【科普】无人机自动避障技术盘点
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">近年来,随着多旋翼消费级无人机市场的飞速增长,其<span style="color: black;">关联</span>技术<span style="color: black;">亦</span>正在<span style="color: black;">出现</span>日新月异的变革,以往多用于特种行业<span style="color: black;">乃至</span>军用<span style="color: black;">制品</span>上的技术(如飞行<span style="color: black;">掌控</span>、图像传输、<span style="color: black;">目的</span>识别和跟踪等)<span style="color: black;">亦</span>越来越多地用于消费级无人机,使其越来越接近于自动化<span style="color: black;">乃至</span>智能化飞行<span style="color: black;">设备</span>人的概念。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cNxzAr3lOmW~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=pP%2F4gMjQzHf15YxN47jJUTCibLQ%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">在消费级无人机技术的<span style="color: black;">许多</span>趋势中,避障能力是实现自动化乃至智能化的关键环节,完善的自主避障系统将能够在很大程度上减少因操作失误<span style="color: black;">导致</span>的无人机损坏和伤及人身和建筑物的事故<span style="color: black;">出现</span>率,而从各消费级无人机厂商的新品和技术发展方向看,避障技术<span style="color: black;">亦</span>将在<span style="color: black;">将来</span>几年趋于完善并<span style="color: black;">作为</span>中高端消费级无人机的标配系统。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">避障技术,顾名思义<span style="color: black;">便是</span>无人机自主躲避<span style="color: black;">阻碍</span>物的智能技术。<span style="color: black;">非常多</span>玩过无人机的小伙伴们都<span style="color: black;">晓得</span>,有避障功能的无人机和<span style="color: black;">无</span>避障功能的无人机,<span style="color: black;">能够</span>说体验是大不相同的!无人机自动避障系统能够<span style="color: black;">即时</span>地避开飞行路径中的<span style="color: black;">阻碍</span>物,<span style="color: black;">极重</span>地减少<span style="color: black;">由于</span>操作失误而带来的各项损失。在减少炸机事故次数的<span style="color: black;">同期</span>,还能给无人机新手<span style="color: black;">极重</span>的<span style="color: black;">帮忙</span>!</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cNyEHMdn3LQ~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=%2BG3%2FVZmHxO%2BpgfpGMOBZdnmoulU%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">无人机避障技术的发展<span style="color: black;">周期</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">按照</span><span style="color: black;">日前</span>无人机避障技术的发展以及其<span style="color: black;">将来</span>的<span style="color: black;">科研</span>态势,无人机避障技术可分为三个<span style="color: black;">周期</span>,一是感知<span style="color: black;">阻碍</span>物<span style="color: black;">周期</span>;二是绕过<span style="color: black;">阻碍</span>物<span style="color: black;">周期</span>;三是场景建模 和路径搜索<span style="color: black;">周期</span>。这三个<span style="color: black;">周期</span>其实是无人机避障技术的<span style="color: black;">功效</span>过程。从无人机<span style="color: black;">发掘</span><span style="color: black;">阻碍</span>物,到<span style="color: black;">能够</span>自动绕开<span style="color: black;">阻碍</span>物,再达到自我规划路径的过程。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cNyR38NpJ4c~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=r2TQ6%2BlDV7N6N2orWctMlYYw3m0%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">第1</span><span style="color: black;">周期</span>,无人机只能是简单地感知<span style="color: black;">阻碍</span>物。当无人机遇到<span style="color: black;">阻碍</span>物时,能快速地识别,并且悬停下来,等待无人机驾驶者的下一步指令!</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">第二<span style="color: black;">周期</span>,无人机能够获取<span style="color: black;">阻碍</span>物的深度图象,并由此精确感知<span style="color: black;">阻碍</span>物的<span style="color: black;">详细</span>轮廓,<span style="color: black;">而后</span>自主绕开<span style="color: black;">阻碍</span>物!这个<span style="color: black;">周期</span>是摆脱飞手操作,实现无人机自主驾驶的<span style="color: black;">周期</span>!</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">第三<span style="color: black;">周期</span>,无人机能够对飞行区域<span style="color: black;">创立</span>地图模型<span style="color: black;">而后</span>规划<span style="color: black;">恰当</span>线路!