LaserOdometry

這一模塊（節點）主要功能是：進行點云數據配準，完成運動估計

利用ScanRegistration中提取到的特征點，建立相鄰時間點云數據之間的關聯，由此推斷lidar的運動。我們依舊從主函數開始：

int main(int argc, char** argv) {ros::init(argc, argv, "laserOdometry");ros::NodeHandle nh;/***** 訂閱6個節點，通過回調函數進行處理 *****/ros::Subscriber subCornerPointsSharp = nh.subscribe<sensor_msgs::PointCloud2>("/laser_cloud_sharp", 2, laserCloudSharpHandler);ros::Subscriber subCornerPointsLessSharp = nh.subscribe<sensor_msgs::PointCloud2>("/laser_cloud_less_sharp", 2, laserCloudLessSharpHandler);ros::Subscriber subSurfPointsFlat = nh.subscribe<sensor_msgs::PointCloud2>("/laser_cloud_flat", 2, laserCloudFlatHandler);ros::Subscriber subSurfPointsLessFlat = nh.subscribe<sensor_msgs::PointCloud2>("/laser_cloud_less_flat", 2, laserCloudLessFlatHandler);ros::Subscriber subLaserCloudFullRes = nh.subscribe<sensor_msgs::PointCloud2> ("/velodyne_cloud_2", 2, laserCloudFullResHandler);ros::Subscriber subImuTrans = nh.subscribe<sensor_msgs::PointCloud2> ("/imu_trans", 5, imuTransHandler);/**** 發布4個節點 *****/ros::Publisher pubLaserCloudCornerLast = nh.advertise<sensor_msgs::PointCloud2>("/laser_cloud_corner_last", 2);ros::Publisher pubLaserCloudSurfLast = nh.advertise<sensor_msgs::PointCloud2>("/laser_cloud_surf_last", 2);ros::Publisher pubLaserCloudFullRes = nh.advertise<sensor_msgs::PointCloud2> ("/velodyne_cloud_3", 2);ros::Publisher pubLaserOdometry = nh.advertise<nav_msgs::Odometry> ("/laser_odom_to_init", 5);//創建對象nav_msgs::Odometry laserOdometry;laserOdometry.header.frame_id = "/camera_init";laserOdometry.child_frame_id = "/laser_odom";//創建TransformBroadcaster對象tf::TransformBroadcaster tfBroadcaster;//創建坐標變換對象tf::StampedTransform laserOdometryTrans;laserOdometryTrans.frame_id_ = "/camera_init";laserOdometryTrans.child_frame_id_ = "/laser_odom";

接下來是設置處理速度并進行初始化

ros::Rate rate(100);bool status = ros::ok(); while (status) {// 進行一次回調函數的調用（完成當前時刻的msg的訂閱和上一時刻的處理后的msg的發布）ros::spinOnce();if (newCornerPointsSharp && newCornerPointsLessSharp && newSurfPointsFlat && newSurfPointsLessFlat && newLaserCloudFullRes && newImuTrans &&fabs(timeCornerPointsSharp - timeSurfPointsLessFlat) < 0.005 &&fabs(timeCornerPointsLessSharp - timeSurfPointsLessFlat) < 0.005 &&fabs(timeSurfPointsFlat - timeSurfPointsLessFlat) < 0.005 &&fabs(timeLaserCloudFullRes - timeSurfPointsLessFlat) < 0.005 &&fabs(timeImuTrans - timeSurfPointsLessFlat) < 0.005) {newCornerPointsSharp = false;newCornerPointsLessSharp = false;newSurfPointsFlat = false;newSurfPointsLessFlat = false;newLaserCloudFullRes = false;newImuTrans = false;/*********初始化 ***********/if (!systemInited) { /********保存上一時刻的數據******/// exchange cornerPointsLessSharp & laserCloudCornerLast pcl::PointCloud<PointType>::Ptr laserCloudTemp = cornerPointsLessSharp;cornerPointsLessSharp = laserCloudCornerLast;laserCloudCornerLast = laserCloudTemp;// exchange surfPointsLessFlat & laserCloudSurfLastlaserCloudTemp = surfPointsLessFlat;surfPointsLessFlat = laserCloudSurfLast;// kd二叉樹 為了方便尋找最近的點kdtreeCornerLast->setInputCloud(laserCloudCornerLast);kdtreeSurfLast->setInputCloud(laserCloudSurfLast);// 發布 laserCloudCornerLast 上一時刻sensor_msgs::PointCloud2 laserCloudCornerLast2;pcl::toROSMsg(*laserCloudCornerLast, laserCloudCornerLast2);laserCloudCornerLast2.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserCloudCornerLast2.header.frame_id = "/camera";pubLaserCloudCornerLast.publish(laserCloudCornerLast2);// 發布 laserCloudSurfLast 上一時刻sensor_msgs::PointCloud2 laserCloudSurfLast2;pcl::toROSMsg(*laserCloudSurfLast, laserCloudSurfLast2);laserCloudSurfLast2.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserCloudSurfLast2.header.frame_id = "/camera";pubLaserCloudSurfLast.publish(laserCloudSurfLast2);// transformSum累計變換了的角度transformSum[0] += imuPitchStart;transformSum[2] += imuRollStart;systemInited = true;continue; //初始化后開始接受下一時刻的訂閱數據}laserCloudOri->clear();coeffSel->clear();//更新位姿中的坐標transform[3] -= imuVeloFromStartX * scanPeriod;transform[4] -= imuVeloFromStartY * scanPeriod;transform[5] -= imuVeloFromStartZ * scanPeriod;

