Visible/near-infrared hyper-spectra technique was applied to detect peroxidase (POD) content in cucumber leaves with early bacterial angular leaf-spot disease (BALD) stress. A total of 120 samples with three BALD infection stages were used to acquire with hyper-spectra with in the range of 380~1 030 nm (512 wavelengths) and corresponding POD content in the leaves were measured with spectrophotometer method. Analysis of variance (ANOVA) were applied to process statistical analysis of the measured POD content and results showed that there was significant difference (p=005) about the POD content in the cucumber leaves with the three stages of BALD stress. SPXY method was employed to divide the samples into a calibration set (n=80) and a prediction set (n=40). Random Frog (RF) and regression coefficient (RC) were used to select the characteristic wavelengths to reduce data dimensions. Partial least square regression (PLSR) was applied to develop the quantitative relationship between spectra and POD content. RF-PLSR was the optimal model to predict POD content with correlation coefficient (r) of 0816 with root mean squared error of prediction (RMSEP) of 11235. The result showed that hyper-spectra technique combined with chemometrics method was promising for detecting POD content in the cucumber leaves with different BALD developments. This study provided a theoretical reference for early detecting disease infection in non-destructive way.