企业官网建设流程全解析

备案网站可以做接码平台么,学校官网网站建设的现状分析,wordpress图片不同分辨率,目前引流最好的平台基于大疆MSDK实现的无人机视觉引导自适应降落功能 概述 最初需求#xff1a;想要无人机在执行完航线任务后#xff0c;一键落到一个指定的位置#xff0c;简化人工控制。 实现一套完整的无人机自主降落功能,通过虚拟摇杆控制使无人机飞向指定位置#xff0c;再利用视觉识别…基于大疆MSDK实现的无人机视觉引导自适应降落功能概述最初需求想要无人机在执行完航线任务后一键落到一个指定的位置简化人工控制。实现一套完整的无人机自主降落功能,通过虚拟摇杆控制使无人机飞向指定位置再利用视觉识别引导无人机精确降落到具体位置。本文中采用自适应降落策略,根据高度动态调整精度要求和下降速度,以实现安全、精确的降落。核心点:虚拟摇杆导航替代FlyTo功能双轴(X/Y)位置偏移实时调整高度自适应降落策略视觉识别引导定位智能避障管理系统架构整体流程graph TD A[用户触发Return to Vehicle] -- B[获取无人机GPS位置] B -- C[计算与目标点距离] C -- D[启动虚拟摇杆导航] D -- E[飞向目标位置 5m/s] E -- F{距离小于10m?} F --|否| E F --|是| G[开始自适应降落] G -- H[视觉识别系统] H -- I[计算X/Y偏移量] I -- J[更新偏移量到ViewModel] J -- K[自适应降落循环] K -- L{高度分段判断} L --|高于50m| M[高空模式] L --|20-50m| N[中空模式] L --|5-20m| O[低空模式] L --|低于5m| P[极低空模式] M -- Q[计算调整速度和下降速度] N -- Q O -- Q P -- Q Q -- R{偏移大于阈值2倍?} R --|是| S[停止下降只调整] R --|否| T[边调整边下降] S -- U{高度小于5m?} T -- U U --|是| V[关闭下视避障] U --|否| K V -- W{高度小于等于0.1m?} W --|否| K W --|是| X[着陆完成清理资源]技术实现思路第一步:让无人机飞到目标位置?问题分析遥控器控制的无人机在执行完航线任务之后飞到给定降落点汽车或其他载具上。最初的想法是使用DJI SDK提供的FlyTo功能,直接指定目标GPS坐标让无人机飞过去。但在实际测试中发现部分机型如M3E并不支持FlyTo功能。机型是否支持FlyTo功能参考文档https://developer.dji.com/doc/mobile-sdk-tutorial/cn/tutorials/intelligent-flight.html解决方案:虚拟摇杆导航既然FlyTo功能不可用,那就用虚拟摇杆功能进行模拟。思路:计算当前位置到目标位置的方位角(bearing)将方位角转换为速度分量(南北/东西)持续发送虚拟摇杆指令,让无人机朝目标飞行实时监测距离,接近目标时停止方位角计算:/* by 01022.hk - online tools website : 01022.hk/zh/browserinfo.html */ private fun calculateBearing(latA: Double, lonA: Double, latB: Double, lonB: Double): Double { val lat1 Math.toRadians(latA) val lat2 Math.toRadians(latB) val dLon Math.toRadians(lonB - lonA) val y Math.sin(dLon) * Math.cos(lat2) val x Math.cos(lat1) * Math.sin(lat2) - Math.sin(lat1) * Math.cos(lat2) * Math.cos(dLon) var bearing Math.toDegrees(Math.atan2(y, x)) bearing (bearing 360) % 360 // 归一化到0-360度 return bearing // 0°正北, 90°正东, 180°正南, 270°正西 }速度分量计算:/* by 01022.hk - online tools website : 01022.hk/zh/browserinfo.html */ val bearing calculateBearing(currentLat, currentLon, targetLat, targetLon) val bearingRad Math.toRadians(bearing) // 使用GROUND坐标系(地面坐标系) val navParam VirtualStickFlightControlParam().apply { rollPitchCoordinateSystem FlightCoordinateSystem.GROUND verticalControlMode VerticalControlMode.POSITION yawControlMode YawControlMode.ANGLE rollPitchControlMode RollPitchControlMode.VELOCITY // 将速度分解为南北和东西分量 pitch NAVIGATION_SPEED * Math.cos(bearingRad) // 南北分量(5m/s) roll NAVIGATION_SPEED * Math.sin(bearingRad) // 东西分量(5m/s) yaw bearing // 让机头指向目标 verticalThrottle targetAlt }GROUND坐标系是绝对方向,不受无人机朝向影响pitch控制南北,roll控制东西。虚拟摇杆参数含义https://developer.dji.com/doc/mobile-sdk-tutorial/cn/basic-introduction/basic-concepts/flight-controller.html#虚拟摇杆第二步:判断何时到达目标点上方附近持续监测距离每100ms检查一次当前位置与目标的距离距离小于预期值ARRIVAL_THRESHOLD,就认为无人机已到达目标点上方附近停止导航开始降落:val navTask object : Runnable { override fun run() { val currentLoc getAircraftLocation() val remainingDistance calculateDistance( currentLoc.latitude, currentLoc.longitude, targetLat, targetLon ) if (remainingDistance ARRIVAL_THRESHOLD) { // 10米内 // 到达目标,停止导航,开始降落 isNavigating false startDynamicAdjustment() } else { // 继续飞行 sendNavigationCommand() virtualStickHandler?.postDelayed(this, 100) } } }第三步:精确降落到指定点无人机虽然到了目标附近(10米内),但有以下问题GPS精度有限(±3米),不够精确。