Cross-linked Polyethylene (XLPE) cable is widely used in urban power supply systems because of its excellent performance. Its inner structure-the water-blocking buffer layer-is prone to suffer from ablation defects after a long time of working, which will threaten the safety of the city's power supply. Therefore, the accurate detection of such cable defects is an urgent problem to be solved. In our previous work, we proposed a Local Source-translation Computed Tomography (L-STCT) method for cable detection. This method focuses on scanning part of the cable around the water-blocking buffer layer, and can intuitively observe the internal structure and identify defects with the reconstructed image. However, the used reconstruction method-Simultaneous Iterative Reconstruction Technique (SIRT)-requires massive iterations for final images, lowering reconstruction efficiency. To meet the high-efficiency reconstruction in engineering applications, we propose a Local-detecting Filtered Backprojection (LFBP) method in this paper, as the analytical method is much fast for image reconstruction. The typical analytic algorithm-Filtered Backprojection (FBP)-requires complete and detruncated projection for artifacts-free reconstruction. As L-STCT scans only part of the cable, the X-ray beam cannot cover the entire object, resulting in projection truncation. At the same time, limited source trajectory and detector width cannot achieve at least 180° angular coverage, resulting in projection missing. The L-STCT scan suffers from the problems of projection truncation and limited-angle computed tomography. Performing analytic reconstruction with raw incomplete and truncated projection will produce severe artifacts, affecting defect identification. To suppress the potential artifacts, the projection smoothing method is introduced according to projection characteristics. Projection smoothing mainly uses the cosine function to interpolate data along the detector direction and source sampling direction respectively. At both ends of the detector, the new part of virtual data gradually smoothing to 0 is splicing with the original data, avoiding data dropping to 0 suddenly and alleviating projection truncation. Similarly, starting from the projection sampled at the start and end source sampling points, smoothing is carried out on the part of the original projection collected near the two ends of the source sampling trajectory. The processed data gradually smooth to 0 along the two ends of the X-ray source trajectory to suppress limited-angle artifacts. In the experiment, we respectively used SIRT, FBP, and LFBP methods to reconstruct the image with the same set of projections, and calculated the SSIM and RMSE values to evaluate the image quality quantitatively. The results show that images reconstructed by the FBP method suffer from severe limited-angle and truncated artifacts, which interfere with the basic structure of the cable and potential defect identification. The images reconstructed by SIRT and LFBP are similar in quality, as both methods suppressed artifacts and clarified the basic structure of the cable. A lower RMSE and higher SSIM mean better image quality. Obviously, for ?100 mm cable imaging, the values of RMSE and SSIM are 0.117 0 and 0.923 8 for SIRT, which is close to that of LFBP (0.130 3 and 0.917 2). It proves that the image quality is comparable for both methods. By contrast, a higher RMSE (0.191 6) and lower SSIM (0.657 7) further indicates that artifacts greatly degrade reconstructed image for FBP. A similar result also can be found for ?160 mm cable imaging. For computational consumption, LFBP costs 1.38 s, FBP costs 1.42 s, and SIRT costs 62.46 s. Due to the added additional projection in the projection smoothing process, LFBP reconstruction is slightly longer than that of FBP but can be completely negligible. As the SIRT method gets the final image after 500 iterations, its reconstruction speed is much lower than the other two methods. Therefore, the LFBP method takes into account the reconstruction image quality and reconstruction efficiency, which is more suitable for practical engineering applications.