For the low altitude target elevation direction finding scenario of bistatic multiple input multiple output (MIMO) radar, due to multipath effects, the transmitting and receiving steering vectors lose orthogonality with the noise subspace due to coupling, resulting in the unavailability of subspace class algorithms mainly based on multiple signal classification (MUSIC) in this scenario, However, subspace algorithms based on spatial smoothing preprocessing suffer from low accuracy in angle estimation due to array aperture loss. To solve the above problems, this paper establishes a bistatic meter wave MIMO radar single target and incoherent multi target mirror reflection signal model. Based on the mathematical transformation and analysis of the signal model, a steering vector matrix that is still orthogonal to the noise subspace is discovered, then, a new guidance vector matrix combined with generalized MUSIC and maximum likelihood algorithm was used to propose a joint estimation method for the elevation direction of departure and arrival of low altitude targets in a bistatic meter wave MIMO radar. Finally, simulation experiments verified the effectiveness of the proposed method and the superiority of the elevation direction finding performance.