Road safety continues to be a major concern in low-middle-income countries (LMICs). In India, approximately 37% of road crash fatalities occur on rural two-lane roads. Segmenting the roadways into sections with an appropriate segment length is the initial and crucial step in statistical crash analysis. The existing segmentation methods depend on individual experience and engineering judgment. This investigation aims to identify critical crash contributing factors affecting fatal crash occurrence on rural two-lane roads, which have been discretized into segments of an appropriate length using power spectral analysis. The study methodology is based on power spectral analysis of crashes in a one-dimensional Spatial Frequency Domain. Based on this finding, the power spectral density and power spectral percentage are calculated in order to determine the power spectral segment length, also termed Optimal Segment Length (OSL). This study utilized three years (2017–2019) of fatal crash incidents and several Negative Binomial (NB) model extensions with OSL to assess the impact of roadside infrastructure on fatal crash occurrence on two rural two-lane National Highways (NH) in Haryana, India. The crash analysis incorporated explanatory variables such as roadway geometric features, traffic characteristics, roadside Points of Interest and population. The model results showed that the Random Parameter Negative Binomial (RPNB) model and the Correlated Random Parameter Negative Binomial (CRPNB) model performed better than the standard NB model. The model findings indicated that variables such as the presence of fuel stations, eateries, culverts, and minor access points along highways were identified as significant risk factors for fatal crashes.