always be 0. So, for example, if a target was at exactly the same location as the player, the X, Y, and Z of the returned Vector3 object would be (0, 0, 0).
Since we know that the target GameObject is no further from the player's character than insideRadarDistance, we can calculate a value in the -1 ... 0 ... +1 range for the X and Z axes by finding the distance on each axis from the target to the player, and then dividing it by insideRadarDistance. An X-value of -1 means that the target is fully to the left of the player (at a distance that is equal to insideRadarDistance), while +1 means it is fully to the right. A value of 0 means that the target has the same X position as the player's character. Likewise, for -1 ... 0 ... +1 values in the Z-axis (this axis represents how far, in front or behind us, an object is located, which will be mapped to the vertical axis in our radar).
Finally, this method constructs and returns a new Vector3 object with the calculated X and Z normalized values and a Y-value of zero.
The normalized value is one that has been simplified in some way so that its context has been abstracted away. In this recipe, what we are interested in is where an object is relative to the player. So, our normal form is to get a value of the X and Z position of a target in the -1 to +1 range for each axis. Since we are only considering the GameObjects within our insideRadarDistance value, we can map these normalized target positions directly onto the location of the radar image in our UI.