这个地图<span style="color: black;">不可</span>仅仅是机械平面模型,而应该是一个能够实时更新的三维立体地图!这将是<span style="color: black;">日前</span>无人机避障技术的最高<span style="color: black;">周期</span>!</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">感知<span style="color: black;">阻碍</span>物</strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">在前避障时代,消费级无人机的<span style="color: black;">运用</span>说明上都会标明必须在开阔场地飞行,<span style="color: black;">况且</span>应当<span style="color: black;">尽可能</span>避免<span style="color: black;">周边</span>有<span style="color: black;">海量</span>人群(当然,这<span style="color: black;">亦</span>与当前技术和市场环境使得消费级无人机<span style="color: black;">制品</span>的<span style="color: black;">靠谱</span>性较差有很大关系),<span style="color: black;">由于</span>一不小心操作失误,<span style="color: black;">或</span>在<span style="color: black;">周边</span>有<span style="color: black;">高挑</span><span style="color: black;">阻碍</span>物时<span style="color: black;">起步</span>一键(低电压、<span style="color: black;">失去控制</span>)返航,则有可能眼睁睁看着无人机傻傻地撞向<span style="color: black;">那样</span><span style="color: black;">显著</span>的<span style="color: black;">阻碍</span>物,这是何等的回天乏力。为了降低这种事故的<span style="color: black;">出现</span>率,各厂商<span style="color: black;">亦</span>都在卯足劲<span style="color: black;">开发</span>避障技术,而在实现方式上,<span style="color: black;">大众</span>的着眼点<span style="color: black;">亦</span>都放在了一个方向——<span style="color: black;">测绘</span>无人机到<span style="color: black;">阻碍</span>物的距离。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cNydIaUA1Dq~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=fQOwx3mMh0zGKq%2BqzXnOENOHhqs%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">咱们</span><span style="color: black;">能够</span>很容易地想到,只要无人机能够<span style="color: black;">测绘</span>出与潜在<span style="color: black;">阻碍</span>物之间的距离,就<span style="color: black;">能够</span>在撞向<span style="color: black;">阻碍</span>物之前停止前进(虽然固定翼无人机<span style="color: black;">暗示</span><span style="color: black;">区别</span>意),于是一场让人感觉回天乏力的事故被轻描淡写地避免了,这种思路简单粗暴,但还是有<span style="color: black;">必定</span><span style="color: black;">功效</span>的。而当前运用较多的<span style="color: black;">阻碍</span>物检测<span style="color: black;">办法</span><span style="color: black;">重点</span>有:</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">超声波测距:这个<span style="color: black;">办法</span><span style="color: black;">非常多</span>人都<span style="color: black;">熟练</span>,家用汽车的倒车雷达<span style="color: black;">便是</span>利用超声波检测<span style="color: black;">阻碍</span>物,该<span style="color: black;">办法</span>的优点<span style="color: black;">便是</span>技术成熟,成本很低;但缺点在于<span style="color: black;">功效</span>距离近(常用的中低端超声波传感器<span style="color: black;">功效</span>距离不超过10m),且对反射面有<span style="color: black;">必定</span><span style="color: black;">需求</span>。<span style="color: black;">因此呢</span>超声波测距传感器常用来<span style="color: black;">测绘</span>无人机与地面的距离(固定翼无人机<span style="color: black;">暗示</span>自己飞的太高太快,超声波传感器用不上)。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">红外/激光TOF:即飞行时间传感器,基本原理<span style="color: black;">便是</span>传感器发射<span style="color: black;">必定</span>频率的红外/激光信号,<span style="color: black;">而后</span><span style="color: black;">按照</span>反射信号与原信号的相位差计算信号的飞行时间,<span style="color: black;">就可</span>换算出距离<span style="color: black;">阻碍</span>物的距离。该<span style="color: black;">办法</span>技术比较成熟,<span style="color: black;">功效</span>距离较超声波更远(数米到数百米),<span style="color: black;">况且</span>高等级的TOF传感器<span style="color: black;">能够</span><span style="color: black;">得到</span><span style="color: black;">阻碍</span>物的深度图像(这一项能力在下文会有应用说明),但缺点在于成本高,抗干扰能力较差(激光TOF稍好)。<span style="color: black;">因此呢</span>该<span style="color: black;">方法</span>在当前市场上<span style="color: black;">制品</span>或样机中有<span style="color: black;">必定</span>规模的应用。