遙遙領(lǐng)先瑞薩抗沖擊寬溫有源晶振，近年來，人工智能被用于各種應(yīng)用，據(jù)說可以保護(hù)人們的安全，豐富他們的生活。瑞薩在2020年推出了RZ/V系列作為視覺人工智能的微處理器(MPU)。RZ/V系列一直備受好評(píng)，各種來源都有許多關(guān)于其高速人工智能推理和與端點(diǎn)人工智能匹配的低功耗性能的詢問。在這篇博客中，我們想介紹一些我們的合作伙伴使用RZ/V系列的實(shí)施示例，以及提供安全性和舒適性的人工智能。

用人工智能密切保護(hù)人們

雖然對(duì)策已經(jīng)到位，但是很難減少常見事故的數(shù)量，例如小孩摔倒、跌倒、被縫隙夾住等。我們的合作伙伴使用RZ/V系列開發(fā)并實(shí)現(xiàn)了更先進(jìn)的姿態(tài)檢測(cè)。其關(guān)鍵特性是高度精確地檢測(cè)手腕、肘部和腳踝，使系統(tǒng)能夠快速提醒用戶注意事故，如從站立位置跌倒、進(jìn)入有樓梯或臺(tái)階的區(qū)域，或者手或頭被游樂場(chǎng)設(shè)備或門卡住等，以確保安全。而這些人工智能的應(yīng)用都離不開有源晶體振蕩器的支持，該系統(tǒng)可以在一個(gè)人把手放在樓梯上或踏上樓梯的瞬間檢測(cè)到危險(xiǎn)；而在以前，該系統(tǒng)只能在人完全爬上樓梯后檢測(cè)到危險(xiǎn)，因?yàn)樗鼘⑷梭w識(shí)別為一個(gè)整體。

該解決方案還支持多人檢測(cè)。當(dāng)部署在醫(yī)院或老年人護(hù)理機(jī)構(gòu)時(shí)，單個(gè)攝像機(jī)可用于同時(shí)保護(hù)多個(gè)人。該系統(tǒng)可以預(yù)測(cè)涉及從床上跌落的事故，并估計(jì)一個(gè)人在房間里的位置以及他們的狀況(站起、坐下或摔倒)，從而減輕護(hù)理人員的負(fù)擔(dān)。

預(yù)防工地事故

毫不夸張地說，在物流運(yùn)輸行業(yè)的倉(cāng)庫(kù)里，在叉車、自動(dòng)導(dǎo)引車(AGV)等無人運(yùn)輸機(jī)器人四處移動(dòng)的地方，在起重機(jī)、挖掘機(jī)等施工設(shè)備四處移動(dòng)的工地里，危險(xiǎn)總是在你身邊。減少叉車與人接觸的事故數(shù)量尤其困難，因此需要強(qiáng)有力的安全措施。我們的合作伙伴開發(fā)的獨(dú)特的對(duì)象檢測(cè)模型以及高精密的溫度補(bǔ)償晶振，在RZ/V系列上實(shí)施，具有高度準(zhǔn)確性，執(zhí)行高速推理，并具有使用單目攝像機(jī)的距離測(cè)量功能。換句話說，通過在這種車輛設(shè)備上結(jié)合具有這種AI模型的相機(jī)模塊，可以加強(qiáng)周圍的監(jiān)控，特別是因?yàn)樗梢詫?shí)時(shí)測(cè)量與正在接近的人的距離。預(yù)計(jì)這將提高現(xiàn)場(chǎng)安全性并保持運(yùn)營(yíng)效率。

進(jìn)一步的發(fā)展包括將安全監(jiān)控領(lǐng)域從工作場(chǎng)所擴(kuò)展到普通人行道，我們可以通過在自行車和踏板車上安裝該系統(tǒng)來促進(jìn)行人安全。

輕量級(jí)、超快速的AI模型實(shí)現(xiàn)了10人檢測(cè)、18個(gè)關(guān)節(jié)計(jì)數(shù)和30FPS，是姿勢(shì)檢測(cè)實(shí)現(xiàn)更好生活的另一個(gè)實(shí)施示例。由于健康意識(shí)的增強(qiáng)，近年來對(duì)健身計(jì)劃的需求一直在增加，而對(duì)于高品質(zhì)的石英晶體振蕩器需求也在不斷增加。如果將該系統(tǒng)應(yīng)用于舞蹈和工作室課程等。它可以通過將參與者與教練進(jìn)行比較來立即糾正參與者的核心和姿勢(shì)，從而帶來顯著的健身益處以及損傷預(yù)防。除了姿勢(shì)矯正之外，由于新冠肺炎而難以處理的身體距離(參與者之間的距離)現(xiàn)在也可以輕松確保，讓您安心。此外，通過向稱為“十字路口”的特定區(qū)域添加數(shù)量計(jì)數(shù)，它可以檢測(cè)是否有人進(jìn)入特定區(qū)域，從而允許AI管理進(jìn)入工作室的人數(shù)，從而減輕操作管理的負(fù)擔(dān)。遙遙領(lǐng)先瑞薩抗沖擊寬溫有源晶振.

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.