Pork is the main meat consumption product in China. Its freshness is closely related to the health of residents. At present, the most common detection methods for meat quality include sensory testing, physical and chemical testing, and microbiological testing, but sensory detection is less reliable and comparable. Physical and chemical testing and microbiological testing have many problems, such as time-consuming, complicated operation and destroying samples, thus establishing a fast and nondestructive detection method has great significance. Raman spectroscopy is fast and nondestructive as a detection technology. Moreover, portable Raman spectroscopy provides a new way for food spot detection and is expected to achieve rapid real-time mass detection in the processing industry. At present, there is no study on the rapid detection of physical and chemical indexes of pork freshness by Raman spectroscopy. Therefore, a portable Raman spectrometer was used in this study to detect the freshness of cold storage lean pork. Collecting the Raman spectroscopy of samples with time and monitoring the corresponding freshness index, including total volatile base nitrogen (TVB-N), pH, L*, a*, and b*. Raman spectra were pre-processed by standard normal variable transformation（SNV）, curve smoothing（SG）, normalize（NL）, multiple scattering correction（MSC）, baseline（BL）, and Detrending（DFA）. Partial least squares regression (PLSR) was used to establish a quantitative prediction model of pork freshness indicators based on full displacements of Raman spectroscopy. The results indicated that the PLSR model based on the Raman spectrum had a good performance predicting pork freshness. The optimal model for TVB-N and pH was SNV-PLSR, and the correlation coefficient was 0.948 and 0.886, respectively. The optimal models for color L*, color a*, and color b* were SNV-PLSR, DFA-PLSR, and MSC-PLSR, respectively. The correlation coefficients were 0.827, 0.858 and 0.900, respectively. The regression coefficient method (RC) was used to screen the optimal spectral bands of each index model, and the PLSR model of the optimal spectral bands of each index was established. The results showed that the TVB-N and pH models could be simplified, and only 20% of the spectral bands can achieve a good prediction effect. TheRP of the TVB-N model and pH model were 0.933 and 0.880, respectively. Raman spectroscopy providing us with a spot detection method shows great potential in rapidly detecting pork freshness, especially in predicting TVB-N content.