加载MAT文件与地面实况数据。装载，装载“caltech_cordova1_laneAndVehicleGroundTruth.mat”）;

创建一个VideoReader对象来读取视频的帧。videoName =加载。videoName;fileReader = videereader (videoName);地面实况数据被组织在一个时间表中。gtdata = loaded.groundTruthData;显示ground truth数据的前几行。头(gtdata)

阅读视频的第一帧。frame = readFrame(fileReader);提取第一帧的所有车道点。lanePoints = gtdata.LaneBoundaries{1};在第一帧中提取车辆包围框。vehicleBBox = gtdata.Vehicles{1};重叠右侧车道点和车辆包围框。frame = insertMarker(frame, lanePoints{2}，“X”）;帧= insertObjectAnnotation(帧，“矩形”vehicleBBox,“汽车”）;显示第一帧的真实数据。图imshow(框架)

设置monoSensorHelper来处理视频。monocamersensor = loaded.sensor;monoSensorHelper = helperMonoSensor(monocamersensor);为测量创建具有相同时间向量的新时间表。测量=时间表(gtdata.Time);%设置时间表列，用于保存车道边界和车辆数据。numFrames = floor(fileReader.FrameRate*fileReader.Duration);测量。LaneBoundaries = cell(numFrames, 2);测量。车辆检测= cell(numFrames, 1);gtdata。LanesInVehicleCoord = cell(numFrames, 2);将视频倒回到t = 0，并创建一个帧索引来保持当前%框架。fileReader。CurrentTime = 0;frameIndex = 0;遍历视频文件，直到没有新的帧。而hasFrame(fileReader) frameIndex = frameIndex+1;frame = readFrame(fileReader);使用processFrame方法计算检测。此方法可以替换为自定义车道检测方法。detection = processFrame(monoSensorHelper, frame);存储估计的车道边界和车辆检测。测量。LaneBoundaries{frameIndex} = [detection . leftegoboundary ..．.detections.rightEgoBoundary];测量。VehicleDetections{frameIndex} = detects . vehicleboxes;为便于比较，将地面真值车道点转换为%车辆坐标系。gtPointsThisFrame = gtdata.LaneBoundaries{frameIndex};vehiclePoints = cell(1, numel(gtPointsThisFrame));为ii = 1:numel(gtPointsThisFrame) vehiclePoints{ii} = imageToVehicle(monocamersensor, gtPointsThisFrame{ii});结束存储以车辆坐标表示的地面真值点。gtdata。LanesInVehicleCoord{frameIndex} = vehiclePoints;结束

%倒带视频到t = 0。fileReader。CurrentTime = 0;阅读视频的第一帧。frame = readFrame(fileReader);birdsEyeImage = transformImage(monoSensorHelper。BirdsEyeConfig,框架);在鸟瞰视图中提取第一帧的右车道点。firstFrameVehiclePoints = gtdata.LanesInVehicleCoord{1};pointsInBEV = vehicleToImage(monoSensorHelper。BirdsEyeConfig, firstFrameVehiclePoints {2});在框架上叠加点。birdsEyeImage = insertMarker(birdsEyeImage, pointsInBEV，“X”，“大小”6);在鸟瞰视图中显示转换后的点。图imshow (birdsEyeImage)

阈值= 0.25;米%[numMatches, numMisses, numFalsePositives, assignments] =.．.evaluateLaneBoundaries(测量。LaneBoundaries,.．.gtdata。LanesInVehicleCoord,.．.阈值);disp (['匹配数:'num2str (numMatches)]);disp ([“失手次数:”num2str (numMisses)]);disp ([“误报数量:”num2str (numFalsePositives)]);

averageDistance = helpcomputelanestatistics(测量。LaneBoundaries,.．.gtdata。LanesInVehicleCoord,.．.作业,@mean);绘制估计值与实际值之间的平均距离。图阻止(gtdata。时间,averageDistance) title(“估计值与实际数据之间的平均距离”网格)在ylabel (“以米为单位的距离”)传说(的左边界，“右边界”）