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">双目视觉:这个<span style="color: black;">办法</span>运用了人眼估计距离的原理,即同一个物体在两个镜头画面中的坐标稍有<span style="color: black;">区别</span>,经过转换<span style="color: black;">就可</span>得到<span style="color: black;">阻碍</span>物的距离,双目视觉<span style="color: black;">办法</span><span style="color: black;">亦</span><span style="color: black;">能够</span><span style="color: black;">得到</span><span style="color: black;">阻碍</span>物的深度图像。这种<span style="color: black;">办法</span>的缺点在于技术难度较高(<span style="color: black;">不外</span>自从有了OpenCV,妈妈再<span style="color: black;">亦</span><span style="color: black;">不消</span>担心我不会写<span style="color: black;">设备</span>视觉程序了),且距离估计的误差随距离变大而指数型增长,只是这一缺点在无人机避障应用中并无大碍。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">电子地图:借助细粒度的数字高程地图和城市3D建筑地图,既能够实现避免重要建筑物受到无人机撞击(即禁飞区功能),<span style="color: black;">亦</span>能够实现<span style="color: black;">非常多</span><span style="color: black;">状况</span>下的无人机避障。而事实上,战斧巡航导弹的远程飞行<span style="color: black;">亦</span>在很大程度上依赖于数字高程地图。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">阻碍</span>物<span style="color: black;">测绘</span>的原理搞懂了,就<span style="color: black;">能够</span><span style="color: black;">起始</span>讨论无人机避障了。最简单的策略莫过于遇到<span style="color: black;">阻碍</span>物时停止前进,<span style="color: black;">而后</span>与<span style="color: black;">阻碍</span>物保持<span style="color: black;">必定</span>的距离。这种遇到<span style="color: black;">阻碍</span>物后就默默悬停等待,仿佛手足无措不知所往的初级策略,<span style="color: black;">便是</span>这么简单的<span style="color: black;">起始</span>,无人机就进入了避障时代。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">绕过<span style="color: black;">阻碍</span>物</strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">很显然,<span style="color: black;">咱们</span>不会满足于让无人机遇到<span style="color: black;">阻碍</span>物后傻傻等待(固定翼<span style="color: black;">暗示</span>自己一秒钟<span style="color: black;">亦</span><span style="color: black;">没法</span>等待),这就需要设计让无人机安全<span style="color: black;">有效</span>地绕过<span style="color: black;">阻碍</span>物继续完成预定飞行的策略。表面上看,连萨摩耶这种囧傻呆萌的狗狗都<span style="color: black;">晓得</span>前面有座大楼时该怎么绕过去(请<span style="color: black;">自动</span>想象<span style="color: black;">为何</span>说到狗狗时要强调前方是大楼,而不是一棵树),让无人机绕过<span style="color: black;">阻碍</span>物的策略应该很简单了,但这其中的门道可多了去了。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">首要</span>,狗狗很清楚前方大楼的轮廓,<span style="color: black;">因此呢</span>只需要往边缘走就<span style="color: black;">能够</span>绕过去,<span style="color: black;">然则</span>无人机想<span style="color: black;">得到</span><span style="color: black;">阻碍</span>物轮廓就很难了,<span style="color: black;">倘若</span>避障传感器是普通超声波,无人机就只能<span style="color: black;">晓得</span>前方有<span style="color: black;">阻碍</span>,<span style="color: black;">然则</span>却无从<span style="color: black;">晓得</span><span style="color: black;">阻碍</span>物的边缘,这<span style="color: black;">便是</span>前面<span style="color: black;">为何</span>要强调“能够<span style="color: black;">得到</span><span style="color: black;">阻碍</span>物深度图像”了,<span style="color: black;">针对</span>能<span style="color: black;">得到</span>深度图像的TOF测距和双目视觉测距方式,只要<span style="color: black;">阻碍</span>物<span style="color: black;">无</span>充满视场,就总能够找到<span style="color: black;">阻碍</span>物的边缘。举例说明,下图所示的是无人机的到的深度图像示意图,灰度越深,<span style="color: black;">显示</span>距离越近,遇到这种<span style="color: black;">状况</span>,很显然的策略<span style="color: black;">便是</span>往左上方飞,即朝向灰度最浅的区域飞行,此时问题仿佛<span style="color: black;">已然</span>得到<span style="color: black;">处理</span>。