AWS IoT - ESP32-CAM and Rekognition - Lab 98

In this lab an ESP32 with camera will connect to IoT Core, subscribe to a topic, obtain pre-sign url to S3 object for an upload, then upload the picture to S3.

Services Covered

IAM
S3
IoT Core
AWS Lambda
Amazon Rekognition

Lab description

In this lab an ESP32 with camera will connect to IoT Core, subscribe to a topic, obtain pre-sign url to S3 object for an upload, then upload the picture to S3. Two Rules in IoT Core will trigger two Lambda functions, one that will generate mentioned presigned url, second will get the S3 object and send it to rekognition and publish the result to IoT topic

Creating Things, Roles, attaching certificates and policies
Creating S3 buckets
Changing CORS configuration of S3 buckets
Creating Lambda function and giving them necessary permissions
Creating presigned urls for S3 objects

Lab diagram

Lab date

24-09-2021

Prerequisites

ESP32-CAM (Ai-Thinker ESP32-S in this project)
FTDI cable for connecting device through USB
AWS account
Visual Studio Code with PlatformIO IDE extension installed or Arduino IDE
POSTMAN to make API calls

Lab source

The Internet of Things on AWS – Official Blog

Lab steps

Create a thing in IoT Core, either in Console or through CLI
Get your IoT device data endpoint (AWS IoT Core -> Settings) or through CLI
```
aws iot describe-endpoint --region <YOUR REGION>
```

Reformat you certificates with ” and n, it’s time consuming but necessary for the device to connect. It’ll look like that:

"-----BEGIN CERTIFICATE-----n"
"MIIDQTCCAimgAwIBAgITBmyfz5m/jAo54vB4ikPmljZbyjANBgkqhkiG9w0BAQsFn"
"ADA5MQswCQYDVQQGEwJVUzEPMA0GA1UEChMGQW1hem9uMRkwFwYDVQQDExBBbWF6n"
"b24gUm9vdCBDQSAxMB4XDTE1MDUyNjAwMDAwMFoXDTM4MDExNzAwMDAwMFowOTELn"

Create a policy for the device, policy.json

{
...
  {
    "Effect": "Allow",
    "Action": "iot:Connect",
    "Resource": "arn:aws:iot:eu-central-1:797321539388:client/ESP32-cam"
  },
  {
    "Effect": "Allow",
    "Action": "iot:Subscribe",
    "Resource": "arn:aws:iot:eu-central-1:797321539388:topicfilter/esp32/sub/data"
  },
...
}

Create S3 Bucket

under permissions add bucket policy allowing “s3:getObject” and “s3:PutObject” for your IAM User

add CORS configuration

[
  {
      "AllowedHeaders": [
          "*"
      ],
      "AllowedMethods": [
          "GET",
          "PUT",
          "POST",
          "DELETE"
      ],
      "AllowedOrigins": [
          "*"
      ],
      "ExposeHeaders": []
  }
]

Create two Lambda functions:
- one to generate presigned url in S3 called esp32-request-url.py and change it’s policy and allow publishing in IoT and getting objects from S3 [esp32-request-url-policy.json]()
- second called esp32-request-rekognition.py will get the object from S3 when triggered by IoT Rule and send it to Rekognition, then it’ll publish result to topic ‘esp32/sub/data’
Create two Rules in IoT Core:
- one with query statement:
```
SELECT * FROM 'esp32/pub/url'
```
  and action to send message to esp32-request-url Lambda function
- second with query statement:
```
SELECT * FROM 'esp32/pub/data'
```
  action will be send message to esp32-request-rekognition Lambda function
Test if your permissions allow for posting to topics and if Lambda generate url and response from Rekognition (you can use Postman for API calls and sending pictures)
The AWS part is complete. Since this is an AWS Challenge I won’t be going in to the details of programming ESP32 here. Please refer to Nathan Glovers GitHub page for more details.

Lab files

thing_policy.json – policy granting ESP32-Cam necessary allowed actions in IoT Core
esp32-request-rekognition.py – Lambda function code to generate presigned url in S3
esp32-request-url-policy.json – execution policy for Lambda function
esp32-request-rekognition.py – Lambda function triggered by IoT set on topic, gets picture from S3, sends it to Rekognition
esp32-request-url-policy.json – execution policy for Lambda function
esp32-request-rekognition-policy.json

Acknowledgements

AWS IoT – ESP32-CAM and Rekognition – Lab 98