We present ground-based photometric observations of secondary eclipses of the hottest known planet KELT-9 b using MuSCAT2 and Sinistro. We detect secondary eclipse signals in i and $z_{\rm s}$ with eclipse depths of $373^{+74}_{-75}$ and $638^{+199}_{-178}$ parts per million, respectively. We perform an atmospheric retrieval on the emission spectrum combined with the data from HST/WFC3, Spitzer, TESS, and CHEOPS to obtain the temperature profile and chemical abundances, including TiO and VO, which have been thought to produce temperature inversion structures in the dayside of ultra-hot Jupiters. While we confirm a strong temperature inversion structure, we find low abundances of TiO and VO with mixing ratios of $\rm {log(TiO)}=-7.80^{+0.15}_{-0.30}$ and $\rm {log(VO)}=-9.60^{+0.64}_{-0.57}$, respectively. The low abundances of TiO and VO are consistent with theoretical predictions for such an ultra-hot atmosphere. In such low abundances, TiO and VO have little effect on the temperature structure of the atmosphere. The abundance of ${\rm e}^{-}$, which serves as a proxy for ${\rm H}^{-}$ ions in this study, is found to be high, with $\rm {log(e^-)}=-4.89\pm {0.06}$. These results indicate that the temperature inversion in KELT-9 b's dayside atmosphere is likely not caused by ${\rm TiO/VO}$, but rather by the significant abundance of ${\rm H}^{-}$ ions. The best-fitting model cannot fully explain the observed spectrum, and chemical species not included in the retrieval may introduce modeling biases. Future observations with broader wavelength coverage and higher spectral resolution are expected to provide more accurate diagnostics on the presence and abundances of ${\rm TiO/VO}$. These advanced observations will overcome the limitations of current data from HST and photometric facilities, which are constrained by narrow wavelength coverage and instrumental systematics.