how to optimize real-time stitching by OpenCV

Hello everyone! I am trying to make a real-time stitching function from camera in Unity. After watching some tutorials, I have success to make a simple real-time stitching from camera and this is my work. However, you may find there are some problems in the video. First, there is shaking every frame from the stitched video; second is the stitched video color is different from the left-hand side; and finally, there is a conspicuous line from the stitched video. Therefore, I want to know how to fix the problem and thanks everyone who seeing my question. Here is my code of the project, and I am using OpenCVforUnity for this project.

   void OnStitch1()
 {
     pr2 = Stitching(second_mat_gray, first_mat_gray);
 
     Utils.matToTexture2D(pr2, _Texture1);
     output.texture = _Texture1;
     output.GetComponent<AspectRatioFitter>().aspectRatio = (float)_Texture1.width / (float)_Texture1.height;
 }
 
 Mat Stitching(Mat first_mat_temp, Mat second_mat_temp)
 {
     //finding the keypoint between two streaming
     detector.detect(first_mat_temp, keypointsSrc);
     extractor.compute(first_mat_temp, keypointsSrc, descriptorsSrc);
 
     detector.detect(second_mat_temp, keypointsScene);
     extractor.compute(second_mat_temp, keypointsScene, descriptorsScene);

    //Matching the keypoints
     matcher.clear();
     matcher.match(descriptorsSrc, descriptorsScene, matches);
 
     matchesList = matches.toList();

     //After matching, trying to find the good match point by distance
     for (int i = 0; i < descriptorsSrc.rows(); i++)
     {
         dist = (double)matchesList[i].distance;
         if (dist < min_dist)
             min_dist = dist;
         if (dist > max_dist)
             max_dist = dist;
     }
 
     good_matches.Clear();
     for (int i = 0; i < matchesList.Count; i++)
     {
         if (matchesList[i].distance < 10 * min_dist)
             good_matches.Add(matchesList[i]);
     }
     GM = good_matches.Count;

     //Adding Keypoint to List for finding Homography with the OpenCV function
     keypoints_objectList = keypointsSrc.toList();
     keypoints_sceneList = keypointsScene.toList();
     keypointsSrc.release();
     keypointsScene.release();
 
     for (int i = 0; i < good_matches.Count; i++)
     {
         objList.Add(keypoints_objectList[good_matches[i].queryIdx].pt);
         sceneList.Add(keypoints_sceneList[good_matches[i].trainIdx].pt);
     }
 
     MatOfPoint2f obj = new MatOfPoint2f();
     MatOfPoint2f scene = new MatOfPoint2f();
 
     obj.fromList(objList);
     scene.fromList(sceneList);
     H = Calib3d.findHomography(obj, scene, Calib3d.RANSAC, 10);

    //Start warp the streaming from second camera to connect the streaming from first camera
    //Trying to stop the shaking, but not smooth enough
     warp_mat = second_mat.clone();
 
     if (GM - gmCount > 10)
     {
         gmCount = good_matches.Count;
         tempH = H.clone();
         Imgproc.warpPerspective(second_mat, warp_mat, H, ims);
     }
     else
     {
         Imgproc.warpPerspective(second_mat, warp_mat, tempH, ims);
     }
     if (gmCount > 400 || GM < 100)
         gmCount = GM;
 
     //Last Step for stitching as a whole Mat
     half = new Mat(warp_mat, _rect);
     first_mat.copyTo(half);
     return warp_mat;
 }