遥遥领先瑞萨抗冲击宽温有源晶振，近年来，人工智能被用于各种应用，据说可以保护人们的安全，丰富他们的生活。瑞萨在2020年推出了RZ/V系列作为视觉人工智能的微处理器(MPU)。RZ/V系列一直备受好评，各种来源都有许多关于其高速人工智能推理和与端点人工智能匹配的低功耗性能的询问。在这篇博客中，我们想介绍一些我们的合作伙伴使用RZ/V系列的实施示例，以及提供安全性和舒适性的人工智能。

用人工智能密切保护人们

虽然对策已经到位，但是很难减少常见事故的数量，例如小孩摔倒、跌倒、被缝隙夹住等。我们的合作伙伴使用RZ/V系列开发并实现了更先进的姿态检测。其关键特性是高度精确地检测手腕、肘部和脚踝，使系统能够快速提醒用户注意事故，如从站立位置跌倒、进入有楼梯或台阶的区域，或者手或头被游乐场设备或门卡住等，以确保安全。而这些人工智能的应用都离不开有源晶体振荡器的支持，该系统可以在一个人把手放在楼梯上或踏上楼梯的瞬间检测到危险；而在以前，该系统只能在人完全爬上楼梯后检测到危险，因为它将人体识别为一个整体。

该解决方案还支持多人检测。当部署在医院或老年人护理机构时，单个摄像机可用于同时保护多个人。该系统可以预测涉及从床上跌落的事故，并估计一个人在房间里的位置以及他们的状况(站起、坐下或摔倒)，从而减轻护理人员的负担。

预防工地事故

毫不夸张地说，在物流运输行业的仓库里，在叉车、自动导引车(AGV)等无人运输机器人四处移动的地方，在起重机、挖掘机等施工设备四处移动的工地里，危险总是在你身边。减少叉车与人接触的事故数量尤其困难，因此需要强有力的安全措施。我们的合作伙伴开发的独特的对象检测模型以及高精密的温度补偿晶振，在RZ/V系列上实施，具有高度准确性，执行高速推理，并具有使用单目摄像机的距离测量功能。换句话说，通过在这种车辆设备上结合具有这种AI模型的相机模块，可以加强周围的监控，特别是因为它可以实时测量与正在接近的人的距离。预计这将提高现场安全性并保持运营效率。

进一步的发展包括将安全监控领域从工作场所扩展到普通人行道，我们可以通过在自行车和踏板车上安装该系统来促进行人安全。

轻量级、超快速的AI模型实现了10人检测、18个关节计数和30FPS，是姿势检测实现更好生活的另一个实施示例。由于健康意识的增强，近年来对健身计划的需求一直在增加，而对于高品质的石英晶体振荡器需求也在不断增加。如果将该系统应用于舞蹈和工作室课程等。它可以通过将参与者与教练进行比较来立即纠正参与者的核心和姿势，从而带来显著的健身益处以及损伤预防。除了姿势矫正之外，由于新冠肺炎而难以处理的身体距离(参与者之间的距离)现在也可以轻松确保，让您安心。此外，通过向称为“十字路口”的特定区域添加数量计数，它可以检测是否有人进入特定区域，从而允许AI管理进入工作室的人数，从而减轻操作管理的负担。遥遥领先瑞萨抗冲击宽温有源晶振.

AI is used for a variety of applications and is said to safeguard people and enrich their lives in recent years. Renesas launched the RZ/V series as microprocessors (MPUs) for Vision AI in 2020. The RZ/V series has been highly acclaimed and there have been many inquiries from various sources regarding its high-speed AI inference and low consumption performance that match endpoint AI. In this blog, we would like to introduce some implementation examples by our partners using the RZ/V series, and AI that provides safety and comfort.

Safeguarding People Closely with AI

Although countermeasures are in place, it has been difficult to reduce the number of familiar accidents such as small children falling over, falling down, getting caught in gaps, etc. Our partner has developed and implemented more advanced pose detection using the RZ/V series. The key feature is highly accurate detection at the level of the wrist, elbow, and ankle, enabling the system to quickly alert users to accidents such as falling from a standing position, entering an area with stairs or steps, or getting one's hand or head caught in playground equipment or doors, etc to ensure safety. This system can detect danger right at the moment when a person puts his or her hand or steps on a staircase; whereas previously, the system could only detect danger after the person had completely climbed the staircase because it recognized the human body as a single mass.

This solution also supports multiple-person detection. A single camera can be used to safeguard multiple people simultaneously when deployed in hospitals or elderly care facilities. The system can predict accidents involving falls from beds and estimate where a person is in the room as well as their condition (standing up, sitting down or falling over), thus reducing the burden on caregivers.

Guarding Against Worksite Accidents

It is no exaggeration to say that danger is always next to you in the warehouses of logistics and transportation industries, where unmanned transport robots such as forklifts and (Automated Guided Vehicles (AGVs) move around, and in construction sites where construction equipment such as cranes and excavators move around. It is particularly difficult to reduce the number of accidents involving forklifts making contact with people, therefore strong safety measures are needed. The unique object detection model developed by our partner, implemented on the RZ/V series, is highly accurate, performs high-speed inference, and has a distance measurement function using a monocular camera. In other words, it is possible to strengthen surrounding monitoring by incorporating a camera module with this AI model on such vehicle equipment, especially since it can measure the distance to approaching people in real time. This is expected to enhance safety on site as well as maintain operational efficiency.

Further development includes expanding the area of safety monitoring from work sites to general sidewalks, and we could contribute to pedestrian safety by installing the system on bicycles and kick scooters.

Support Better Living

The lightweight, ultra-fast AI model achieving 10-person detection, 18 joint counts and 30 FPS is another implementation example of pose detection for better living. The demand for fitness programs has been increasing in recent years due to growing health consciousness. If this system is applied to dance and studio lessons, etc., it could instantly correct the participant's core and posture by comparing the participant with the instructor, resulting in significant fitness benefits as well as injury prevention. In addition to posture correction, physical distancing (distance between participants) that has been difficult to deal with due to COVID-19 can now be easily ensured, providing peace of mind. Furthermore, by adding a number count to specific areas termed as “line cross,” it could detect whether a person had entered a specific area allowing AI to manage the number of people entering the studio, thereby reducing the burden on the operation management.