Mech-Vision 2.0.0 Release Notes

This document introduces the new features, improvements, and resolved issues of Mech-Vision 2.0.0.

New Features

Restructured and Improved Solution Library

The solution library has been improved in Mech-Vision 2.0.0. Based on users’ actual requirements, the cases have been re-categorized, and new cases have been added to each category. See the following for details on each category.

  • Application templates: General and standardized solution templates are provided for common application scenarios, facilitating rapid on-site deployment and solution tuning.

  • Hands-on examples: Frequently used Procedures and workflow fragments are available to fulfill common on-site requirements. They are applicable not only in fundamental teaching but can also be seamlessly copied and pasted to achieve essential functionalities.

  • Typical cases:Refined from numerous successful real-life scenarios, typical cases serve as references for enhancing your solution. These solutions can be run with one click, suitable for showcasing 3D robot guidance applications.

Brand New “Target Object Editor”

Mech-Vision 2.0.0 provides a brand new target object editor, which integrates features for configuring target object, including generating and editing the point cloud models, adding pick points, and setting up collision models. With the clear configuration workflows, you can follow the step-by-step instructions to complete the target object configuration and then apply the configured target object in the subsequent matching process. Meanwhile, starting from version 2.0.0, Mech-Viz project must belong to the same solution as the associated Mech-Vision project to share the tool and target object configurations.

For more notes about the target object editor, please refer to Upgrade Notes.

New “3D Target Object Recognition” Step

The 3D Target Object Recognition Step is introduced in Mech-Vision 2.0.0. It integrates the “3D Target Object Recognition” tool, including functions such as point cloud preprocessing, 3D matching, and removing poses of overlapped objects, thereby facilitating rapid target object recognition.

New “3D Matching” Step

The 3D Matching Step, which integrates the functions of fine matching and coarse matching, is introduced in Mech-Vision 2.0.0. This Step matches the point cloud model of the target object with the original point cloud, removes poses of overlapped and coinciding objects, and outputs accurate poses of the target objects. This Step is suitable for most scenarios requiring coarse and fine matching for poses of the same type of objects.

New “Custom Alert” Feature

The Custom Alert feature is introduced in Mech-Vision 2.0.0. It allows users to set trigger conditions, alert codes, and alert messages. When an anomaly is detected in the data during the project’s execution, the system automatically triggers alerts to help users quickly identify and resolve issues, ensuring stable production.

New “System Memory Risk Alert” Feature

Mech-Vision 2.0.0 adds a “System Memory Risk Alert” feature that prompts you to restart the software if it has been running continuously for too long or if system memory is insufficient.

Other New Steps

No. Step Description

1

Predict Pick Points

Integrates the functions and usage scenarios of the “Predict Pick Points (Any Objects)” and “Predict Pick Points V2” Steps; applicable to supermarket and box-shaped object picking scenarios.

2

Generate Picking Strategy for Target Object

Generates the picking strategy for the target object based on the pick points,object center points, and other data output by the prior Steps.

New “Continuous Run” Feature

A new “Continuous Run” feature is introduced in Mech-Vision 2.0.0. The Max number of continuous executions and Interval after project execution can be customized.

Improvements

Improved Solution and Project-Related Settings

Starting from Mech-Vision 2.0.0, the “solution” will be used as the only storage unit, as explained below.

  • When creating a new solution, the projects within it will be automatically registered, i.e., these projects can be called by external services by default.

  • Automatically loading projects that do not belong to any solution is no longer supported, i.e., such projects cannot be called by external services.

  • The “Save Project As” and “Backup Project” options are no longer supported. To back up a project, please click “Back up Solution” in the context menu.

Improved Production Interface

The production interface has been improved in Mech-Vision 2.0.0. See the following for details.

  • Compatibility Issues

    • The production interface configured in Mech-Vision prior to version 2.0.0 cannot be opened in Mech-Vision 2.0.0, which requires a re-configure of the production interface in the latest software.

  • Improved Production Interface Configurator

    • When configuring the Basic info, the limit on the number of production units has been modified. Now up to ten production units can be added.

    • When configuring the new target object, you can Import STL file or use the Target object editor to make the point cloud model of the target object. Using the “Matching Model and Pick Point Editor” to make the point cloud model is no longer supported.

    • When configuring the New target object, if “Model-making project” is used to make the point cloud model, you should specify the Step port instead of the Step to select the point cloud and pick points.

