Chinese have a saying: "Food", people value of the diet has always been throughout the long history of the development of Chinese civilization. But in recent years, frequent food safety incidents, the restaurant industry "Houchu powerhouse" Compliance problems have gradually embarked on the focus of public opinion, become the focus of regulation where government. So, faced with "food safety greater than days," the demands of society, cloud computing, big data and artificial intelligence technology can play what role?
"Ming Liang kitchen stove" What's hiding something?
Since 2015, the State Food and Drug Administration's "bright bright kitchen stove" project has been extended nationwide. The purpose is to make food enterprises "Houchu powerhouse" walked from behind the front desk, to guide the industry transparency, visualization, to achieve universal supervision, food and beverage companies to fix Forced internal strength, improve the management level, so as to further effectively protect food safety.
However, for the regulatory agencies, food stores video inspection systems are mostly localized, lack of networking capabilities, you can not view the playback anywhere, while vast amounts of video data can not identify critical information and real-time analysis, information monitoring platform construction has become urgent demand. For food and beverage companies, is also facing food factories, central kitchens, chain stores and other video localization systems, networking is limited, the headquarters of the issue can not be viewed in real-time viewing. At the same time, food and beverage companies are faced with in a proactive regulatory process rely only on manual analysis of video in low efficiency, operation and maintenance monitoring platform for the deployment of high cost and other problems.
With the full advent of the cloud, "bright kitchen light kitchen" also ushered in the best opportunity on the cloud. Then, in the cloud, big data, blessing AI technology, catering companies and regulatory efficiency "bright kitchen light kitchen" What if the coverage will be effective to enhance it? Mounted on the cloud, "Ming Liang kitchen stove" what is different?
Visible, able to understand the true · · Houchu intelligent management platform
In fact, the food and beverage industry, "Liang Ming kitchen stove" project has a number of stores and more widely distributed, difficult to deploy transformation, content delivery safe, a large potential demand for business analysis AI features.
In order to solve industry problems, Ali cloud based on widely distributed cloud computing infrastructure, equipped with video access, device management, video storage, video distribution, and other full link live look back at the video processing capabilities, integrated into artificial intelligence technology, to build a one-stop video access and intelligent visual platform to help "clear bright kitchen stove" true networking, intelligent lightweight. At the same time, using the model more in line with AI algorithms catering business scenarios, the behavior of staff catering, health data, monitoring the effectiveness of picture identification such as high efficiency, targeted to provide food and beverage industry bright kitchen light kitchen solutions, enabling the full Food industry to enhance regulatory efficiency.
Solution has the following advantages:
1, the deployment operation and maintenance simple, asset-light, fast start
- Stores and factories only deploy front-end equipment, operation and maintenance costs low deployment
- Network, compute, elastic storage expansion, the business can focus camera, cloud flexible scheduling, low-cost fast start
- Back-end video + AI using cloud services, on-demand, pay-per-volume, asset-light
2, high-quality, all-access video link distribution deal with security
- Relying on CDN and broadcast services, proven components, maturity, stability, security, performance guarantees. CDN edge node using a transceiver, a good cross inter-network coverage, the device supports ten million concurrent Push stream
- Support for cloud video storage and custom storage strategy, while supporting real-time live, look back at history, as well as strategies such as video streaming screenshots
- SAG support to conduct VPC hybrid network to ensure data isolation and encryption transmission
3, open standards for easy integration
- Devices, platforms, open protocol, quick to GB GB / T28181 or rtmp access multi-vendor cameras, network storage devices, all ability OpenAPI open, easy top business integration
4, AI high recognition accuracy
- Covering the focus detection algorithm scene bright kitchen light kitchen in food testing scenarios concern
- AI recognition accuracy rate of 95% or more, may be used as regulatory standards
5, AI algorithm iterative upgrade convenience
- Cloud algorithm to improve the capacity for sustainable training, the type of algorithm can be continuously enriched beneficial effect of AI applications to lead the industry, not the local operator constraints force
AI enabling clear bright kitchen stove, intelligent visual show its mettle
Application of AI technology for food and beverage industry improve operational and regulatory efficiency is essential, Ali cloud catering industry bright bright kitchen stove solve behind the program is relying on intelligent vision (IntelligenceVision) product recognition and AI algorithm iterations.
