doTraining = false;outputFolder = fullfile (tempdir,“Pandaset”)；lidarURL=[“https://ssd.mathworks.com/万博1manbetxsupportfiles/lidar/data/”．..“Pandaset_LidarData.tar.gz”];helperDownloadPandasetData (outputFolder lidarURL);

路径= fullfile (outputFolder,“激光雷达”）;lidarData = fileDatastore(路径,“ReadFcn”@ (x) pcread (x));

gtPath = fullfile (outputFolder,“长方体”，“PandaSetLidarGroundTruth.mat”）;data =负载(gtPath“lidarGtLabels”）;标签= timetable2table (data.lidarGtLabels);boxLabels =标签(:2:3);

图ptld = read(lidarData);ax = pcshow (ptCld.Location);集(ax,“XLim”50 [-50],“YLim”, 40 [-40]);变焦(ax, 2.5);轴从；

重置(lidarData);

xMin = 0.0;% x轴上最小值。yMin = -39.68;% y轴最小值。zMin = -5.0;% z轴最小值。xMax = 69.12;% x轴上的最大值。yMax = 39.68;% y轴上的最大值。zMax=5.0；% z轴上最大值。xStep = 0.16;%沿X轴的分辨率。yStep = 0.16;%沿Y轴的分辨率。dsFactor = 2.0;%将采样因素。%计算伪图像的尺寸。Xn = round(((xMax - xMin) / xStep));Yn = round(((yMax - yMin) / yStep));定义点云参数。yMin pointCloudRange = [xMin xMax, yMax, zMin, zMax);voxelSize = [xStep, yStep];

[croppedPointCloudObj, processedLabels] = cropFrontViewFromLidarData (．..lidarData、boxLabels pointCloudRange);

电脑= croppedPointCloudObj {1};gtLabelsCar = processedLabels.Car {1};gtLabelsTruck = processedLabels.Truck {1};helperDisplay3DBoxesOverlaidPointCloud (pc。位置、gtLabelsCar．..“绿色”gtLabelsTruck,“洋红”，“裁剪点云”）;

重置(lidarData);

rng (1);shuffledIndices = randperm(大小(processedLabels, 1));idx = floor(0.7 * length(shuffledIndices));trainData = croppedPointCloudObj (shuffledIndices (1: idx):);testData = croppedPointCloudObj (shuffledIndices (idx + 1:结束):);trainLabels = processedLabels (shuffledIndices (1: idx):);testLabels = processedLabels (shuffledIndices (idx + 1:结束):);

writefile = true;dataLocation = fullfile (outputFolder,“InputData”）;[trainData, trainLabels] = saveptCldToPCD (trainData trainLabels,．..dataLocation, writefile);

像= fileDatastore (dataLocation,“ReadFcn”@ (x) pcread (x));

bds = boxLabelDatastore (trainLabels);

cds=联合收割机（lds、bds）；

augData =阅读(cds);augptCld = augData {1};augLabels = augData {1,2};augClass = augData {1,3};labelsCar = augLabels (augClass = =“汽车”:);labelsTruck = augLabels (augClass = =“卡车”:);helperDisplay3DBoxesOverlaidPointCloud (augptCld。位置、labelsCar“绿色”，．..labelsTruck,“洋红”，“之前数据增强”）;

重置(cds);

一会= {“汽车”，“卡车”};sampleLocation = fullfile (tempdir,“GTsamples”）;[sampledGTData,指数]= sampleGroundTruthObjectsFromLidarData (cd,一会,．..“MinPoints”, 20岁,“sampleLocation”，样本位置）；

numObjects = (10,10);addsaugmented = transform(cd，@(x) augmentGroundTruthObjectsToLidarData(x，．..sammagnitdata、索引、类名、numObjects）；

cdsAugmented=变换（cdsAugmented，@（x）扩充数据（x））；

augData =阅读(cdsAugmented);augptCld = augData {1};augLabels = augData {1,2};augClass = augData {1,3};labelsCar = augLabels (augClass = =“汽车”:);labelsTruck = augLabels (augClass = =“卡车”:);helperDisplay3DBoxesOverlaidPointCloud (augptCld。位置、labelsCar“绿色”，．..labelsTruck,“洋红”，数据增加的）;