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cNys8LMIDPD~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=Cto2soTCIUXdsP1%2FO42vyn4xCHg%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">然则</span>别高兴太早,这种策略<span style="color: black;">能够</span>满足大<span style="color: black;">都数</span>应用场景,<span style="color: black;">然则</span>问题还远<span style="color: black;">无</span><span style="color: black;">处理</span>,看下面的左图,一架无人机刚绕从一座<span style="color: black;">高挑</span>建筑旁绕过去,<span style="color: black;">而后</span>得到了如左图的<span style="color: black;">阻碍</span>物深度图像,<span style="color: black;">倘若</span><span style="color: black;">根据</span>前面的策略,肯定是要往颜色最前的<span style="color: black;">地区</span>飞行,好那我要是告诉你其实右图是无人机和两座悬崖几何关系的俯视图,请你告诉我无人机会怎么飞,<span style="color: black;">倘若</span><span style="color: black;">根据</span>前述的策略,这又必将是一场刻骨铭心的事故。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cO9R20Eq47u~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=rESb5ViZN%2BtOwpCY%2BoS6F6ZmPcw%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">亦</span>许有人对windows98时代的迷宫屏保还有印象,屏保中,<span style="color: black;">运用</span><span style="color: black;">始终</span>沿着左侧墙壁和<span style="color: black;">始终</span>沿着右侧墙壁都会<span style="color: black;">最后</span>走出迷宫,这是<span style="color: black;">由于</span>普通迷宫的拓扑结构就是两条平行线中间有一个通道,<span style="color: black;">根据</span>这种思路,无人机遇到下图这种简单策略躲<span style="color: black;">不外</span>去的<span style="color: black;">阻碍</span>时,完全<span style="color: black;">能够</span>采取类似的方式,就<span style="color: black;">始终</span>向左<span style="color: black;">或</span>向右寻找出路。即<span style="color: black;">倘若</span>上面右图的局部场景的完整形态<span style="color: black;">倘若</span>如下图所示的话,沿着图中的两条曲线为路径都<span style="color: black;">能够</span>绕过去,<span style="color: black;">倘若</span>场景比下图更<span style="color: black;">繁杂</span>,绕过去的路可就需要苦苦追寻了。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cO9hIHCvH1F~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=oJR%2BH%2BFxaKjUY8Ful4m8%2BsQ%2Fqww%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">虽然关于<span style="color: black;">设备</span>人在未知场景中的避障<span style="color: black;">办法</span><span style="color: black;">科研</span>非常多,<span style="color: black;">然则</span><span style="color: black;">因为</span>终究是未知场景,其中必然有<span style="color: black;">繁杂</span>的搜索过程。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">场景建模和路径搜索</strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">再回到狗狗绕过大楼的例子,看下图,<span style="color: black;">倘若</span>狗狗左侧右侧的路都走过,<span style="color: black;">况且</span>右侧其实<span style="color: black;">无</span>那棵树的话,很显然的它下一次绕过这座楼的时候基本上会<span style="color: black;">选取</span>左侧的路(<span style="color: black;">然则</span>当右侧有颗邪恶的树之后,结论仿佛有所改变),这是<span style="color: black;">由于</span>它大脑里<span style="color: black;">已然</span>有了一幅地图,即有了这个场景的模型。</p><img src="https://p26-sign.toutiaoimg.com/pgc-image/RR9cO9v5QLC6u8~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=OFr%2FljSNGYCep7aZFOGkgFsphlw%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">无人机<span style="color: black;">亦</span>是如此,无论是基于电子地图,还是其他<span style="color: black;">源自</span>,还是SLAM(即时定位与地图构建)<span style="color: black;">得到</span>了场景模型,就<span style="color: black;">能够</span>在机载计算机里用算法去搜索优化的避障路径。当前关于这种已知场景路径规划的<span style="color: black;">科研</span><span style="color: black;">非常多</span>,算法<span style="color: black;">亦</span>是层出不穷(算法太多太<span style="color: black;">繁杂</span>,本文暂不展开讨论),<span style="color: black;">亦</span>是无人机避障发展的必然趋势。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOA8HP5N5dx~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=kFS%2BMUJBgPpigw54BIecMfR1udE%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">与传统的<span style="color: black;">设备</span>人避障技术<span style="color: black;">科研</span>相比,当前无人机的避障还<span style="color: black;">处在</span>很初级的<span style="color: black;">周期</span>,但<span style="color: black;">因为</span>消费级无人机市场的<span style="color: black;">火热</span>,<span style="color: black;">大众</span><span style="color: black;">亦</span>都在争先恐后地开展此类<span style="color: black;">科研</span>,<span style="color: black;">能够</span>预见,<span style="color: black;">将来</span>的避障时代中,将会有各自各样<span style="color: black;">此刻</span>难以想到的<span style="color: black;">办法</span>用于无人机避障,有了这些技术的辅助,无人机的操作<span style="color: black;">亦</span>将越来越安全,越来越简单。