Visual Measuement

Visual measurement based on CNN and reinforcement learning. The target of the project is to measure the objects based on monocular camera, compare with the actual size of objects, and quantify the error.

---

Project Plan

Week 1

The first assignment is mainly about digital image representation and processing, whcih can refer to Digital Image Processing written by R.C.Gonzalez.

• Gray-scale transformation
• Histogram
• Convolution
• Spatial filtering
• Smoothing and Sharpening
• Edge Filter
• Frequency Domain Filter
• Image Segementation
• Edge Detection Operator about points and lines
• Morphological processing

Week 2

Realise the size measurement of objects based on the principle of Hough Transform.

• Programming Language: Python, MATLAB or C++
• Learn the basic principle of Hough Transform
• Realise camera-based measurment based

Reference:

• 基于图像分割的目标尺寸特征测量
• (OpenCV)图像目标尺寸检测
• python-opencv尺寸测量

Week 3

• Learn the knowledge about CNN and realize handwritten recognition according to MNIST data set.

Reference

• CNN(Andrew_NG)
• CNN(TensorFlow)

• Learn the ranging principle based on monocular vision, and realize the target ranging based on monocular camera.

Reference

• 单目测距原理与实现
• 单目测距原理与实现（代码可运行）

Week 4

• Learn the principles in the 2020 ECCV paper "Single View Metrology in the Wild".
• Measure the objects based on monocular camera, compare with the actual size of objects, and quantify the error.

Reference

• Single-View-Metrology
• Translation of the paper
• 2020 ECCV: "Single View Metrology in the Wild".