    • When configuring the General settings, you can select the default interface to enter when the software is restarted. This setting will only take effect if the solution is already auto-loaded.

  • Improved Production Interface

    • Monitoring of the camera network speed is supported, and the camera bandwidth logs can be viewed in the “Troubleshooting” interface of the “Maintenance” window.

    • Alert records can be filtered. You can choose to view all alerts, default alerts, or custom alerts.

    • The layout for displaying the target object information has been improved to make the information presentation more logical and easier to read.

  • Other Improvements

    • A new Production interface setting has been added to the menu bar  Options in Mech-Vision. You can set whether to display high-quality point clouds in the production interface or use grid view for the standby view.

Improved Step Library

Mech-Vision 2.0.0 has recategorized and streamlined the Step library, as explained below.

  • According to actual project construction requirements, frequently used Steps have been recategorized to better align with user habits.

  • Some outdated Steps and all Procedures have been removed to streamline the Step library.

    Click here to view other deleted Steps that can be replaced by other Steps
    No. Step Alternative Step

    1

    Remove Cloud Points from Point Cloud

    Delete Points in Point Cloud

    2

    3D Coarse Matching

    3D Coarse Matching V2

    3

    3D Coarse Matching (Multiple Models)

    3D Matching and Classification (Multiple Models)

    4

    3D Fine Matching (Multiple Models)

    3D Matching and Classification (Multiple Models)

    5

    Remove Overlapped Objects

    Remove Overlapped Objects (Lite)

    6

    Remove Overlapped Objects V2

    Remove Overlapped Objects (Lite)

    7

    Generate Cylinder Point Cloud Model

    Create Object Model

    8

    Collect Poses in 3D ROI

    Validate Existence of Poses in 3D ROI

    9

    Extract Point Cloud in 3D Box

    Extract 3D Points in Cuboid

    10

    Extract Image Regions by Mask

    Apply Masks to Image

    11

    Read Images

    Read Images V2

    12

    Read Point Cloud

    Read Point Cloud V2

    13

    Calc Absolute Values

    Numeric Operation

    14

    Cloud Num Limit

    Trim Input List

    15

    Cloud Processing (GPU)

    Calc Normals of Point Cloud and Filter It, Calc Normals and Estimate Edges of Point Cloud

    16

    From CloudXYZ To CloudNormal

    Calc Normals of Point Cloud and Filter It, Calc Normals and Estimate Edges of Point Cloud

    17

    Filter by Labels

    Validate Labels and Output Flags + Filter

    18

    Get First Image from Image List

    Trim Input List

    19

    Merge Point Cloud Lists

    Merge Data

    20

    Merge Label List

    Merge Data

    21

    Merge LineSegment Lists

    Merge Data

    22

    Merge Pose Lists

    Merge Data

    23

    Normal Estimation

    Calc Normals of Point Cloud and Filter It, Calc Normals and Estimate Edges of Point Cloud

    24

    Number Scaling

    Numeric Operation

    25

    Point Cloud Filter

    Validate Point Clouds

    26

    Pose Filter

    Validate Poses by Included Angles to Reference Direction

    27

    Smooth Depth Map

    Image Filtering

    28

    Sort Point Clouds

    Calc Specified Property of Point Clouds+Sort List and Output Index List+Reorder by Index List

    29

    Subtract Real Numbers

    Numeric Operation

    30

    Trim Pose List

    Trim Input List

    31

    Map Depth to RGB

    Rectify Image by Frame Transformation

    32

    Sort 3D Poses

    Sort 3D Poses V2

    33

    From Variants to Numbers

    Convert Data Type

    34

    From Variants to Variant

    Convert Data Type

    35

    Find Corners

    Detect Vertex

    36

    From Variant to Variants

    Convert Data Type

    37

    Generate Ring Point Cloud

    Create Object Model

    38

    3D Target Object Recognition

    3D Target Object Recognition

    39

    Draw Min Circumscribed Rectangles of Masks

    Detect Shape Feature of Region

    40

    Adjust Poses

    Adjust Poses V2

    41

    Get Highest Layer Regions in Depth Map

    Get Highest Layer Clouds

    42

    Find Corners

    Detect Vertex

    43

    Allocator

    Accept All
    Click here to view other deleted Steps that cannot be replaced by other Steps

