To use a Scatchard plot, you have to convert the fluorescence intensity measurements into concentrations of bound and free ligand. The Scatchard plot is a linearization. To avoid having to prepare this secondary plot, you can simply plot fluorescence versus concentration of the titrant and fit the data to the binding isotherm equation or Hill equation using nonlinear regression. This will give you the Kd of the binding pair and the Hill coefficient. (If you are working in the tight binding region, you will have to use the Morrison equation instead).

Personally, I don't think you can use fluorescence intensity to determine the concentration of ligands, which is essential to measure Kd.

Fluorescence intensity is an arbitrary value of the protein, instrument, solvent, salt concentration, etc. In other words, the fluorescence intensity you measure is not transferrable to other researchers like molar absorptivity (Beer's Law).

If you have a ligand that is uniquely fluorescent in a bound and free state, you can measure an apparent Kd by the 50% binding.

As Steingrimur said, in general it will not be possible to calculate concentrations of ligand and protein (both free and bound) from fluorescence measurements. However, if the binding affinity is not high and the employed ligand concentration is much higher than that of the protein, you can use the total ligand concentration as the free ligand concentration [L] (neglecting ligand depletion as a result of binding), and using the initial and limiting value (at high ligand concentration) of fluorescence you can calculate the saturation fraction, Fb. Then, you have [L] and Fb, and you can represent Fb/[L] as a function of Fb, which is the Scatchard representation.

The Scatchard representation suffers severe drawbacks and it should be used qualitatively. Therefore, as Adam said, it is advised to plot fluorescence as a function of total ligand concentration and analize the data considering the exact binary equilibrium to calculate free and bound ligand and protein (i.e. Morrison equation).

If it is the intrinsic tryptophan fluorescence of the protein that is being measured and the concentration of the ligand is being titrated, then you potentially have two different signals that can be used to measure binding of the ligand: the fluorescence intensity (it could either increase or decrease), or the peak emission wavelength of fluorescence (blue shift or red shift). If the ligand is itself fluorescent, this experiment can be impractical. If the ligand absorbs either the excitation or emission light, or both, then it becomes necessary to correct for this "inner filter effect." The correction is done be repeating the titration, substituting N-acetyltryptophanamide (NATA) for the protein and calculating a correction factor for the reduced NATA fluorescence at each ligand concentration.

If it is the ligand's fluorescence that is being measured, then the ligand concentration is fixed at a level sufficient to give a useful fluorescence intensity and the protein concentration is titrated. Either of two signals can be used for the binding measurement: the fluorescence intensity (increasing or decreasing) or the fluorescence polarization (or anisotropy) increase. Fluorescence coming from the protein should be measured by repeating the titration without the ligand, and this fluorescence, if any, should be subtracted. If a fluorescence polarization measurement is being performed, the parallel and perpendicular fluorescence intensities of the protein should be subtracted separately, then the polarization (or anisotropy) should be calculated from the corrected values.