在初始化之后，就能將相鄰兩個時刻的點云哪來進行配準了，具體的做法我們看看論文怎么說

在這里，由于需要高速的處理，論文并沒有采用一一對應的點云配準（事實上，對于點云來說，這也是很難做到的），而是在上一時刻尋找兩點確定一條與當前時刻的角點距離最近的直線作為該角點的配準對應，同理，在上一時刻尋找三點確定一個與當前時刻的平面點距離最近的平面作為該平面點的配準對應。

作者給出了距離的計算公式，于是整個配準過程變成了優化過程——不斷迭代使得距離之和最小。我們來看看作者給出的距離公式

根據下圖能較容易推導出該公式

由三角形面積公式可知2S=c*d=absinC 則d=absinC/c=|a×b|/c

接下來是點到平面的距離

根據下圖能較容易推導出該公式

?

首先看分子：b和c叉乘，得到平面法向量設為h。即要求a和h的點乘的模。等于|a||h|cosα。再看分母：b和c叉乘的模，即法向量h的模|h|。因此，分子除以分母即為a在h上投影的長度|a|cosα，也就是a到平面的距離。

理解了原理，我們來看這一部分的代碼

// 特征點的數量足夠多再開始匹配，cornerPointsSharpNum和cornerPointsSharp都是最新時刻t+1的數據if (laserCloudCornerLastNum > 10 && laserCloudSurfLastNum > 100) {std::vector<int> indices;pcl::removeNaNFromPointCloud(*cornerPointsSharp,*cornerPointsSharp, indices);int cornerPointsSharpNum = cornerPointsSharp->points.size();// 多次迭代 LM求解前后時刻的位姿變化for (int iterCount = 0; iterCount < 25; iterCount++) {laserCloudOri->clear();coeffSel->clear();/******** 特征角點的處理 ********/for (int i = 0; i < cornerPointsSharpNum; i++) {// 每一次迭代都將特征點都要利用當前預測的坐標轉換轉換至k+1 sweep的初始位置處對應于函數 TransformToStart()// 每一次迭代都將特征點都要利用當前預測的坐標轉換轉換至k+1 sweep的結束位置處對應于函數 TransformToEnd()// pointSel是當前時刻t+1的cornerPointsSharp轉換到初始imu坐標系后的點坐標，對角點一個一個做處理，設為i點TransformToStart(&cornerPointsSharp->points[i], &pointSel);// 每5次循環尋找一次鄰域點，否則沿用上次循環的鄰域點if (iterCount % 5 == 0) {std::vector<int> indices;pcl::removeNaNFromPointCloud(*laserCloudCornerLast,*laserCloudCornerLast, indices);// nearestKSearch是PCL中的K近鄰域搜索，搜索上一時刻LessCornor的K鄰域點// 搜索結果: pointSearchInd是索引; pointSearchSqDis是近鄰對應的平方距離(以25作為閾值)kdtreeCornerLast->nearestKSearch(pointSel, 1, pointSearchInd, pointSearchSqDis);/******** 在上一時刻t的LessSharp中用KD樹尋找和點i最近的點j ********/int closestPointInd = -1, minPointInd2 = -1;if (pointSearchSqDis[0] < 25) {closestPointInd = pointSearchInd[0];//最近鄰域點的所在層數int closestPointScan = int(laserCloudCornerLast->points[closestPointInd].intensity);/******** 在上一時刻t的LessSharp中點j附近2層中最近的點l ********/float pointSqDis, minPointSqDis2 = 25;for (int j = closestPointInd + 1; j < cornerPointsSharpNum; j++) {// closestPointScan +(-) 2.5表示只接受(后)前2層(共4層)范圍內的數據if (int(laserCloudCornerLast->points[j].intensity) > closestPointScan + 2.5) { break;}pointSqDis = (laserCloudCornerLast->points[j].x - pointSel.x) * (laserCloudCornerLast->points[j].x - pointSel.x) + (laserCloudCornerLast->points[j].y - pointSel.y) * (laserCloudCornerLast->points[j].y - pointSel.y) + (laserCloudCornerLast->points[j].z - pointSel.z) * (laserCloudCornerLast->points[j].z - pointSel.z);if (int(laserCloudCornerLast->points[j].intensity) > closestPointScan) {if (pointSqDis < minPointSqDis2) {minPointSqDis2 = pointSqDis;minPointInd2 = j;}}}for (int j = closestPointInd - 1; j >= 0; j--) {if (int(laserCloudCornerLast->points[j].intensity) < closestPointScan - 2.5) {break;}pointSqDis = (laserCloudCornerLast->points[j].x - pointSel.x) * (laserCloudCornerLast->points[j].x - pointSel.x) + (laserCloudCornerLast->points[j].y - pointSel.y) * (laserCloudCornerLast->points[j].y - pointSel.y) + (laserCloudCornerLast->points[j].z - pointSel.z) * (laserCloudCornerLast->points[j].z - pointSel.z);if (int(laserCloudCornerLast->points[j].intensity) < closestPointScan) {if (pointSqDis < minPointSqDis2) {minPointSqDis2 = pointSqDis;minPointInd2 = j;}}}}pointSearchCornerInd1[i] = closestPointInd;pointSearchCornerInd2[i] = minPointInd2;}/******** 計算點i到jl的距離de 理想情況下ijl共線 ********/if (pointSearchCornerInd2[i] >= 0) {tripod1 = laserCloudCornerLast->points[pointSearchCornerInd1[i]];tripod2 = laserCloudCornerLast->points[pointSearchCornerInd2[i]];float x0 = pointSel.x;float y0 = pointSel.y;float z0 = pointSel.z;float x1 = tripod1.x;float y1 = tripod1.y;float z1 = tripod1.z;float x2 = tripod2.x;float y2 = tripod2.y;float z2 = tripod2.z;// 公式分子float a012 = sqrt(((x0 - x1)*(y0 - y2) - (x0 - x2)*(y0 - y1))* ((x0 - x1)*(y0 - y2) - (x0 - x2)*(y0 - y1)) + ((x0 - x1)*(z0 - z2) - (x0 - x2)*(z0 - z1))* ((x0 - x1)*(z0 - z2) - (x0 - x2)*(z0 - z1)) + ((y0 - y1)*(z0 - z2) - (y0 - y2)*(z0 - z1))* ((y0 - y1)*(z0 - z2) - (y0 - y2)*(z0 - z1)));// 公式分母float l12 = sqrt((x1 - x2)*(x1 - x2) + (y1 - y2)*(y1 - y2) + (z1 - z2)*(z1 - z2));// 對x,y,z的偏導float la = ((y1 - y2)*((x0 - x1)*(y0 - y2) - (x0 - x2)*(y0 - y1)) + (z1 - z2)*((x0 - x1)*(z0 - z2) - (x0 - x2)*(z0 - z1))) / a012 / l12;float lb = -((x1 - x2)*((x0 - x1)*(y0 - y2) - (x0 - x2)*(y0 - y1)) - (z1 - z2)*((y0 - y1)*(z0 - z2) - (y0 - y2)*(z0 - z1))) / a012 / l12;float lc = -((x1 - x2)*((x0 - x1)*(z0 - z2) - (x0 - x2)*(z0 - z1)) + (y1 - y2)*((y0 - y1)*(z0 - z2) - (y0 - y2)*(z0 - z1))) / a012 / l12;float ld2 = a012 / l12;pointProj = pointSel;pointProj.x -= la * ld2;pointProj.y -= lb * ld2;pointProj.z -= lc * ld2;float s = 1;if (iterCount >= 5) {s = 1 - 1.8 * fabs(ld2);//增加權重，距離越遠，影響影子越小}coeff.x = s * la;coeff.y = s * lb;coeff.z = s * lc;coeff.intensity = s * ld2;if (s > 0.1 && ld2 != 0) {laserCloudOri->push_back(cornerPointsSharp->points[i]);coeffSel->push_back(coeff);}}}/******** 特征平面點的處理 ********/for (int i = 0; i < surfPointsFlatNum; i++) {/********當前時刻t+1轉換到imu初始坐標系下，對平面點一個一個做處理，設為點i ********/TransformToStart(&surfPointsFlat->points[i], &pointSel);if (iterCount % 5 == 0) {kdtreeSurfLast->nearestKSearch(pointSel, 1, pointSearchInd, pointSearchSqDis);int closestPointInd = -1, minPointInd2 = -1, minPointInd3 = -1;/******** 上一時刻t的LessFlat中的距離點i最近的點j ********/if (pointSearchSqDis[0] < 25) {closestPointInd = pointSearchInd[0];int closestPointScan = int(laserCloudSurfLast->points[closestPointInd].intensity);/******** 在附近4層尋找距離點j最近的兩點l,m ********/float pointSqDis, minPointSqDis2 = 25, minPointSqDis3 = 25;for (int j = closestPointInd + 1; j < surfPointsFlatNum; j++) {if (int(laserCloudSurfLast->points[j].intensity) > closestPointScan + 2.5) {break;}pointSqDis = (laserCloudSurfLast->points[j].x - pointSel.x) * (laserCloudSurfLast->points[j].x - pointSel.x) + (laserCloudSurfLast->points[j].y - pointSel.y) * (laserCloudSurfLast->points[j].y - pointSel.y) + (laserCloudSurfLast->points[j].z - pointSel.z) * (laserCloudSurfLast->points[j].z - pointSel.z);if (int(laserCloudSurfLast->points[j].intensity) <= closestPointScan) {if (pointSqDis < minPointSqDis2) {minPointSqDis2 = pointSqDis;minPointInd2 = j;}} else {if (pointSqDis < minPointSqDis3) {minPointSqDis3 = pointSqDis;minPointInd3 = j;}}}for (int j = closestPointInd - 1; j >= 0; j--) {if (int(laserCloudSurfLast->points[j].intensity) < closestPointScan - 2.5) {break;}pointSqDis = (laserCloudSurfLast->points[j].x - pointSel.x) * (laserCloudSurfLast->points[j].x - pointSel.x) + (laserCloudSurfLast->points[j].y - pointSel.y) * (laserCloudSurfLast->points[j].y - pointSel.y) + (laserCloudSurfLast->points[j].z - pointSel.z) * (laserCloudSurfLast->points[j].z - pointSel.z);if (int(laserCloudSurfLast->points[j].intensity) >= closestPointScan) {if (pointSqDis < minPointSqDis2) {minPointSqDis2 = pointSqDis;minPointInd2 = j;}} else {if (pointSqDis < minPointSqDis3) {minPointSqDis3 = pointSqDis;minPointInd3 = j;}}}}pointSearchSurfInd1[i] = closestPointInd;pointSearchSurfInd2[i] = minPointInd2;pointSearchSurfInd3[i] = minPointInd3;}/*******計算點i到平面jlm的距離dh 理想情況下ijlm共面 *******/if (pointSearchSurfInd2[i] >= 0 && pointSearchSurfInd3[i] >= 0) {tripod1 = laserCloudSurfLast->points[pointSearchSurfInd1[i]];tripod2 = laserCloudSurfLast->points[pointSearchSurfInd2[i]];tripod3 = laserCloudSurfLast->points[pointSearchSurfInd3[i]];// pa,pb,pc既為偏導函數的分子部分也為距離函數分母行列式內各方向分量值，ps為分母部分float pa = (tripod2.y - tripod1.y) * (tripod3.z - tripod1.z) - (tripod3.y - tripod1.y) * (tripod2.z - tripod1.z);float pb = (tripod2.z - tripod1.z) * (tripod3.x - tripod1.x) - (tripod3.z - tripod1.z) * (tripod2.x - tripod1.x);float pc = (tripod2.x - tripod1.x) * (tripod3.y - tripod1.y) - (tripod3.x - tripod1.x) * (tripod2.y - tripod1.y);float pd = -(pa * tripod1.x + pb * tripod1.y + pc * tripod1.z);float ps = sqrt(pa * pa + pb * pb + pc * pc);pa /= ps;pb /= ps;pc /= ps;pd /= ps;//距離沒有取絕對值float pd2 = pa * pointSel.x + pb * pointSel.y + pc * pointSel.z + pd;// 平面偏導公式，似乎后面沒有用到pointProj = pointSel;pointProj.x -= pa * pd2;pointProj.y -= pb * pd2;pointProj.z -= pc * pd2;float s = 1;if (iterCount >= 5) {s = 1 - 1.8 * fabs(pd2) / sqrt(sqrt(pointSel.x * pointSel.x+ pointSel.y * pointSel.y + pointSel.z * pointSel.z));//加上影響因子}coeff.x = s * pa;coeff.y = s * pb;coeff.z = s * pc;coeff.intensity = s * pd2;if (s > 0.1 && pd2 != 0) {laserCloudOri->push_back(surfPointsFlat->points[i]);coeffSel->push_back(coeff);}}}// 如果點云數量小于10，可以不用計算位姿矩陣了int pointSelNum = laserCloudOri->points.size();if (pointSelNum < 10) {continue;}似乎后面沒有用到pointProj = pointSel;pointProj.x -= pa * pd2;pointProj.y -= pb * pd2;pointProj.z -= pc * pd2;float s = 1;if (iterCount >= 5) {s = 1 - 1.8 * fabs(pd2) / sqrt(sqrt(pointSel.x * pointSel.x+ pointSel.y * pointSel.y + pointSel.z * pointSel.z));//加上影響因子}coeff.x = s * pa;coeff.y = s * pb;coeff.z = s * pc;coeff.intensity = s * pd2;if (s > 0.1 && pd2 != 0) {laserCloudOri->push_back(surfPointsFlat->points[i]);coeffSel->push_back(coeff);}}}// 如果點云數量小于10，可以不用計算位姿矩陣了int pointSelNum = laserCloudOri->points.size();if (pointSelNum < 10) {continue;}