智能视觉，可以为多层次视觉AI需求提供一站式服务，支持直接通过API调用标准算法，或将已有个性化算法模型导入平台进行推理，还能以零技术门槛的方式，通过上传并标注数据集，自助训练定制化模型，实现创新算法的调用。
智能视觉AI算法具有三大优势：
1、智能的数据收集及标注流程
数据的重要性对于深度学习来说毋庸置疑，没有优良的行业垂直数据，即使有再好的算法模型，也不可能取得好的结果。如何能够更高效地收集数据并且正向反馈给模型，这是所有AI公司的痛点。针对业务数据收集难的问题，阿里云创造了一套智能的利用算法收集数据和数据标注流程，完成了从数据到算法再到数据的闭环，能使业务数据快速回流，帮助业务快速迭代算法，形成良性循环。
2、进阶的后厨场景的算法模型
在用餐高峰期，后厨的工作场地是非常拥挤的，同时摄像头也会被水汽油烟影响，传统的检测算法在这些场景下会严重受限甚至不可用，考虑到厨房垂直场景的特点，阿里云结合当前业内领先算法，针对进行了一系列的改进，后厨环境下目标检测算法模型的速度远远快于当前的开源算法模型，同时检测准确率也高于当前的开源算法10%以上。
3、精准的属性分类模型
阿里云联合集团内部食品卫生安全标杆部门，精心设计深度神经网络特征并结合注意力机制，最后呈现更准确的属性分类模型，能够排除各种厨房用品和工作人员操作的姿势的干扰，比各开源模型有质的飞越。
智能视觉在明厨亮灶场景中的应用：
1、人员身份核验
对厨房内人员进行身份比对，判断是否为正规持健康证上岗的工作人员
2、人员穿戴规范检测
对厨房内人员进行穿戴识别，判断是否正确佩戴有口罩、厨师帽、手套
3、人员行为规范检测
对厨房内人员进行为识别，判断是否有违规行为 ，如抽烟、玩手机等
4、垃圾桶盖检测
识别厨房内垃圾桶盖未闭合的情况
5、老鼠蟑螂检测
对厨房内出现老鼠、蟑螂的情况进行识别。夜间需红外热感摄像头支持，需清晰度足够高
6、画面有效性检测
对视频画面过亮、过暗、黑屏、花屏、遮挡、角度不佳等情况进行检测
7、自助训练AI模型
支持企业从视频监控、视频直播、OSS中读取视频流或图片作为数据集上传、打标，一键训练、模型验证和迭代
（后厨AI识别效果示例-截帧）
技术赋能盒马鲜生与饿了么食安升级
目前，阿里云一站式视频接入与智能视觉平台，已经在盒马鲜生和饿了么成功落地。
盒马是阿里巴巴集团旗下以技术驱动的新零售平台，对内有严格的后厨食安标准规范，并有门店巡检系统。盒马希望通过整合门店视频接入+AI分析能力，将后厨巡检工作自动化，以提高巡检效率，及时发现并整改问题。
阿里云为盒马提供明厨亮灶解决方案，通过国标摄像头采集视频并推送视频至云端，盒马后台可以进行实时远程直播观看，对于未戴口罩/帽子/手套等违规行为，系统可以进行AI识别并实时预警，盒马在接到预警后，联动到巡检系统，记录门店巡检结果，并实时推送给门店负责人立即整改。整个方案大幅缩短了巡检时间，降低巡检差旅费用，提高了食品安全人员的工作效率。更重要的是通过实时的识别问题和跟进整改结果，可以有效提高食品安全管理的实时性。
饿了么已经在去年开通了明厨亮灶频道，阿里云为饿了么提供基础视频汇聚能力与场景化AI算法，共同打造智能化、可视化、系统化的明厨亮灶平台。项目中，通过视频的汇聚与分发，公众可以通过APP在线观看餐馆后厨的直播，提升餐饮商家对食安问题的重视程度。同时，基于AI算法对外卖后厨巡查视频中的人员着装、食品存放、场所环境、制作加工、标识、摄像头等进行违规甄别并及时整改，让消费者能够放心点餐。
如何接入“明厨亮灶”？
第一步，登录阿里云官网，选择 企业应用 > 视频云 > 智能视觉/视频监控。
第二步，分别在产品详情页面，单击 立刻开通。
第三步，在智能视觉产品页中选择“授权管理”，对“视频监控”与“MNS访问授权”选择 去授权。
The fourth step, namely system model, the industry scene to add to my model as a "bright kitchen light kitchen" model.
The fifth step, through the API interface, based on the model ID to start using the service.