Aiming at the problem of how to obtain accurate positional information of objects in unstructured scenes by depth cameras with limited hardware device resources, a target position detection method based on bi-directional fusion of texture and depth information is proposed here. In the learning phase, the two networks adopt the full-flow bidirectional fusion (FFB6D) module, the texture information extraction part introduces the lightweight Ghost module to reduce the computation of the network, and adds the attention mechanism CBAM that can enhance the useful features, and the depth information extraction part extends the local features and multilevel feature fusion to obtain more comprehensive features; in the output phase, in order to improve the efficiency utilizes the instance semantic In the output stage, the segmentation results are utilized to filter the background points for efficiency, and then 3D keypoint detection is performed, and finally the position information is obtained by the least squares fitting algorithm. Validated on LineMod, occluded LineMod and YCB-Video public datasets, its accuracy reaches 99.8%, 66.3% and 94%, respectively, and its network speed is improved by 30% compared with the previous one, the improved position estimation method guarantees the accuracy and improves the speed at the same time.