使用单摄像机计算估计的点位置。[estVehiclePoints, estImagePoints] = helperGetCorrespondingPoints(monocamersensor，.．.测量。LaneBoundaries,.．.gtdata。LanesInVehicleCoord,.．.作业);将估计车道添加到测量时间表中。。测量。EstimatedLanes = estImagePoints;测量。lanesinvehiclepoints = estVehiclePoints;用可视化所需的所有变量创建一个新的时间表。。名称= {“LanePoints”；“DetectedLanePoints”};类型= labelType({“行”；“行”}）;labelDefs = table(名称，类型，“VariableNames”,{“名字”，“类型”}）;visualizeInFrame =时间表(gtdata. visualizeInFrame =时间表。时间,.．.gtdata。LaneBoundaries,.．.测量。EstimatedLanes,.．.“VariableNames”、名称);创建groundTruth对象。dataSource = groundTruthDataSource(videoName);datatovisualization = groundTruth(dataSource, labelDefs, visualizeInFrame);将上一节的所有结果保存在distanceData中。垫子%临时文件夹。dataToLoad = [tempdir .“distanceData.mat”];保存(dataToLoad,“monoSensorHelper”，“videoName”，“测量”，“gtdata”，“averageDistance”）;

函数stat = helpcomputelanestatistics (estModels, gtPoints, assignments, fcnHandle) numFrames = length(estModels);默认情况下，左右估计NaN表示缺乏%的数据。stat = NaN*ones(numFrames, 2);为frameInd = 1:numFrames默认情况下进行左右估计。stat(frameInd，:) = NaN*ones(2,1);为idx = 1:length(estModels{frameInd})忽略错误的阳性分配。如果分配{frameInd}(idx) == 0继续；结束estModelInFrame中的第k个边界与第k个边界匹配在gtPointsInFrame中由赋值索引的元素。thisModel = estModels{frameInd}(idx);thisGT = gtPoints{frameInd}{assignments{frameInd}(idx)};thisGT = driving.internal.piecewiseLinearBoundary(thisGT);如果mean(thisGTModel.Points(:，2)) > 0%左车道xPoints = thisGTModel.Points(:，1);yDist = 0 (size(xPoints));为索引= 1:数字(xPoints) gtYPoints = thisGTModel.computeBoundaryModel(xPoints(索引));testYPoints = thisModel.computeBoundaryModel(xPoints(index));yDist(index) = abs(testYPoints-gtYPoints);结束stat(frameInd, 1) = fcnHandle(yDist);其他的%右车道xPoints = thisGTModel.Points(:，1);yDist = 0 (size(xPoints));为索引= 1:数字(xPoints) gtYPoints = thisGTModel.computeBoundaryModel(xPoints(索引));testYPoints = thisModel.computeBoundaryModel(xPoints(index));yDist(index) = abs(testYPoints-gtYPoints);结束stat(frameInd, 2) = fcnHandle(yDist);结束结束结束结束

函数[vehiclePoints, imagePoints] = helperGetCorrespondingPoints(monocamersensor, estModels, gtPoints, assignments) numFrames = length(estModels);imagePoints = cell(numFrames, 1);vehiclePoints = cell(numFrames, 1);为frameInd = 1:numFrames如果isempty(assignments{frameInd}) imagePointsInFrame = [];vehiclePointsInFrame = [];其他的estModelInFrame = estModels{frameInd};gtPointsInFrame = gtPoints{frameInd};imagePointsInFrame = cell(length(estModelInFrame)， 1);vehiclePointsInFrame = cell(length(estModelInFrame)， 1);为idx = 1:length(estModelInFrame)忽略错误的阳性分配。如果分配{frameInd}(idx) == 0 imagePointsInFrame{idx} = [NaN NaN];继续；结束estModelInFrame中的第k个边界与第k个边界匹配在gtPointsInFrame中由赋值索引的元素。thisModel = estModelInFrame(idx);thisGT = gtPointsInFrame{assignments{frameInd}(idx)};xPoints = thisGT(:， 1);yPoints = thisModel.computeBoundaryModel(xPoints);vehiclePointsInFrame{idx} = [xPoints, yPoints];imagePointsInFrame{idx} = vehicleToImage(monocamersensor， [xPoints yPoints]);结束结束vehiclePoints{frameInd} = vehiclePointsInFrame;imagePoints{frameInd} = imagePointsInFrame;使用imagePoints[]代替{}来符合groundTruth对象。如果isempty(imagePoints{frameInd}) imagePoints{frameInd} = [];结束如果isempty(vehiclePoints{frameInd}) vehiclePoints{frameInd} = [];结束结束结束