2022 Smart Car Competition: First Round of Preparation Guide for Smart Vision Group

When formulating the rules of this competition, we communicated with Teacher Zhuo many times. Based on the principle of "both inheritance and difference", we focused on retaining the image recognition part of last year's intelligent vision group and expanded it appropriately, upgrading the image recognition from two categories (animals and fruits) to three categories (with vehicles added). At the same time, in order to improve the viewing experience of the subsequent finals, we added recognition of subcategories within the larger categories. For example, in the animal category, we need to distinguish between cats, dogs, horses, etc.

In addition, in order to increase the fun, last year's target shooting was changed to picture moving, which slightly reduced the difficulty of the task. This task is planned to be used in the final stage.

A major change in the rules of the Intelligent Vision Group this year compared to last year is the track. Instead of the car following the track on a specified track, it will autonomously search for the target in a venue without a fixed track. The coordinates of the target on the venue are given, and the car will follow the map to drive to the target and identify it.

This task requires the car to plan its own route and use the inertial navigation principle to determine its own position, direction and speed. The accuracy of inertial navigation determines the time it takes to find the target, and therefore greatly affects the performance of the entire competition.

Considering that inertial navigation is being used in smart car competitions for the first time and the accuracy of inertial navigation is difficult to grasp, the rules add several AprilTag signs at fixed locations in the venue. The car can use cameras to find and identify the signs and use the principle of triangulation to determine its own position.

Through inertial navigation, triangulation positioning, wheel speed and mileage control and other means, the vehicle can be accurately positioned to complete target search, identification and pickup.

In order to help students sort out some concepts and lay a solid foundation for subsequent training, NXP engineers will conduct a basic training for the intelligent vision group of the next competition at 2 pm on December 15, 2021, which will mainly involve AI intelligent recognition and inertial navigation.

In addition, in response to questions from many students about what kind of NXP MCU can be used, a brief introduction will also be given during the training.

• December 15, 2021, 2:00 p.m.

  
  

• Yan Zhang, System Engineer at NXP
  

• Yang Xi, System Engineer at NXP

  
  

• Introduction to NXP MCU

• Introduction to Intelligent Vision Module

  - OpenART module introduction

• Model training, deployment methods, and debugging methods

  - Model acquisition and training

  - Model deployment

  - Model debugging

• Key factors affecting the actual accuracy of the model

  - Dataset augmentation

  - Environmental interference

  - Model input data preprocessing

  
  

• Inertial Navigation and Inertial Devices (Accelerometers and Gyroscopes)

• 2D inertial navigation algorithm (matlab example)

• Inertial navigation error characteristics and correction

• Inertial navigation correction information - integrated navigation: integrating more sensors (odometer/UWB/camera/GPS)

• Introduction to Fusion Algorithm (Kalman)

Friends who wish to participate in this training

↓↓↓

Since the AI ​​recognition part of next year's competition is the same as this year's, students can also refer to the following previous tweets when preparing for the competition:

• Introducing NXP’s MCU Machine Learning Education Kit – OpenART, click to view>>

• Reference answers for the AI ​​Vision Group of the Smart Car Competition, click to view>>

• Smart Car Competition, AI Vision Group Competition Question Analysis, click to view >>

• AI Vision Group’s Immortal Step Model Quantization, click to view>>

• AI Vision Team’s Immortal Step Model Tuning, click to view>>

• AI Vision Team’s Advanced Gameplay: Returning from Python to C Language, click to view>>

• AI machine learning in action: electromagnetic smart car, click to view>>

• AI in Action: Electromagnetic Intelligent Car (Continued) - Data Processing, Training Details, Click to View>>