风力影响,有时候受风的影响无人机会偏离。解决方案:视觉识别位置调整工作原理:无人机摄像头识别地面的特定图像(如二维码、标记点)视觉算法计算偏移量(X轴左右,Y轴前后,Z轴距图像距离)将偏移量传给无人机无人机调整位置,边降落边对准数据结构:private var xOffset: Double 0.0 // X轴偏移(米),正右,负左 private var yOffset: Double 0.0 // Y轴偏移(米),正前,负后 private var zDistance: Double 0.0 // Z轴距离(米),距降落点高度外部接口:// 视觉识别系统调用这些方法更新偏移量(~1Hz) fun setXOffset(offset: Double) { xOffset offset } fun setYOffset(offset: Double) { yOffset offset } fun setZDistance(distance: Double) { zDistance distance }采用自适应策略一边降落一遍调整关键点:在不同的高度我们允许的偏移量阈值不同的高度较高的时候偏移量就算比较大也可以下降随着高度降低我们允许的偏移量阈值会不断缩小要求越来越向中间对齐真实偏移超出偏移量阈值的2倍就停止下降只进行对齐调整真实偏移超出偏移量的1倍就以0.1m/s的慢速一边降落一边调整在偏移量范围内且高度 20m以0.5m/s的速度快速下降在偏移量范围内且高度在5m-20m之间以0.2m/s的速度下降在偏移量范围内且高度 5m以0.2m/s速度下降实现:// 1. 根据高度动态计算允许的误差 private fun getOffsetThreshold(altitude: Double): Double { return when { altitude 50.0 - 1.0 // 高空:允许1米偏移误差 altitude 20.0 - 0.5 // 中空:允许0.5米偏移误差 altitude 5.0 - 0.3 // 低空:允许0.3米偏移误差 else - 0.2 // 极低空:要求0.2米精度 } } // 2. 根据高度和偏移量动态计算下降速度 private fun getDescentSpeed(altitude: Double, xOffset: Double, yOffset: Double): Double { val threshold getOffsetThreshold(altitude) return when { xOffset threshold * 2 || yOffset threshold * 2 - 0.0 // 偏移太大:停止下降 xOffset threshold || yOffset threshold - 0.1 // 偏移较大:慢降 altitude 20.0 - 0.5 // 中高空:快降 altitude 5.0 - 0.2 // 低空:慢降 else - 0.2 // 极低空:极慢降 } }控制逻辑:graph TD A[获取当前高度和偏移量] -- B{高度判断} B --|大于50m| C[偏离阈值1m] B --|20-50m| D[偏离阈值0.5m] B --|5-20m| E[偏离阈值0.3m] B --|小于5m| F[偏离阈值0.2m] C -- G{偏移判断} D -- G E -- G F -- G G --|偏移大于阈值的2倍| H[停止下降,只调整] G --|偏移大于阈值| I[慢降0.1m/s并且调整] G --|偏移小于阈值| J[快降并且微调] H -- K[发送虚拟摇杆指令] I -- K J -- K K -- L{高度小于等于0.1m?} L --|否| A L --|是| M[着陆完成]第四步:处理避障降落后停桨。问题:下视避障会阻止降落无人机的下视避障系统会将地面识别为障碍物,在接近地面时自动停止下降我们在高度为5m的时候关闭下视避障落到地面后调用KeyStartAutoLanding进行停桨。参考文档https://sdk-forum.dji.net/hc/zh-cn/articles/14578693771033-如何使用虚拟摇杆降落低空时关闭下视避障var downwardObstacleDisabled false //确保关闭下视避障操作只成功执行一次 // 高度5m时关闭下视避障 if (currentAltitude 5.0 !downwardObstacleDisabled) { downwardObstacleDisabled true setObstacleAvoidanceEnable(false, PerceptionDirection.DOWNWARD) } //关闭下视避障调用方法 private fun setObstacleAvoidanceEnable(enabled: Boolean,direction: PerceptionDirection){ if (direction null) { Log.e(Perception, 方向参数为空无法设置避障) return } PerceptionManager.getInstance().setObstacleAvoidanceEnabled( //调用大疆MSDK方法关闭下视避障 enabled, direction, object : CommonCallbacks.CompletionCallback { override fun onSuccess() { toastResult?.postValue(DJIToastResult.success( 成功设置【${direction.name}】方向的避障为${if (enabled) 开启 else 关闭}) ) Log.i( Perception, 成功设置【${direction.name}】方向的避障为${if (enabled) 开启 else 关闭} ) } override fun onFailure(error: IDJIError) { downwardObstacleDisabled false toastResult?.postValue(DJIToastResult.failed( 设置【${direction.name}】方向的避障失败$error )) Log.e( Perception, 设置【${direction.name}】方向的避障失败$error ) } } ) }第五步:降落循环完整逻辑private fun startDynamicAdjustment() { isAdjusting true virtualStickHandler Handler(Looper.getMainLooper()) val adjustTask object : Runnable { override fun run() { if (!isAdjusting) return // 1. 获取当前状态 val currentAltitude FlightControllerKey.KeyAltitude.create().get(0.0) val currentXOffsetAbs Math.abs(xOffset) val currentYOffsetAbs Math.abs(yOffset) // 2. 检查是否着陆 if (currentAltitude 0.1) { stopLanding() return } // 3. 