之后則是通過構建雅克比矩陣，完成LM求解，從而估計運動，我們仔細閱讀論文了解其原理

接下來看看代碼

cv::Mat matA(pointSelNum, 6, CV_32F, cv::Scalar::all(0));cv::Mat matAt(6, pointSelNum, CV_32F, cv::Scalar::all(0));cv::Mat matAtA(6, 6, CV_32F, cv::Scalar::all(0));cv::Mat matB(pointSelNum, 1, CV_32F, cv::Scalar::all(0));cv::Mat matAtB(6, 1, CV_32F, cv::Scalar::all(0));cv::Mat matX(6, 1, CV_32F, cv::Scalar::all(0));/****** 構建雅可比矩陣，求解 *******/for (int i = 0; i < pointSelNum; i++) {pointOri = laserCloudOri->points[i];coeff = coeffSel->points[i];float s = 1;float srx = sin(s * transform[0]);float crx = cos(s * transform[0]);float sry = sin(s * transform[1]);float cry = cos(s * transform[1]);float srz = sin(s * transform[2]);float crz = cos(s * transform[2]);float tx = s * transform[3];float ty = s * transform[4];float tz = s * transform[5];// J對rx,ry,rz,tx,ty,tz求解偏導，不是很明白怎么得到的代碼 float arx = (-s*crx*sry*srz*pointOri.x + s*crx*crz*sry*pointOri.y + s*srx*sry*pointOri.z + s*tx*crx*sry*srz - s*ty*crx*crz*sry - s*tz*srx*sry) * coeff.x+ (s*srx*srz*pointOri.x - s*crz*srx*pointOri.y + s*crx*pointOri.z+ s*ty*crz*srx - s*tz*crx - s*tx*srx*srz) * coeff.y+ (s*crx*cry*srz*pointOri.x - s*crx*cry*crz*pointOri.y - s*cry*srx*pointOri.z+ s*tz*cry*srx + s*ty*crx*cry*crz - s*tx*crx*cry*srz) * coeff.z;float ary = ((-s*crz*sry - s*cry*srx*srz)*pointOri.x + (s*cry*crz*srx - s*sry*srz)*pointOri.y - s*crx*cry*pointOri.z + tx*(s*crz*sry + s*cry*srx*srz) + ty*(s*sry*srz - s*cry*crz*srx) + s*tz*crx*cry) * coeff.x+ ((s*cry*crz - s*srx*sry*srz)*pointOri.x + (s*cry*srz + s*crz*srx*sry)*pointOri.y - s*crx*sry*pointOri.z+ s*tz*crx*sry - ty*(s*cry*srz + s*crz*srx*sry) - tx*(s*cry*crz - s*srx*sry*srz)) * coeff.z;float arz = ((-s*cry*srz - s*crz*srx*sry)*pointOri.x + (s*cry*crz - s*srx*sry*srz)*pointOri.y+ tx*(s*cry*srz + s*crz*srx*sry) - ty*(s*cry*crz - s*srx*sry*srz)) * coeff.x+ (-s*crx*crz*pointOri.x - s*crx*srz*pointOri.y+ s*ty*crx*srz + s*tx*crx*crz) * coeff.y+ ((s*cry*crz*srx - s*sry*srz)*pointOri.x + (s*crz*sry + s*cry*srx*srz)*pointOri.y+ tx*(s*sry*srz - s*cry*crz*srx) - ty*(s*crz*sry + s*cry*srx*srz)) * coeff.z;float atx = -s*(cry*crz - srx*sry*srz) * coeff.x + s*crx*srz * coeff.y - s*(crz*sry + cry*srx*srz) * coeff.z;float aty = -s*(cry*srz + crz*srx*sry) * coeff.x - s*crx*crz * coeff.y - s*(sry*srz - cry*crz*srx) * coeff.z;float atz = s*crx*sry * coeff.x - s*srx * coeff.y - s*crx*cry * coeff.z;float d2 = coeff.intensity;// A=[J的偏導]; B=[權重系數*(點到直線的距離/點到平面的距離)] 求解公式: AX=B// 為了讓左邊滿秩，同乘At-> At*A*X = At*BmatA.at<float>(i, 0) = arx;matA.at<float>(i, 1) = ary;matA.at<float>(i, 2) = arz;matA.at<float>(i, 3) = atx;matA.at<float>(i, 4) = aty;matA.at<float>(i, 5) = atz;matB.at<float>(i, 0) = -0.05 * d2;}cv::transpose(matA, matAt);matAtA = matAt * matA;matAtB = matAt * matB;cv::solve(matAtA, matAtB, matX, cv::DECOMP_QR);//QR分解得到X不是很明白怎么得到的代碼 float arx = (-s*crx*sry*srz*pointOri.x + s*crx*crz*sry*pointOri.y + s*srx*sry*pointOri.z + s*tx*crx*sry*srz - s*ty*crx*crz*sry - s*tz*srx*sry) * coeff.x+ (s*srx*srz*pointOri.x - s*crz*srx*pointOri.y + s*crx*pointOri.z+ s*ty*crz*srx - s*tz*crx - s*tx*srx*srz) * coeff.y+ (s*crx*cry*srz*pointOri.x - s*crx*cry*crz*pointOri.y - s*cry*srx*pointOri.z+ s*tz*cry*srx + s*ty*crx*cry*crz - s*tx*crx*cry*srz) * coeff.z;float ary = ((-s*crz*sry - s*cry*srx*srz)*pointOri.x + (s*cry*crz*srx - s*sry*srz)*pointOri.y - s*crx*cry*pointOri.z + tx*(s*crz*sry + s*cry*srx*srz) + ty*(s*sry*srz - s*cry*crz*srx) + s*tz*crx*cry) * coeff.x+ ((s*cry*crz - s*srx*sry*srz)*pointOri.x + (s*cry*srz + s*crz*srx*sry)*pointOri.y - s*crx*sry*pointOri.z+ s*tz*crx*sry - ty*(s*cry*srz + s*crz*srx*sry) - tx*(s*cry*crz - s*srx*sry*srz)) * coeff.z;float arz = ((-s*cry*srz - s*crz*srx*sry)*pointOri.x + (s*cry*crz - s*srx*sry*srz)*pointOri.y+ tx*(s*cry*srz + s*crz*srx*sry) - ty*(s*cry*crz - s*srx*sry*srz)) * coeff.x+ (-s*crx*crz*pointOri.x - s*crx*srz*pointOri.y+ s*ty*crx*srz + s*tx*crx*crz) * coeff.y+ ((s*cry*crz*srx - s*sry*srz)*pointOri.x + (s*crz*sry + s*cry*srx*srz)*pointOri.y+ tx*(s*sry*srz - s*cry*crz*srx) - ty*(s*crz*sry + s*cry*srx*srz)) * coeff.z;float atx = -s*(cry*crz - srx*sry*srz) * coeff.x + s*crx*srz * coeff.y - s*(crz*sry + cry*srx*srz) * coeff.z;float aty = -s*(cry*srz + crz*srx*sry) * coeff.x - s*crx*crz * coeff.y - s*(sry*srz - cry*crz*srx) * coeff.z;float atz = s*crx*sry * coeff.x - s*srx * coeff.y - s*crx*cry * coeff.z;float d2 = coeff.intensity;// A=[J的偏導]; B=[權重系數*(點到直線的距離/點到平面的距離)] 求解公式: AX=B// 為了讓左邊滿秩，同乘At-> At*A*X = At*BmatA.at<float>(i, 0) = arx;matA.at<float>(i, 1) = ary;matA.at<float>(i, 2) = arz;matA.at<float>(i, 3) = atx;matA.at<float>(i, 4) = aty;matA.at<float>(i, 5) = atz;matB.at<float>(i, 0) = -0.05 * d2;}cv::transpose(matA, matAt);matAtA = matAt * matA;matAtB = matAt * matB;cv::solve(matAtA, matAtB, matX, cv::DECOMP_QR);//QR分解得到X