重置(cdsAugmented);

定义突出柱子的数量。P=12000；定义每根柱子的点数。N = 100;从训练数据中估计锚盒。anchorBoxes = calculateAnchorsPointPillars (trainLabels);一会= trainLabels.Properties.VariableNames;定义pointcolumns检测器。探测器= pointPillarsObjectDetector (anchorBoxes pointCloudRange,类名,．..“VoxelSize”voxelSize,“NumPillars”、磷、“NumPointsPerPillar”N);

executionEnvironment =“汽车”；如果canUseParallelPool dispatchInBackground = true;其他的dispatchInBackground = false;结束选择= trainingOptions (“亚当”，．..“阴谋”，“没有”，．..“MaxEpochs”现年60岁的．..“MiniBatchSize”3.．..“GradientDecayFactor”, 0.9,．..“SquaredGradientDecayFactor”, 0.999,．..“LearnRateSchedule”，“分段”，．..“InitialLearnRate”, 0.0002,．..“LearnRateDropPeriod”15岁的．..“LearnRateDropFactor”, 0.8,．..“ExecutionEnvironment”executionEnvironment,．..“DispatchInBackground”dispatchInBackground,．..“BatchNormalizationStatistics”，“移动”，．..“ResetInputNormalization”假的,．..“CheckpointPath”, tempdir);

如果doTraining[检测器，info] = trainPointPillarsObjectDetector(cdsAugmented，检测器，options);其他的pretrainedDetector =负载(“pretrainedPointPillarsDetector.mat”，“探测器”）;探测器= pretrainedDetector.detector;结束

45 ptCloud = testData {1};gtLabels = testLabels(45岁);%指定仅使用的检测的置信阈值%置信度得分高于此值。confidenceThreshold = 0.5;(盒、分数、标签)=检测(ptCloud探测器,“阈值”, confidenceThreshold);boxlabelsCar(标签' = = =框“汽车”，：；boxlabelsTruck=box（标签'==“卡车”:);%在点云上显示预测。helperDisplay3DBoxesOverlaidPointCloud (ptCloud。位置、boxlabelsCar“绿色”，．..boxlabelsTruck,“洋红”，“预测边界框”）;

numInputs = 50;%从长方体标签生成旋转的矩形。bds = boxLabelDatastore (testLabels (1: numInputs,:));groundTruthData = transform(bds，@(x) createRotRect(x));%设置阈值。nmsPositiveIoUThreshold = 0.5;confidenceThreshold = 0.25;detectionResults =检测(探测器,testData (1: numInputs,:)．..“阈值”, confidenceThreshold);%转换为旋转矩形格式计算指标为i = 1:height(detectionResults) box = detectionResults. boxes {i};detectionResults。盒子我{}=盒(:,(1、2、4、5、7]);结束指标= evaluateDetectionAOS (detectionResults groundTruthData,．..nmsPositiveIoUThreshold);disp(指标(:,1:2))

函数helperDownloadPandasetData (outputFolder lidarURL)%从给定URL下载数据集到输出文件夹。LIDARDatarFile=fullfile（输出文件夹，“Pandaset_LidarData.tar.gz”）;如果~存在(lidarDataTarFile“文件”mkdir (outputFolder);disp (“正在下载PandaSet激光雷达驱动数据(5.2 GB)……”）;websave (lidarDataTarFile lidarURL);解压(lidarDataTarFile outputFolder);结束%解压缩文件。如果（~exist（fullfile）（outputFolder，“激光雷达”)，“dir”））．..& & (~ (fullfile (outputFolder,存在“长方体”)，“dir”))untar（LIDARDatar文件，输出文件夹）；结束结束函数helperDisplay3DBoxesOverlaidPointCloud (ptCld labelsCar carColor,．..labelsTruck、truckColor titleForFigure)%显示点云与不同颜色的包围盒%的类。图;ax = pcshow (ptCld);showShape (“长方体”labelsCar,“父”ax,“不透明度”, 0.1,．..“颜色”carColor,“线宽”, 0.5);持有在；展示(“长方体”labelsTruck,“父”ax,“不透明度”, 0.1,．..“颜色”truckColor,“线宽”, 0.5);标题(titleForFigure);变焦(ax, 1.5);结束