    Extract Empty Regions in Depth Map within 3D ROI

    Calc Point Cloud Curvatures

    Smooth Point Cloud and Estimate Normals

    Correct Point Cloud Distortion

    Project Points onto Plane

    Align Plane Point Clouds

    Filter Ring Point Cloud List

    Detect Occluded Objects

    Generate Cloud Wall

    Remove Poses outside Bin

    Calc Calibration Matrix of Structured Light Sensor

    Draw Polygon Vertices

    Suppress Neighboring Poses with Low Scores (NMS)

    Calc Histogram

    Depth Encoding

    Get Highest Score Result

    Histogram Matcher

    Path Target Matching

    Convert 2D Poses to 3D Poses

    Calc Mask Distances

    Validate Masks Containing 2D Poses

    Extract 2D Path Function

    Load 2D Path

    Group 2D Poses

    Adjust Poses by Offsets

    Adjust Poses by Tilt

    Adjust Inaccurate Poses Caused by Camera Distortion

    Rectify Ring Object Poses

    Keep Poses Distributed in Regular Polygon

    Generate Pose/Offset

    Compare Z Values of Poses with Threshold

    Validate 2D Poses by Mask

    Validate Underlying Poses

    Find Correspondence between Poses and Offsets

    Read 3D ROI Center

    Read STL

    Save Regions around Poses as 3D ROIs

    Copy Images

    Measure Result

    Verify Pick Points

    Save Result To XML File

    Depth Clustering Along Scan Lines

    Measure Plane Height Differences Along Direction Parallel To Axes

    Measure Gap Width

    Calc Plane Width

    Calc Profile

    Calc Profile By Sampling

    Divide Point Cloud into Smaller Parts Evenly

    Compare Two Depth Maps

    Convert Disparity Image to Depth Map

    Merge Depth Maps

    Draw Coherent Lines

    Background Subtraction

    From Actual Dimensions to Dimensions in Pixels

    Extract Regions of Large Normal Deviations

    Generate Discrete Poses Revolving around Reference Pose

    Replace Elements In List

    Invalidate Depth Pixels outside 3D ROI

    Calc Projection Length along Reference Direction

    Generate Test Image

    Calc Box Dimensions

    Generate Test Point Cloud

    Calc Edge Point Cloud Normals

    Save Path Targets

    Remove All Overlapping Poses

    Load Targets in Path and Apply Affine Transform

    From Numbers to Variants

    Calc Diameter and Thickness

    Offset Poses in Cylinder

    Detect Line Segments

    Test

    OCR

    Send Point Cloud to External Service

    Filter Point Cloud by Model and Poses

    Read QR Code

    Periodic Trigger

    Trigger

    Find Hole 2D

    Calc Mean Gray Value

    Evaluate Image Clarity

    Count Color Info

    2D Shape Matching

    Template Matching

    Record Criterion Pose and Calc Transformation

    Make Template
    Click to view the deleted Procedures

    Calc Hole Center Poses and Diameters

    Calc Oblong Hole Center Poses and Axis Lengths

    Extract Planar Point Clouds

    Select Point Cloud from Depth Map

    Transform Poses to Custom Frame

    Sort Clouds Based on XOY Distance to Camera Center

    Sort Pick Points

    Sort by Three Values

    Sort by Two Values

    Point Cloud Preprocessing

    Calc Color Image for Highest Layer

    Calc Mask for Highest Layer

    Transform Clouds from Current Coordinate to Specific Coordinate

    3D Matching

    Calc and Adjust Poses from Planar Point Clouds

    Apply Masks to Color Image

    Binary Classification Based on Elements Number

    Correct system drift in EIH setup

    Filter Out Point Clouds that Exceed the Size Limit

    Filter Out Poses Outside ROI

    Filter Out Poses that Exceed the Angle Limit

    Save Images and Step Parameters
    NOTE

    You can find the corresponding project construction example in the Hands-on Examples category in the Solution Library.

Improved Steps

No. Step Description

1

Capture Images from Camera

  • Image data of various formats can be read.

    • 2D image: JPG, JPEG, PNG, BMP

    • Depth map: PNG, TIFF

  • The external 2D cameras, LMI cameras, and other third-party cameras are no longer supported.

  • The “Background Removal Settings” parameter is removed, and removing the background information from the depth map is no longer supported.

2

Output

  • The Collision Detection Settings and Update Scene Object Settings are supported.

  • The Step was renamed from “Procedure Out” to “Output”.

