Reflective tomography lidar (RLT) uses laser detectors to obtain target echo projection data from multiple angles, and then reconstructs the target contour image. Although the spatial resolution is not limited by the detection distance, the acquired detection data is often incomplete due to the limited detection angle. Therefore, while introducing the principles and techniques of lidar reflective tomography, this paper aims at the target contour image reconstruction with the projection data under incomplete angles. Especially in lack of projection data under many angles, a method combines structural sparsity of projection data with super-resolution convolutional neural network (CNN) is proposed for RTL target contour image reconstruction. Unlike introducing sparse solving model in traditional RTL imaging, this paper restores the projection data under missing angles through structural sparsity of the projection data. And then, super-resolution CNN technique is adopted to further recover the restored full-angle projection data. After, the complete reconstruction of target contour image is realized by using RTL imaging method. The laser echo simulation is designed based on the panel method, and combined with the field-measured laser echo data experiment, the image reconstruction capability of the proposed method under different detection conditions is verified.