Mech-Eye series features

Mech-Eye industrial-use 3D camera

●Comprises an extensive line of 9 models

●Provides a wide field of vision (135 × 90 mm to 3,500 × 2,800 mm)

●Has undergone more than 10,000 hours of continuous operational testing to ensure safety and reliability

Mech-Vision image processing software

●Features leading-edge deep learning technology

●Can recognize a variety of workpieces without CAD inputs

●Is easy to set up thanks to Denso Wave’s proprietary Mech-Eye GUI

Mech-Viz robot control software

●Visualizes recognition results and the robotic environment in 3D

●Supports simulated and real-time interference avoidance and track planning

●Allows you to quickly set up a robot model simply by choosing the model

 

Advantages

Exceptional durability and cost performance

The Mech-Eye industrial-use camera has been tested continuously for 10,000 hours to ensure exceptional durability. In addition, the product offers excellent cost performance since it can accommodate a broad range of target objects and environments utilizing multiple algorithms.

Advanced image processing

Mech-Eye can accommodate a variety of target objects and environments using user-selected recognition methods. It can also accommodate bulk or aligned workpieces, glossy or contrasting workpieces to which tape has been applied, and other workpieces that defy easy recognition.
In addition, users can build image processing systems to suit various applications by using Mech-Vision to add multiple algorithms.

Easy installation and operation

Users need only select the optimal camera from the extensive line of models and complete a simple installation process.

The dedicated GUI is easy to use. Users can also learn how to control the product at Denso Robotics School.

 

Denso Wave’s proprietary Mech-Eye GUI

●Select from the two recognition methods described below. There’s no need for model CAD data.

(1) Matching:Recognizes objects based on profiles created by imaging models and registering their shapes.
This method is suited to bulk workpieces.

(2) Clustering:Recognizes objects whose area matches that of a designated workpiece.
This method is suited to workpieces with a simple shape.

●Configuration can be accomplished entirely by touch operations, ensuring anyone can use the system intuitively.

*Results can be displayed on a Smart TP (if using the RC9) or on a computer monitor (if using the RC8).

 

System architecture

 

Use cases

·Palletizing
·Sorter loading
·Sorting
·Positioning
·Assembly

 

Target objects

Mech-Eye can generate high-quality point cloud data for various objects (boxes, hemp sacks, metal parts, etc.).

• 

  Tightly packed boxes with patterns or tape

• 

  Hemp sacks with patterns

• 

  Bulk boxes

• 

  Bulk metal parts

 

Application videos

• Use of AI assisted 3D Camera for Palletizing Returnable Containers

• Use of 3D Camera and Gripper for Picking Irregularly Shaped

 

Depalletizing of stacked cardboard boxes

Overview and issues

Overview and issues
・Applications in which cardboard boxes stacked on pallets are depalletized

・Applications in which cardboard boxes of mixed shapes are stacked randomly on pallets require 3D image recognition.


Solution
・Use Mech-Eye to realize stable recognition, even when cardboard boxes are stacked randomly.

・·Accommodate multiple types of boxes with deep learning technology.

・Pick a wide variety of box shapes without registering models.

 

Mech-Eye camera

	UHP-140	NANO ULTRA	NANO
Model
Optimal working distance [mm]	300 ± 20	250～800	300～600
Near/Far FOV [mm]	135 × 90 @ 0.28m～ 150 × 100 @ 0.32m	220 × 165 @ 0.25m～ 770 × 550 @ 0.8m	220 × 150 @ 0.3m～ 440 × 300 @ 0.6m
Resolution	2048 × 1536	2400 × 1800	1280 × 1024
Point repeatability Z*¹	2.6μm @ 0.3m Region*² : 0.09μm @ 0.3m	0.1mm @ 0.6m	0.1mm @ 0.5m
VDI/VDE accuracy*³	0.03mm @ 0.3m	0.1mm @ 0.6m	0.1mm @ 0.5m
Typical capture time [s]	0.6～	0.5～	0.6～
Baseline [mm]	80	86	68
Dimensions [mm]	260 × 65 × 142	125 × 46 × 76	145 × 51 × 85
Weight [kg]	1.9	0.7	0.7
Operating temperature [℃]	0～45	0～45	0～45
Communications port	Gigabit ethernet	Gigabit ethernet	Gigabit ethernet
Input	24V DC, 3.75A	24V DC, 3.75A	24V DC, 1.5A
Safety and EMC	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL
IP rating	IP65	IP65	IP65
Spec [mm]

 

　　　　　　

