| Binocular stereo vision is one of the key technologies in the field of computer vision.Because of its advantages such as non-contact,simple structure,and low cost,it is widely used in three-dimensional scene reconstruction,industrial inspection,robot navigation,medical devices and emerging automatic Human driving,virtual reality,face recognition and other fields.As the core technology of binocular vision,stereo matching technology has been one of the hot topics studied by scholars at home and abroad in recent years.The purpose of stereo matching is to determine the pair of corresponding matching points in the left and right image pairs by the matching algorithm so as to convert the two-dimensional image data into three-dimensional data by matching the parallaxes of the pair of points.The current stereo matching algorithm has certain limitations.For some occasions with high measurement accuracy or real-time requirements,the measurement accuracy and the real-time performance of the algorithm need to be improved.How to apply the stereo matching algorithm to binocular stereo vision effectively System,taking into account the matching accuracy and real-time performance is still a problem to be solved.In this paper,we focus on the local stereo matching algorithm based on adaptive weights and local entropy,and the graph-cutting global stereo matching algorithm based on adaptive weights.Based on the proposed algorithm,we design a specific example of the binocular stereo vision system-robot vision.Guidance and work-piece measurement to further illustrate the practicality of this research.The main innovations of this article are as follows:Aiming at the problem that the traditional local stereo matching algorithm has poor matching effect in the weak texture region and the overall matching accuracy is low,a local stereo matching algorithm based on adaptive weights and local entropy is proposed.According to the similarity and similarity of the Gestalt theory,this algorithm proposes a new weight value calculation method in gray-scale space and Euclidean space of the image,which can improve the accuracy of the algorithm while improving the computational speed of the algorithm;in the cost function Add data smoothing items,and use the local entropy of the image to determine the weighting coefficient of the smooth item;use the median filter,left and right consistency check,weighted median filter,and sub-pixel processing to perform multi-step parallax processing on the initial disparity map obtained by disparity selection.In particular,when the matching window is relatively small,a high-precision disparity map can also be obtained.Using the four international standard stereoscopic image pairs provided on the Middlebury website,the algorithm was tested.Experiments show that the proposed algorithm can obtain more accurate disparity maps.Aiming at the low precision of the traditional stereograph matching global stereo matching algorithm,this paper proposes a global stereo matching algorithm based on adaptive weights.The algorithm first uses the new weight value proposed in this paper as the data item of the energy function,and uses the gradient information of the image as the smoothing item of the image.Then,the model is solved using the graph cut theory and the extended algorithm;the optimal parallax distribution is obtained.Using the four international standard stereoscopic image pairs provided on the Middlebury website,the algorithm was tested.The experiments show that the average correct matching rate of the proposed algorithm is 94.2%.In order to verify the practicability of the image stereo matching algorithm proposed in this paper,the robot vision guidance and workpiece size measurement experiment based on the adaptive stereograph-based global stereo matching algorithm are designed.Using the algorithm proposed in this paper for stereo matching can obtain a more accurate parallax map.The result of robot vision guided experiments shows that the algorithm can meet the guidance requirements of industrial robots;the experimental results of workpiece measurement show that the algorithm has higher The measurement accuracy in the x direction is 0.22mm;the measurement accuracy in the y direction is 0.24mm;the measurement accuracy in the z direction is 0.33 mm. |