低空时关闭下视避障 if (currentAltitude 5.0 !downwardObstacleDisabled) { downwardObstacleDisabled true setObstacleAvoidanceEnable(false, PerceptionDirection.DOWNWARD) } // 4. 计算自适应参数 val offsetThreshold getOffsetThreshold(currentAltitude) val descentSpeed getDescentSpeed(currentAltitude, currentXOffsetAbs, currentYOffsetAbs) // 5. 构建虚拟摇杆指令 val adjustParam VirtualStickFlightControlParam().apply { rollPitchCoordinateSystem FlightCoordinateSystem.BODY verticalControlMode VerticalControlMode.VELOCITY rollPitchControlMode RollPitchControlMode.VELOCITY // 水平调整 roll if (currentXOffsetAbs offsetThreshold) { if (xOffset 0) ADJUSTMENT_SPEED else -ADJUSTMENT_SPEED } else 0.0 pitch if (currentYOffsetAbs offsetThreshold) { if (yOffset 0) ADJUSTMENT_SPEED else -ADJUSTMENT_SPEED } else 0.0 // 垂直下降 verticalThrottle -descentSpeed } // 6. 发送指令 VirtualStickManager.getInstance().sendVirtualStickAdvancedParam(adjustParam) // 7. 100ms后再次执行(10Hz) virtualStickHandler?.postDelayed(this, 100) } } virtualStickHandler?.post(adjustTask) }以上就实现了一整套视觉引导的自适应降落方案安全注意事项WARNING必须在空旷、安全环境测试建议先用DJI模拟器测试视觉识别必须持续更新(~1Hz)准备好随时手动接管代码/** * One-key return to vehicle function (using Virtual Stick instead of FlyTo) * 1. Get aircraft current location * 2. Calculate distance to vehicle using Haversine formula * 3. If distance 500m, reject with error * 4. Use Virtual Stick to navigate to vehicle location * 5. Switch to precision adjustment when close enough */ fun returnToVehicle(callback: CommonCallbacks.CompletionCallback) { // Get aircraft current location val aircraftLocation getAircraftLocation() if (aircraftLocation null || !isLocationValid(aircraftLocation.latitude, aircraftLocation.longitude)) { callback.onFailure(DJICommonError.FACTORY.build(无法获取无人机位置信息)) return } // Vehicle coordinates (hardcoded for now, will be replaced with API later) // TODO: Replace with actual vehicle GPS coordinates from API val vehicleLatitude 22.579 // Example coordinates val vehicleLongitude 113.941 // Example coordinates // Calculate distance using Haversine formula val distance calculateDistance( aircraftLocation.latitude, aircraftLocation.longitude, vehicleLatitude, vehicleLongitude ) // Distance validation: reject if 500m if (distance 500) { callback.onFailure(DJICommonError.FACTORY.build( 距离过远: ${String.format(%.2f, distance)}m, 超出 500m 限制 )) return } // Start virtual stick navigation to vehicle location toastResult?.postValue(DJIToastResult.success(开始飞向车辆位置)) //TODO 这个targetAlt需要后期经过计算算出来。 navigateToTarget(vehicleLatitude, vehicleLongitude, 100.0, callback) } /** * Navigate to target location using Virtual Stick */ private fun navigateToTarget( targetLat: Double, targetLon: Double, targetAlt: Double, callback: CommonCallbacks.CompletionCallback ) { VirtualStickManager.getInstance().enableVirtualStick(object : CommonCallbacks.CompletionCallback { override fun onSuccess() { VirtualStickManager.getInstance().setVirtualStickAdvancedModeEnabled(true) isNavigating true startNavigation(targetLat, targetLon, targetAlt, callback) } override fun onFailure(error: IDJIError) { callback.onFailure(error) } }) } /** * Start navigation loop using Virtual Stick */ private fun startNavigation( targetLat: Double, targetLon: Double, targetAlt: Double, callback: CommonCallbacks.CompletionCallback ) { virtualStickHandler Handler(Looper.getMainLooper()) val navTask object : Runnable { override fun run() { if (!isNavigating) { return } val currentLoc getAircraftLocation() if (currentLoc null) { virtualStickHandler?.postDelayed(this, 100) return } // Calculate remaining distance val remainingDistance calculateDistance( currentLoc.latitude, currentLoc.longitude, targetLat, targetLon ) println(targetLat:targetLat targetLon:targetLon currentLoc.latitude:currentLoc.latitude currentLoc.longitude:currentLoc.longitude remainingDistance:remainingDistance) // Check if arrived if (remainingDistance ARRIVAL_THRESHOLD) { // Arrived at target, stop navigation isNavigating false virtualStickHandler?.removeCallbacksAndMessages(null) callback.onSuccess() toastResult?.postValue(DJIToastResult.success(已到达车辆位置,开始精确定位)) //开始调节云台角度,俯仰角为-90°旋转时间1s startGimbalAngleRotation(GimbalAngleRotationMode.ABSOLUTE_ANGLE,-90.0,0.0,0.0,1.0) // Start precision adjustment startDynamicAdjustment() } else { // Continue navigation val bearing calculateBearing( currentLoc.latitude, currentLoc.longitude, targetLat, targetLon ) val navParam VirtualStickFlightControlParam().apply { rollPitchCoordinateSystem FlightCoordinateSystem.GROUND // Use ground coordinate system verticalControlMode VerticalControlMode.POSITION yawControlMode YawControlMode.ANGLE rollPitchControlMode RollPitchControlMode.VELOCITY // Calculate velocity components based on bearing val bearingRad Math.toRadians(bearing) pitch NAVIGATION_SPEED * Math.sin(bearingRad) // North-South component roll NAVIGATION_SPEED * Math.cos(bearingRad) // East-West component yaw bearing // Point towards target verticalThrottle targetAlt // Target altitude } VirtualStickManager.getInstance().sendVirtualStickAdvancedParam(navParam) virtualStickHandler?.postDelayed(this, 100) } } } virtualStickHandler?.post(navTask) } fun startGimbalAngleRotation(mode: GimbalAngleRotationMode,pitch: Double,yaw: Double,roll: Double,duration: Double){ val rotation GimbalAngleRotation().apply { setMode(mode) setPitch(pitch) setYaw(yaw) setRoll(roll) setDuration(duration) } KeyManager.getInstance().performAction( KeyTools.createKey(GimbalKey.KeyRotateByAngle), rotation, object: CommonCallbacks.CompletionCallbackWithParamEmptyMsg{ override fun onSuccess(result: EmptyMsg?) { toastResult?.postValue(DJIToastResult.success(云台旋转成功)) Log.i(Gimbal,云台旋转成功yaw:${rotation.yaw},pitch:${rotation.pitch},roll:${rotation.roll}) } override fun onFailure(error: IDJIError) { toastResult?.postValue(DJIToastResult.failed(云台旋转失败,$error)) Log.e(Gimbal,云台旋转失败$error) } } ) } /** * Start dynamic position adjustment loop with adaptive descent * Adjusts position while descending, with stricter requirements at lower altitudes */ private fun startDynamicAdjustment() { isAdjusting true virtualStickHandler Handler(Looper.getMainLooper()) // Send adjustment commands at 10Hz val adjustTask object : Runnable { override fun run() { if (!isAdjusting) { return } // TODO 获取脚本检测出的z轴距离 val currentAltitude FlightControllerKey.KeyAltitude.create().get(0.0) val currentXOffsetAbs Math.abs(xOffset) val