接下來有一個迭代第一步的處理，猜測是出現退化進行修正。然后更新位姿之后進行收斂判斷

if (iterCount == 0) {cv::Mat matE(1, 6, CV_32F, cv::Scalar::all(0));cv::Mat matV(6, 6, CV_32F, cv::Scalar::all(0));cv::Mat matV2(6, 6, CV_32F, cv::Scalar::all(0));cv::eigen(matAtA, matE, matV); // 計算矩陣的特征向量E及特征向量的反對稱陣VmatV.copyTo(matV2);isDegenerate = false;float eignThre[6] = {10, 10, 10, 10, 10, 10};for (int i = 5; i >= 0; i--) {if (matE.at<float>(0, i) < eignThre[i]) {for (int j = 0; j < 6; j++) {matV2.at<float>(i, j) = 0;}isDegenerate = true; // 存在比10小的特征值則出現退化} else {break;}}matP = matV.inv() * matV2;}if (isDegenerate) {cv::Mat matX2(6, 1, CV_32F, cv::Scalar::all(0));matX.copyTo(matX2);matX = matP * matX2;}transform[0] += matX.at<float>(0, 0);transform[1] += matX.at<float>(1, 0);transform[2] += matX.at<float>(2, 0);transform[3] += matX.at<float>(3, 0);transform[4] += matX.at<float>(4, 0);transform[5] += matX.at<float>(5, 0);for(int i=0; i<6; i++){if(isnan(transform[i]))transform[i]=0;}float deltaR = sqrt(pow(rad2deg(matX.at<float>(0, 0)), 2) +pow(rad2deg(matX.at<float>(1, 0)), 2) +pow(rad2deg(matX.at<float>(2, 0)), 2));float deltaT = sqrt(pow(matX.at<float>(3, 0) * 100, 2) +pow(matX.at<float>(4, 0) * 100, 2) +pow(matX.at<float>(5, 0) * 100, 2));if (deltaR < 0.1 && deltaT < 0.1) {break;}

