A new algorithm is proposed, through Fourier transformation projection theory using frequency-domain analysis, to detect local linear cracks on the freeway surface. And valid results are obtained. The steps of discrete ridgelet algorithm and the ridgelet transform results of the standard image are represented in detail. The basic condition of reconstructing the original image through Radon transformation is proposed. The method is used to detect actual complex surface, and denoise background noise combined with the histogram equilibrium algorithm. Estimated sample threshold is used to treat the level coefficients of ridgelet transform and get rid of the Random noise. The signal-to-noise ratio of ridgelet transform is better than that of the two-dimensional wavelet transform (greater than 20 dB in the state of low-frequency) and wavelet transform plus Weiner filtering (greater than 3 dB on the average) through counting the different reconstruction coefficients. The local linear crack is picked up through binary image and the resolution factor is within 2 mm.