currentYOffsetAbs Math.abs(yOffset) //关闭降落保护下视避障失效 if(currentAltitude 5 !downwardObstacleDisabled){ downwardObstacleDisabled true setObstacleAvoidanceEnable(false, PerceptionDirection.DOWNWARD) } // 检查是否落地 if (currentAltitude 0.1) { stopLanding() return } // Get adaptive thresholds based on altitude val offsetThreshold getOffsetThreshold(currentAltitude) val descentSpeed getDescentSpeed(currentAltitude, currentXOffsetAbs,currentYOffsetAbs) // Log for debugging println(自动调整 - 高度:%.2fm, x偏移:%.2fm,y偏移:%.2fm, 阈值:%.2fm, 下降速度:%.2fm/s.format( currentAltitude, currentXOffsetAbs,currentYOffsetAbs,offsetThreshold, descentSpeed )) // Calculate adjustment parameters val adjustParam VirtualStickFlightControlParam().apply { rollPitchCoordinateSystem FlightCoordinateSystem.BODY verticalControlMode VerticalControlMode.VELOCITY yawControlMode YawControlMode.ANGULAR_VELOCITY rollPitchControlMode RollPitchControlMode.ANGLE // Calculate roll value based on offset // Positive offset (need to move forward) - positive roll // Negative offset (need to move backward) - negative roll if (currentXOffsetAbs offsetThreshold) { // Need adjustment roll if (xOffset 0) ADJUSTMENT_SPEED else -ADJUSTMENT_SPEED } else { // Within threshold, no adjustment needed roll 0.0 } if (currentYOffsetAbs offsetThreshold) { pitch if(yOffset 0) ADJUSTMENT_SPEED else -ADJUSTMENT_SPEED } else { pitch 0.0 } yaw 0.0 verticalThrottle -descentSpeed // Descend at adaptive speed } VirtualStickManager.getInstance().sendVirtualStickAdvancedParam(adjustParam) virtualStickHandler?.postDelayed(this, 100) } } virtualStickHandler?.post(adjustTask) toastResult?.postValue(DJIToastResult.success(开始动态位置调整)) } /** * Stop landing and cleanup */ private fun stopLanding() { virtualStickHandler?.removeCallbacksAndMessages(null) //调用KeyStartAutoLanding进行停桨 FlightControllerKey.KeyStartAutoLanding.create().action({ toastResult?.postValue(DJIToastResult.success(桨叶动力关闭)) Log.i(stopLanding,桨叶动力关闭成功) },{ toastResult?.postValue(DJIToastResult.failed(桨叶动力关闭失败)) Log.i(stopLanding,桨叶动力关闭失败) }) cleanupVirtualStick() toastResult?.postValue(DJIToastResult.success(降落完成)) } /** * Get offset threshold based on current altitude * Higher altitude allows larger offset, lower altitude requires stricter precision */ private fun getOffsetThreshold(altitude: Double): Double { return when { altitude HIGH_ALTITUDE - 1.0 // High altitude: allow 1m offset altitude MID_ALTITUDE - 0.5 // Mid altitude: allow 0.5m offset altitude LOW_ALTITUDE - 0.4 // Low altitude: allow 0.3m offset else - 0.3 // Very low altitude: require 0.2m precision } } /** * Get descent speed based on current altitude and offset * Larger offset or lower altitude results in slower descent */ private fun getDescentSpeed(altitude: Double, xOffset: Double,yOffset: Double): Double { val threshold getOffsetThreshold(altitude) return when { xOffset threshold * 2 || yOffset threshold * 2 - 0.0 // Offset too large: stop descending xOffset threshold || yOffset threshold - 0.1 // Offset large: slow descent altitude MID_ALTITUDE - 0.5 // Mid-high altitude: fast descent altitude LOW_ALTITUDE - 0.2 // Low altitude: slow descent else - 0.1 // Very low altitude: very slow descent } }