The spectrometric features of mud crabs (Scylla paramamosain) were analyzed aiming for future automatic grading of their internal qualities such as meat yield and ovarian fullness. Since they are currently evaluated and graded manually, there is a strong demand to automate them based on the objective spectrometric analysis. However, developing a practical spectrometric system and grading model with adequate performance and acceptable manufacturability remains challenging. In particular, the effective spectrometric datasets are essential to develop machine-learning based grading models, which require repetitive prototyping of various spectrometric systems optimized for in vitro, semi-in vivo, and in vivo conditions. In this study, we measured transmission spectra of essential crab components (meat, ovary, liver, and shell) under in-vitro condition and analyzed their spectrometric features in detail as the first prototyping stage. The spectral data were acquired with standard and concise spectrometers across various time points from just after killing up to 24 h. Their PCA results with the standard spectrometer in the wavebands from 450 to 850 nm revealed significant differences between the crab components as well as their time-course degradation, demonstrating their effectiveness as essential spectrometric features in machine learning. The results with the concise spectrometer in the wavebands from 410 to 940 nm also revealed component-specific spectra even in three-waveband domains, suggesting the applicability of conventional color CMOS camera for limited-accuracy applications as well as the usefulness of NIR wavebands to increase the performance. These findings are useful in preparing effective spectroscopic datasets for machine-learning based quality grading models, and may strongly assist the development of practical spectrometric systems.