	PRO S	PRO M	Deep*⁴
Model
Optimal working distance [mm]	500～1000	1000～2000	1200～3500
Near/Far FOV [mm]	370 × 240 @ 0.5m～ 800 × 450 @ 1.0m	800 × 450 @ 1.0m～ 1500 × 890 @ 2.0m	1200 × 1000 @ 1.2m～ 3500 × 2800 @ 3.5m
Resolution	1920 × 1200	1920 × 1200	2048 × 1536 (Depth) 2000 x 1500 (RGB)
Point repeatability Z*¹	0.05mm @ 1.0m	0.2mm @ 2.0m	1.0mm @ 3.0m
VDI/VDE accuracy*³	0.1mm @ 1.0m	0.2mm @ 2.0m	3.0mm @ 3.0m
Typical capture time [s]	0.3～	0.3～	0.5～
Baseline [mm]	180	270	300
Dimensions [mm]	265 × 57 × 100	353 × 57 × 100	366 × 77 × 92
Weight [kg]	1.6	1.9	2.4
Operating temperature [℃]	0～45	0～45	-10～45
Communications port	Gigabit ethernet	Gigabit ethernet	Gigabit ethernet
Input	24V DC, 3.75A	24V DC, 3.75A	24V DC, 3.75A
Safety and EMC	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL
IP rating	IP65	IP65	IP65
Spec [mm]

 

	LSR S*⁴	LSR L*⁴	LSR XL*⁴
Model
Optimal working distance [mm]	500～1500	1200～3000	1600～3500
Near/Far FOV [mm]	480 × 360 @ 0.5m～ 1500 × 1200 @ 1.5m	1200 × 1000 @ 1.2m～ 3000 × 2400 @ 3.0m	1280 × 1280 @ 1.6m～ 3000 × 2800 @ 3.5m
Resolution	2048 × 1536 (Depth) 4000 x 3000/2000 x 1500 (RGB)	2048 × 1536 (Depth) 4000 x 3000/2000 x 1500 (RGB)	2448 × 2040 (Depth) 4000 x 3000/2000 x 1500 (RGB)
Point repeatability Z*¹	0.2mm @ 1.5m	0.5mm @ 3.0m	0.2mm @ 3.0m
VDI/VDE accuracy*³	1.0mm @ 1.5m	1.0mm @ 3.0m	1.0mm @ 3.0m
Typical capture time [s]	0.5～	0.5～	0.6～1.1
Baseline [mm]	140	380	800
Dimensions [mm]	228 × 77 × 126	459 × 77 × 86	942 × 88 × 116
Weight [kg]	1.9	2.9	4.5
Operating temperature [℃]	-10～45	-10～45	-10～45
Communications port	Gigabit ethernet	Gigabit ethernet	Gigabit ethernet
Input	24V DC, 3.75A	24V DC, 3.75A	24V DC, 3.75A
Safety and EMC	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL	CE/FCC/VCCI/KC/ISED/NRTL
IP rating	IP67	IP65	IP65
Spec [mm]

*¹ The standard deviation of the single point Z value for 100 measurements. The measurement target is a ceramic plate.
*² The standard deviation of the difference of the average Z value in two local regions for 100 measurements.
The measurement target is a ceramic plate.
*³ Standard: VDI/VDE2634 Part II.
*⁴ DEEP/LSR S/LSR L/LSR XL：Class 2 laser device.
 

IPC

	Beckhoff IPC C6650	Mech-Eye Standard IPC (without GPU)	Mech-Eye Standard IPC (with GPU)
OS	Windows 10 (64bit)	Windows10 IoT Enterprise LTSC2021
CPU	Intel® CoreTM i7, 3.6GHz, 4cores	Intel® CoreTM i5-12400 2.5GHz 6 cores
Memory	32GB	16GB
HDD	1TB	512GB
Power supply	100-240V AC, 600W	100-240V AC,180W	100-240V AC,500W
Graphics card	NVIDIA Quadro P2200	-	NVIDIA GeForce GTX1660S

Mech-DLK deep learning software features

·Mech-DLK is an optional software package that lets customers create deep-learning data.
·Once purchased, it can be used to create multiple learning datasets without additional fees.
·Data security is assured since data doesn’t need to be transferred to any external site.

*DLK: Deep learning kit

 

Example Mech-DLK uses

 

Mech-DLK architecture

 

Recommended operating environment

	Mech-DLK Pro-Train/Standard
Operating system (OS)	Windows 10 or later
CPU	IntelⓇ CoreTM i7-6700 or better
Memory	16 GB or greater
Graphics card	GeForce RTX 3060 or greater
Graphics card driver	Ver. 472.50 or later
Graphics card performance	NVIDIA GeForce series card with computational capability of 6.1 or greater