3

Deep Learning Model Package Inference

  • Starting from Mech-Vision 2.0.0, only the Object Detection, Instance Segmentation, Classification, and Defect Segmentation model packages can be loaded in the “Deep Learning Model Package Inference” Step. The Unsupervised Segmentation, Fast Positioning, Text Detection, and Text Recognition model packages are no longer supported.

  • Starting from Mech-Vision 2.0.0, only model packages exported from Mech-DLK 2.4.1 or above can be loaded in the “Deep Learning Model Package Inference” Step.

Solutions/projects with deep learning models created in Mech-Vision 1.7.4 or later versions can be opened and run in Mech-Vision 2.0.0, while solutions/projects created in Mech-Vision earlier than version 1.7.4 cannot be opened.

4

Create Object Model

Inputting the dimensions of the object model is supported.

5

From Depth Map to Point Cloud

The “Background Removal Settings” parameter is removed, and removing the background information from the depth map is no longer supported.

6

Delete Points in Point Cloud

The Step has been renamed from “Remove Cloud Points from Point Cloud” to “Delete Points in Point Cloud.”

Improved Interface and User Interaction

The interface and user interaction have been improved in Mech-Vision 2.0.0. See the following for details.

  • Improved Home Interface

    • The style and layout of the toolbar has been improved.

    • The layout settings of the home interface has been improved. After restarting the software, the interface will revert to the default layout and will not maintain the interface layout as it was before the software was closed.

  • Improved Step Section and Interaction

    • Added ☆ markers to some important Steps for better identification.

    • The connection guidance for Step output ports has been introduced. An arrow will appear when the cursor hovers over an output port.

    • A “Config wizard” button is added to super Steps. Clicking this button will open the corresponding tool.

    • The method of viewing visual output results by long-pressing and dragging the left mouse button is no longer supported. Visual output results can now only be viewed by enabling Debug Output and the Step visualization feature (by clicking the “eye” icon on the Step block).

    • The Run the project down from the current Step feature is no longer supported.

  • Improved Visualization Area

    • The mouse interaction in the visualization area has been improved and the animated guidance information has been added to the operation prompt window.

Features No Longer Supported

The following feature are no longer supported from Mech-Vision 2.0.0.

  • Camera Calibration (Quick)

  • Intrinsic Parameter Calibration

  • Measurement Mode

  • Glue Wizard

  • Camera Viewer

  • Matching Model and Pick Point Editor

    Please use the target object editor to make the point cloud model and add pick points.

  • EIH System Drift Auto-Correction

    This feature is under maintenance and is temporarily converted to a plug-in that cannot be selected.

Others

  • The Python engine used in Mech-Vision is upgraded to version 3.9.13.

Resolved Issues

The following issues have been resolved in Mech-Vision 2.0.0:

  • The software crashed when a new project with the same name as an existing project is created under the solution.

  • After configuring permissions for the solution, double-clicking the VIS file could not open the project properly.

  • The text and axis size settings in the 3D simulation area of the “Path Planning” tool will only take effect after reopening the “Path Planning” tool.

  • Resetting the path planning configuration in the “Path Planning” tool may cause Mech-Vision to crash.

  • After modifying the configurations in the deep learning model package management tool, the configurations were lost upon reopening the software.

  • In the configuration before calibration for gantry robots, after selecting “Rotation around the Z-axis” as the movement affecting the camera position, the setting did not take effect.

  • During the gantry robot calibration process, using multiple calibration boards to add calibration points caused the calibration to fail.

  • When manually entering the robot flange pose during calibration, the software failed to update the Euler angle convention based on the robot model information.

  • When the graphics card information could not be retrieved properly, the software incorrectly judged the GPU usage as high, resulting in the point cloud not being displayed.

  • There was a low probability of software crashes when using the downsampling-related functions.

  • When running scripts with the “Calc Results by Python” Step, garbled texts appear in the error message if an exception occurs.

  • There was a low probability of software crashes when using the “Mask Gridding” Step to process certain data.

  • The path generated by the “Generate Path from Contour” Step was incorrect.

For more notes after the upgrade, please refer to Mech-Vision 2.0.0 Upgrade Notes.

We Value Your Privacy

We use cookies to provide you with the best possible experience on our website. By continuing to use the site, you acknowledge that you agree to the use of cookies. If you decline, a single cookie will be used to ensure you're not tracked or remembered when you visit this website.

  • Preference
  • I accept
  • I don't accept