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">无人机避障技术种类</strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">红外避障</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">红外线的应用<span style="color: black;">咱们</span>并不陌生:从电视、空调的遥控器,到酒店的自动门,都是利用的红外线的感应原理。而<span style="color: black;">详细</span>到无人机避障上的应用,红外线避障的<span style="color: black;">平常</span>实现方式<span style="color: black;">便是</span>“三角<span style="color: black;">测绘</span>原理”。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">红外感应器<span style="color: black;">包括</span>红外发射器与CCD检测器,红外线发射器会发射红外线,红外线在物体上会<span style="color: black;">出现</span>反射,反射的光线被CCD检测器接收之后,<span style="color: black;">因为</span>物体的距离D<span style="color: black;">区别</span>,反射<span style="color: black;">方向</span><span style="color: black;">亦</span>会<span style="color: black;">区别</span>,<span style="color: black;">区别</span>的反射<span style="color: black;">方向</span>会产生<span style="color: black;">区别</span>的偏移值L,<span style="color: black;">晓得</span>了这些数据再经过计算,就能得出物体的距离了,如下图所示。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOALHLx02Qh~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=BmOgvmmlsRGqVmfrcDU6E0FFX%2Fg%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">超声波避障</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">超声波其实<span style="color: black;">便是</span>声波的一种,<span style="color: black;">由于</span>频率高于20kHz,<span style="color: black;">因此</span>人耳听不见,并且指向性更强。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">超声波测距的原理比红外线更加简单,<span style="color: black;">由于</span>声波遇到<span style="color: black;">阻碍</span>物会反射,而声波的速度已知,<span style="color: black;">因此</span>只需要<span style="color: black;">晓得</span>发射到接收的时间差,就能<span style="color: black;">容易</span>计算出<span style="color: black;">测绘</span>距离,再结合发射器和接收器的距离,就能算出<span style="color: black;">阻碍</span>物的<span style="color: black;">实质</span>距离,如下图所示。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOJZGEeFqgI~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=J%2Fs%2FaADFHI7kk58TFpwAi9i0QHs%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">超声波测距相比红外测距,价格更加便宜,相应的感应速度和精度<span style="color: black;">亦</span>逊色<span style="color: black;">有些</span>。<span style="color: black;">一样</span>,<span style="color: black;">因为</span>需要主动发射声波,<span style="color: black;">因此</span><span style="color: black;">针对</span>太远的<span style="color: black;">阻碍</span>物,精度<span style="color: black;">亦</span>会随着声波的衰减而降低,<span style="color: black;">另外</span>,<span style="color: black;">针对</span>海绵等吸收声波的物体<span style="color: black;">或</span>在大风干扰的<span style="color: black;">状况</span>下,超声波将<span style="color: black;">没法</span>工作。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">激光避障</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">激光避障与红外线类似,<span style="color: black;">亦</span>是发射激光<span style="color: black;">而后</span>接收。<span style="color: black;">不外</span>激光传感器的<span style="color: black;">测绘</span>方式<span style="color: black;">非常多</span>样,有类似红外的三角<span style="color: black;">测绘</span>,<span style="color: black;">亦</span>有类似于超声波的时间差+速度。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">但无论是哪种方式,激光避障的精度、反馈速度、抗干扰能力和有效范围都要<span style="color: black;">显著</span>优于红外和超声波。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">但<span style="color: black;">这儿</span><span style="color: black;">重视</span>,不管是超声波还是红外、<span style="color: black;">也</span>或是<span style="color: black;">这儿</span>的激光测距,都只是一维传感器,只能给出一个距离值,并<span style="color: black;">不可</span>完成对现实三维世界的感知。