In order to solve the problem of external damage and flashover caused by the tree barriers in the channel of transmission lines

the characteristics of binocular stereo vision that it could create the scene of 3D reconstruction were adopted to study the monitoring technology of tree barriers around transmission lines. Based on the over green feature of RGB image and image segmentation method

a model of tree interest region was established

by which the problem of low efficiency of stereo matching in complex background environments could be effectively solved. Aiming at the problem that the current block matching algorithm’s accuracy is not high in target edge matching and distance measurement

we designed a variable window block matching method based on region of interest

and adopted variable window as cost aggregation method to fit tree obstacles and to improve the matching accuracy of edge

so as to improve the distance measurement. In the experimental stage

the matching and ranging accuracy of the fixed window matching algorithm

multi window matching algorithm and the algorithm proposed in this paper are compared by using Middlebury image set and simulation scene image. Through the experimental data

it can be found that

for the two algorithms mentioned above

the algorithm in this paper has higher matching accuracy in the non edge area and depth discontinuity area. At the same time

it also has a smaller error value in ranging; meanwhile

with the increase of the ranging distance

the increment rate of error value promotion is also the slowest.