最為艱難的計算過程結束了，接下來就是坐標轉換了。。。。。這部分其實也讓我很頭疼，各個坐標系有點暈，仔細理一理

先看代碼

float rx, ry, rz, tx, ty, tz;/******** 旋轉角的累計變化量 ********/// AccumulateRotation作用將局部旋轉坐標轉換至全局旋轉坐標AccumulateRotation(transformSum[0], transformSum[1], transformSum[2], -transform[0], -transform[1] * 1.05, -transform[2], rx, ry, rz);/******** 接著轉移到世界坐標系下 *******/float x1 = cos(rz) * (transform[3] - imuShiftFromStartX) - sin(rz) * (transform[4] - imuShiftFromStartY);float y1 = sin(rz) * (transform[3] - imuShiftFromStartX) + cos(rz) * (transform[4] - imuShiftFromStartY);float z1 = transform[5] * 1.05 - imuShiftFromStartZ;float x2 = x1;float y2 = cos(rx) * y1 - sin(rx) * z1;float z2 = sin(rx) * y1 + cos(rx) * z1;tx = transformSum[3] - (cos(ry) * x2 + sin(ry) * z2);ty = transformSum[4] - y2;tz = transformSum[5] - (-sin(ry) * x2 + cos(ry) * z2);// 插入imu旋轉，更新位姿PluginIMURotation(rx, ry, rz, imuPitchStart, imuYawStart, imuRollStart, imuPitchLast, imuYawLast, imuRollLast, rx, ry, rz);transformSum[0] = rx;transformSum[1] = ry;transformSum[2] = rz;transformSum[3] = tx;transformSum[4] = ty;transformSum[5] = tz;/******** rx,ry,rz轉化為四元數發布 ********/geometry_msgs::Quaternion geoQuat = tf::createQuaternionMsgFromRollPitchYaw(rz, -rx, -ry);laserOdometry.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserOdometry.pose.pose.orientation.x = -geoQuat.y;laserOdometry.pose.pose.orientation.y = -geoQuat.z;laserOdometry.pose.pose.orientation.z = geoQuat.x;laserOdometry.pose.pose.orientation.w = geoQuat.w;laserOdometry.pose.pose.position.x = tx;laserOdometry.pose.pose.position.y = ty;laserOdometry.pose.pose.position.z = tz;pubLaserOdometry.publish(laserOdometry);/********* laserOdometryTrans 是用于tf廣播 ********/laserOdometryTrans.stamp_ = ros::Time().fromSec(timeSurfPointsLessFlat);laserOdometryTrans.setRotation(tf::Quaternion(-geoQuat.y, -geoQuat.z, geoQuat.x, geoQuat.w));laserOdometryTrans.setOrigin(tf::Vector3(tx, ty, tz));tfBroadcaster.sendTransform(laserOdometryTrans);// TransformToEnd的作用是將k+1時刻的less特征點轉移至k+1時刻的sweep的結束位置處的雷達坐標系下int cornerPointsLessSharpNum = cornerPointsLessSharp->points.size();for (int i = 0; i < cornerPointsLessSharpNum; i++) {TransformToEnd(&cornerPointsLessSharp->points[i], &cornerPointsLessSharp->points[i]);}int surfPointsLessFlatNum = surfPointsLessFlat->points.size();for (int i = 0; i < surfPointsLessFlatNum; i++) {TransformToEnd(&surfPointsLessFlat->points[i], &surfPointsLessFlat->points[i]);}/***** 這樣if條件的意義 *******/frameCount++;if (frameCount >= skipFrameNum + 1) {int laserCloudFullResNum = laserCloudFullRes->points.size();for (int i = 0; i < laserCloudFullResNum; i++) {TransformToEnd(&laserCloudFullRes->points[i], &laserCloudFullRes->points[i]);}}/******* t+1時刻數據成為上一個時刻的數據 ************/pcl::PointCloud<PointType>::Ptr laserCloudTemp = cornerPointsLessSharp;cornerPointsLessSharp = laserCloudCornerLast;laserCloudCornerLast = laserCloudTemp;laserCloudTemp = surfPointsLessFlat;surfPointsLessFlat = laserCloudSurfLast;laserCloudSurfLast = laserCloudTemp;laserCloudCornerLastNum = laserCloudCornerLast->points.size();laserCloudSurfLastNum = laserCloudSurfLast->points.size();if (laserCloudCornerLastNum > 10 && laserCloudSurfLastNum > 100) {kdtreeCornerLast->setInputCloud(laserCloudCornerLast);kdtreeSurfLast->setInputCloud(laserCloudSurfLast);}/*———————————— 2HZ發布一次？ ————————————*/if (frameCount >= skipFrameNum + 1) {frameCount = 0;sensor_msgs::PointCloud2 laserCloudCornerLast2;pcl::toROSMsg(*laserCloudCornerLast, laserCloudCornerLast2);laserCloudCornerLast2.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserCloudCornerLast2.header.frame_id = "/camera";pubLaserCloudCornerLast.publish(laserCloudCornerLast2);sensor_msgs::PointCloud2 laserCloudSurfLast2;pcl::toROSMsg(*laserCloudSurfLast, laserCloudSurfLast2);laserCloudSurfLast2.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserCloudSurfLast2.header.frame_id = "/camera";pubLaserCloudSurfLast.publish(laserCloudSurfLast2);sensor_msgs::PointCloud2 laserCloudFullRes3;pcl::toROSMsg(*laserCloudFullRes, laserCloudFullRes3);laserCloudFullRes3.header.stamp = ros::Time().fromSec(timeSurfPointsLessFlat);laserCloudFullRes3.header.frame_id = "/camera";pubLaserCloudFullRes.publish(laserCloudFullRes3);}}status = ros::ok();rate.sleep();}return 0; }

代碼剩余的部分中，回調函數的功能較為簡單，不再贅述。

最后同樣附上框圖方便理解

?

總結

以上是生活随笔為你收集整理的LOAM_velodyne学习（二）的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。