[guided]We need to justify the curvature identity rather than treat it as a formal slogan. The family $\pi:\mathcal X\to U$ is smooth, proper, and Kähler, and it is polarized by the fixed Kähler class. Hence the middle cohomology local system $R^n\pi_*\mathbb C$ carries a polarized variation of Hodge structure with flat bundle $\mathcal H^n$, Hodge filtration $F^\bullet$, and Gauss-Manin connection $\nabla$.
By [Griffiths Transversality][citetheorem:9129], differentiating a section of $F^n$ in a holomorphic tangent direction lands in $F^{n-1}$. Applied to the chosen local frame $\Omega_{\bullet}$ and the tangent vector $v\in T_0S$, this gives
\begin{align*}
\nabla_v\Omega_{\bullet}\big|_0\in F^{n-1}H^n(X,\mathbb C).
\end{align*}
The Hodge decomposition at the central fibre splits $F^{n-1}H^n(X,\mathbb C)$ as $H^{n,0}(X)\oplus H^{n-1,1}(X)$. Since $H^{n,0}(X)$ is spanned by $\Omega$, there are unique $a_v\in\mathbb C$ and $\eta_v\in H^{n-1,1}(X)$ such that
\begin{align*}
\nabla_v\Omega_{\bullet}\big|_0=a_v\Omega+\eta_v.
\end{align*}
The same decomposition for the tangent vector $w\in T_0S$ defines unique $a_w\in\mathbb C$ and $\eta_w\in H^{n-1,1}(X)$ by
\begin{align*}
\nabla_w\Omega_{\bullet}\big|_0=a_w\Omega+\eta_w.
\end{align*}
Now apply [Griffiths Curvature Formula for Graded Hodge Bundles][citetheorem:9144] to the polarized variation of Hodge structure just described, with $p=n$. Its hypotheses are exactly the polarized variation data: a flat Gauss-Manin bundle, a Hodge filtration, a polarization, and the Hodge metric induced by that polarization. For the Hodge line $F^n$, the Higgs field is the second fundamental form obtained by projecting $\nabla\Omega_{\bullet}$ from $F^{n-1}$ to $F^{n-1}/F^n\cong H^{n-1,1}(X)$. In the directions $v$ and $w$, these projected components are precisely $\eta_v$ and $\eta_w$.
The cited Griffiths formula gives the Chern curvature of the Hodge line as the negative of the Hodge-metric pairing of this second fundamental form. In the local frame $\Omega_{\bullet}$, the curvature Hermitian form of the Hodge metric is represented by $-\partial\bar\partial\log h$ under the convention fixed in the statement. Thus evaluating this curvature form on $(v,\bar w)$ gives $-\partial_v\partial_{\bar w}\log h|_0$, and the Griffiths sign identifies that quantity with the positive Hodge-Riemann pairing of the horizontal components. With the primitive Hodge-Riemann convention fixed in the statement, the curvature formula gives
\begin{align*}
-\partial_v\partial_{\bar w}\log h\big|_0=\frac{(\eta_v,\eta_w)_{\mathrm{HR}}}{h(0)}.
\end{align*}
Since $h(0)=\|\Omega\|^2$ by definition of the Hodge norm at the central fibre, this becomes
\begin{align*}
-\partial_v\partial_{\bar w}\log h\big|_0=\frac{(\eta_v,\eta_w)_{\mathrm{HR}}}{\|\Omega\|^2}.
\end{align*}
The logarithm is what removes the line component $a_v\Omega$: curvature measures the horizontal second fundamental form, not the change caused by rescaling the chosen frame of the line.[/guided]