当然,<span style="color: black;">因为</span>激光的波束极窄,<span style="color: black;">能够</span><span style="color: black;">同期</span><span style="color: black;">运用</span>多束激光<span style="color: black;">构成</span>阵列雷达,近年来此技术<span style="color: black;">逐步</span>成熟,多用于自动驾驶车辆上,但<span style="color: black;">因为</span>其体积庞大,价格昂贵,故不太适用于无人机。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">视觉避障</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">处理</span><span style="color: black;">设备</span>人<span style="color: black;">怎样</span>“看”的问题,<span style="color: black;">亦</span><span style="color: black;">便是</span><span style="color: black;">大众</span>常听到的计算机视觉(Computer Vision)。其<span style="color: black;">基本</span>在于<span style="color: black;">怎样</span>能够从二维的图像中获取三维信息,从而<span style="color: black;">认识</span><span style="color: black;">咱们</span>身处的这个三维世界。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">视觉识别系统<span style="color: black;">一般</span><span style="color: black;">来讲</span><span style="color: black;">能够</span><span style="color: black;">包含</span>一个或两个摄像头。单一的照片只<span style="color: black;">拥有</span>二维信息,犹如2D电影,并无直接的空间感,<span style="color: black;">仅有</span>靠<span style="color: black;">咱们</span>自己依靠“物体遮挡、近大远小”等生活经验脑补。故单一的摄像头获取到的信息及其有限,并<span style="color: black;">不可</span>直接得到<span style="color: black;">咱们</span>想要的效果(当然能够<span style="color: black;">经过</span><span style="color: black;">有些</span>其他手段,辅助获取,<span style="color: black;">然则</span>此项还不成熟,并<span style="color: black;">无</span>大规模验证)。类比到<span style="color: black;">设备</span>视觉中,单个摄像头的<span style="color: black;">照片</span>信息<span style="color: black;">没法</span>获取到场景中<span style="color: black;">每一个</span>物体与镜头的距离关系,即缺少第三个维度。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">如下图所示,单一的<span style="color: black;">照片</span><span style="color: black;">拥有</span>很强的迷惑性和不确定性</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOJpERdA0EB~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=EIVSd4Ye78DKsdp%2FZjumAabg%2Flg%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">双目立体视觉犹如3D电影(<span style="color: black;">上下</span>眼看到的场景略有差异),能够直接给人带来<span style="color: black;">剧烈</span>的空间临场感。类比<span style="color: black;">设备</span>视觉,从单个摄像头升级到两个摄像头,即立体视觉(Stereo Vision)能够直接<span style="color: black;">供给</span>第三个维度的信息,即景深(depth),能够更为简单的获取到三维信息。双目视觉最<span style="color: black;">平常</span>的例子<span style="color: black;">便是</span><span style="color: black;">咱们</span>的双眼:<span style="color: black;">咱们</span>之<span style="color: black;">因此</span>能够准确的拿起面前的杯子、判断汽车的远近,都是<span style="color: black;">由于</span>双眼的双目立体视觉,而3D电影、VR眼镜的发明,<span style="color: black;">亦</span>都是双目视觉的应用。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">双目视觉的基本原理是利用两个平行的摄像头进行拍摄,<span style="color: black;">而后</span><span style="color: black;">按照</span>两幅图像之间的差异(视差),利用一系列<span style="color: black;">繁杂</span>的算法计算出特<span style="color: black;">选定</span>的距离,当数据足够时还能生成深度图。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOK0w3Shhx~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=n6qd5Yt8Y7vbz9g1bumN9Kk21s8%3D" style="width: 50%; margin-bottom: 20px;"><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOKC5Gipd2g~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=2Oz0CdZkoZrzehBtN5ZMdKpUXnk%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">其实,各个避障技术在无人机上都有用武之地,只是应用场景有所<span style="color: black;">区别</span>,<span style="color: black;">尤其</span><span style="color: black;">针对</span>前视避障而言,有些技术就不适用了。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">红外和超声波技术,<span style="color: black;">由于</span>都需要主动发射光线、声波,<span style="color: black;">因此</span><span style="color: black;">针对</span>反射的物体有<span style="color: black;">需求</span>,<span style="color: black;">例如</span>:红外线会被黑色物体吸收,会穿透透明物体,还会被其他红外线干扰;而超声波会被海绵等物体吸收,<span style="color: black;">亦</span>容易被桨叶气流干扰。