Aiming at the problem of poor disparity quality in binocular views with weak texture and weak geometric features, a stereo matching algorithm adapted to cross matching blocks of weak texture and geometric features is proposed. Adopting a bilinear disparity interpolation framework to improve time efficiency; Calculate the overall trend of disparity variation through a small number of regular points in a binocular view, in order to determine the maximum disparity of the view and the disparity of the regular points in less time; To enhance the quality of disparity generation in views with weak texture features, the cost of cross matching blocks is calculated by fusing information such as census sequences, colors, and gradients; Based on the local geometric features of the initial disparity image, the cost calculation is performed by rotating the optimal angles of the cross matching block and sub blocks to enhance the disparity generation quality of views with weak geometric features. The experimental results show that compared with algorithms such as SGM, AD-Census, PMS, ELAS, etc., the generated disparity images have better overall quality. For weak texture and weak geometric feature local areas, the disparity quality is significantly improved. Based on 6 pairs of test images, compared with the PMS algorithm, the MSE of disparity images is reduced by an average of 30.97%, and the average execution time is 1/4 of it.