Terms and concepts

The

Scene classification	noun	meaning
Project & Task	project	A project contains multiple Tasks, and different Tasks can be switched freely.
	Task	A Task contains various configurations required for the current Task operation, including Camera, Robot, Target, Visual Parameter, ROI, etc. A Task only corresponds to one Scene (depalletizing, orderly loading and unloading, disordered grabbing, positioning assembly), and cannot be modified.
Target property	grid file	refers to the file format used to store 3D model data, mainly including OBJ, STL and PLY formats.
	Point Cloud	Point Cloud is a set of points with coordinate information in three-dimensional space. Each point contains at least three coordinate information (X, Y, Z), which is used to accurately describe the geometric shape of the object surface.
Eye-hand calibration	Eye-hand calibration	determines the relative pose (position + attitude) between the Robot coordinate system ("hand") and the Camera coordinate system ("eye"), so that the vision system can accurately guide the Robot to complete the grabbing task.
	Intrinsic Parameter	Parameter that describes the internal properties of the Camera and describes the optical properties of the Camera itself, regardless of the Camera's position in the world coordinate system. These Parameters remain unchanged during the use of Camera.
	Extrinsic Parameter	is the Camera's pose Parameter (rotation matrix + translation vector) in the world coordinate system, describing its spatial position and orientation.
	Euler Angles	is a way to describe the attitude of an object in three-dimensional space. It uses three angular parameters (pitch angle, yaw angle and roll angle) to represent the rotation of the object in three-dimensional space.
	TCP	TCP (Tool Center Point), tool center point, refers to the point located at the tip of the End-effector. In order to complete tasks such as Target grabbing, when the Robot is moved to a certain point in the space, the essence is to move the tool center point to that point.
Visual Model	Visual Model	deep learning model can perform inference on input images and obtain inference results such as masks, bounding boxes, key points, and scores of objects.
	Key points	Feature points with clear semantic or geometric meaning in the XQZ three-dimensional model are used to describe the local structure or global posture of the target. In pose estimation based on key points, the overall pose (position + pose) of the target is inferred by detecting the positions of these points.
	pose	represents the set of position (Position) and attitude (Orientation) of an object in space.
	mask	uses selected images, graphics or objects to block the processed image (all or part of it) to control the image processing area or process. The specific image or object used for overlaying is called a mask.
	bounding box	is a rectangular box used to locate target objects in computer vision and machine learning. The position, size and range of the target are marked by coordinates.
Visual Process	Coarse Matching	Coarse Matching is the process of matching the key points of the template Point Cloud with the key points predicted by the model on the actual Point Cloud.
	Fine matching	Fine matching is the process of matching the template Point Cloud with the actual Point Cloud so that the Target posture in the actual Point Cloud coincides with the Target posture of the template Point Cloud as much as possible, thereby optimizing the Target posture in the actual Point Cloud.
	ROI	In machine vision and image processing, the area to be processed is selected from the processed image, called ROI. In PickWiz, ROI 3D and ROI 2D need to be set separately.

Previous pageInstallation Guide

Next pageSoftware Interface