Road Stereo Target Recognition Based On Image Processing

In recent years,urban road management departments often set up some roadblocks or signs in some sections,some of them are plane patterns,and some of them are three-dimensional directional signs.These plane patterns have a high degree of three-dimensional and authenticity,which makes it impossible for the assistant driving system to judge their authenticity.When there is a sudden situation on the road,it is likely to cause serious accidents due to misjudgment.This paper studies how to distinguish the road plane pattern and the three-dimensional target,and gives a recognition method.The main research contents are as follows:(1)By learning the difference between the plane pattern and the three-dimensional target,and using the stable feature of projective invariant,computer vision theory and algebraic invariant theory are combined to recognize the three-dimensional target.(2)Three filtering principles are added,and an improved algorithm based on orb feature point matching is proposed.The algorithm is used to calculate the projective invariants of point elements,which improves the accuracy and convenience of calculation.(3)In view of the limitation of coplanar five point invariant recognition method,this paper proposes that the projective geometry of an object can be calculated by selecting the spatial points of the object through image matching,and verifies the invariance of the same object at different angles through experiments.(4)The three-dimensional modeling software Solidworks is used to simulate coplanar five-point and six-point geometric invariants respectively.The results show that the calculated value of the spatial six-point of a real 3D object at different perspectives is an invariant,and a flat pattern with 3D effects is a variation,and the calculation method is used to test the real and fake 3D objects on the real road.Finally,the actual measurement experiment also verifies that the geometric quantity calculated from the six points in space can be used for the detection of three-dimensional targets on the road.