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">况且</span>,主动式测距还会产生两台<span style="color: black;">设备</span>相互干扰的问题。相比之下,虽然双目视觉<span style="color: black;">亦</span>对光线有<span style="color: black;">需求</span>,<span style="color: black;">然则</span><span style="color: black;">针对</span>反射物的<span style="color: black;">需求</span>要低<span style="color: black;">非常多</span>,两台<span style="color: black;">设备</span><span style="color: black;">同期</span><span style="color: black;">运用</span><span style="color: black;">亦</span>不会互相干扰,普适性更强。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">最重要的是,<span style="color: black;">平常</span>的红外和超声波<span style="color: black;">日前</span>都是单点测距,只能<span style="color: black;">得到</span>特定方向上的距离数据,而双目视觉<span style="color: black;">能够</span>在小体积、低功耗的前提下,<span style="color: black;">得到</span>眼前场景的比较高分辨率的深度图,这就让避障功能有了<span style="color: black;">更加多</span>的发展空间,<span style="color: black;">例如</span>避障之后的智能飞行、路径规划等。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">激光技术虽然<span style="color: black;">亦</span>能实现类似双目视觉的功能,<span style="color: black;">然则</span>受限于技术发展,<span style="color: black;">日前</span>的激光元件<span style="color: black;">广泛</span>价格贵、体积大、功耗高,应用在消费级无人机上既不经济<span style="color: black;">亦</span>不实用。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;">无人机避障实现的难点</strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">避障功能从构思到实现,走的每一步几乎都便随着无数的<span style="color: black;">困难</span>。仅仅是写出有效的视觉识别<span style="color: black;">或</span>地图重构的算法还只是<span style="color: black;">第1</span>步,能让它在无人机<span style="color: black;">这般</span>一个计算能力和功耗都有限制的平台上流畅稳定的跑起来,才是真正困难的<span style="color: black;">地区</span>。</p><img src="https://p3-sign.toutiaoimg.com/pgc-image/RR9cOKOF0542V~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1724919746&x-signature=G2OcNgQDxczMkxGtDVMYELcE1Zg%3D" style="width: 50%; margin-bottom: 20px;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">另外</span>,<span style="color: black;">怎样</span>处理功能的边界<span style="color: black;">亦</span>是一个问题,<span style="color: black;">例如</span>双目视觉在视线良好的<span style="color: black;">状况</span>下<span style="color: black;">能够</span>工作,<span style="color: black;">那样</span>当有灰尘遮挡的<span style="color: black;">状况</span>下呢?这就需要<span style="color: black;">持续</span>的实验和试错,并且<span style="color: black;">连续</span>的优化算法,<span style="color: black;">保准</span>各项功能在各类场景下都能正常工作,不会给出错误的指令。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">避障功能<span style="color: black;">做为</span>近年来无人机<span style="color: black;">制品</span>的大趋势,带来的最直接的好处<span style="color: black;">便是</span>,以往<span style="color: black;">有些</span>人为疏忽<span style="color: black;">导致</span>的撞击,<span style="color: black;">此刻</span>都能经由避障功能去避免,既<span style="color: black;">保证</span>了无人机飞行安全的<span style="color: black;">同期</span>,<span style="color: black;">亦</span>避免了对<span style="color: black;">周边</span>人员财产的损害,让飞无人机的门槛进一步得到了降低。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">而长远来看,无人机想要普及到农业、建筑、运输、<span style="color: black;">媒介</span>等<span style="color: black;">行业</span>,「智能化」肯定是必经之路。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">毕竟<span style="color: black;">仅有</span>在飞行功能上做到智能<span style="color: black;">掌控</span>,才有余量去满足<span style="color: black;">区别</span>行业的<span style="color: black;">需要</span>。如今由“避障功能”而衍生出的一系列“智能飞行”功能,无疑<span style="color: black;">便是</span>“无人机智能化”的<span style="color: black;">周期</span>性<span style="color: black;">表现</span>之一。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">END</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">本文已获传感器技术转载授权。著作权解释权属原创者所有,本文TechSugar编辑部<span style="color: black;">举荐</span>阅读!</p>
回顾历史,我们感慨万千;放眼未来,我们信心百倍。 楼主发的这篇帖子,我觉得非常有道理。
页:
[1]