Machine Vision based Froth Flotation Monitoring and Tracking System
65 بار بازدید -
12 ماه پیش
-
Title: - AI-DeepFrothNet: Continuous Monitoring
Title: - AI-DeepFrothNet: Continuous Monitoring and Tracking of Froth Flotation Working Condition by Root Cause Analysis & Optimized Predictive Control using Machine Vision Technology
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Implementation Plan:
---------------------------------
Step 1: Initially we load the video frames from the Froth Flotation Dataset (FFD).
Step 2: Next we perform Tripartite Froth Image Pre-Processing process, In these pre-processes are three stages which are briefed below,
2.1: Noise Removal & Contrast Enhancement- The RGB images acquired from the dataset are firstly denoised, enhanced, and adjusted its illumination using Putrefaction Enrichment and Tuning Network (PETNet). Next we perform contrast enhancement by using Enrichment Net.
2.2: Data Augmentation- In theis method we have performed image rotation (40 ° , 70 ° , 145 ° , 270 ° ), vertical spinning and horizontal spinning.
2.3: RGB to HyperSpectral Image (HSI) Conversion- In this method the RGB images are converted to HSI to enhanced images by using the DL algorithm named Skipped Attention Gated Recurrent Unit (SkA-GRU).
Step 3: Next we perform the Frame by Frame Flotation Status Classification & Tracking process, In this process we used a DL based object detection algorithm named You Look Only Once-V7 (YOLO-V7). The detected foams in the video frame can be classified into three types such as (1) overloaded minerals with no or limited foams, (2) normal, and (3) underloaded minerals with some foams associated with gangue materials. We perform tracking of detected foams to obtain their working condition throughout the flotation process, In this process we used Siamese Network based Kalman Filter (SNKF).
Step 4: Next, we perform DRL Assisted Root Cause Analysis & Grade-Recovery Prediction process, In this process the overloaded and underloaded frames are provided for culpability identification and root cause analysis by using Multi Agent Deep Q Learning (MA-DQL). Next From the root cause results and features for the overloaded and underloaded frames, the grade-recovery plot is predicted using the SkA-GRU algorithm.
Step 5: Next, The predicted grade-recovery plot for the abnormal frames and their corresponding root causes are provided as input to the optimized controller. In this method we used Gazelle Optimization Algorithm (GOA) logic.
Step 6: Finally, The performance of the proposed work is evaluated in terms of following metrics,
6.1: Accuracy
6.2: Precision
6.3: Recall
6.4: F1-Score
6.5: Grade Vs Recovery
6.6: Confusion Matrix (Actual working condition Vs Detected Working Condition)
6.7: Confusion Matrix (Actual Grade-Recovery Vs Predicted Grade-Recovery)
=========================================================================================================================================================
Software Requirement:
--------------------------------------
1. Tool: Matlab-R2020a
2. OS: Windows 10 – (64-bit)
==========================================================================================================================================================
Note: -
-----------
We perform the EXISTING process based on the REFERENCE 1 Title: - Intelligent Detection Method for Froth Flotation Based on YOLOv5
=============================================================
#MachineVision
#FrothFlotation
#MonitoringSystem
#TrackingSystem
#IndustrialAutomation
#ProcessControl
#DataAnalytics
#SmartTechnology
#EfficiencyImprovement
#InnovationInMining
--------------------------------------------------------------------------------------------------------------------------
All types of network simulation tools support for your network research ideas such as
NS3, OMNET++ , MATLAB , MININET , NS2 , OPNET , PYTHON , COOJA , CONTIKI OS , QUALNET ,
OPNET , NETSIM , GNS3 , RTOOL , OPTISYSTEM , ONESIM ….will be supported under one roof.
We would be delighted to extend our support in utilizing network simulation tools to effectively
resolve any network research challenges that you may encounter.
For further support
visit us at : https://networksimulationtools.com
Mail us at : [email protected]
call us at : 94448 47435
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Implementation Plan:
---------------------------------
Step 1: Initially we load the video frames from the Froth Flotation Dataset (FFD).
Step 2: Next we perform Tripartite Froth Image Pre-Processing process, In these pre-processes are three stages which are briefed below,
2.1: Noise Removal & Contrast Enhancement- The RGB images acquired from the dataset are firstly denoised, enhanced, and adjusted its illumination using Putrefaction Enrichment and Tuning Network (PETNet). Next we perform contrast enhancement by using Enrichment Net.
2.2: Data Augmentation- In theis method we have performed image rotation (40 ° , 70 ° , 145 ° , 270 ° ), vertical spinning and horizontal spinning.
2.3: RGB to HyperSpectral Image (HSI) Conversion- In this method the RGB images are converted to HSI to enhanced images by using the DL algorithm named Skipped Attention Gated Recurrent Unit (SkA-GRU).
Step 3: Next we perform the Frame by Frame Flotation Status Classification & Tracking process, In this process we used a DL based object detection algorithm named You Look Only Once-V7 (YOLO-V7). The detected foams in the video frame can be classified into three types such as (1) overloaded minerals with no or limited foams, (2) normal, and (3) underloaded minerals with some foams associated with gangue materials. We perform tracking of detected foams to obtain their working condition throughout the flotation process, In this process we used Siamese Network based Kalman Filter (SNKF).
Step 4: Next, we perform DRL Assisted Root Cause Analysis & Grade-Recovery Prediction process, In this process the overloaded and underloaded frames are provided for culpability identification and root cause analysis by using Multi Agent Deep Q Learning (MA-DQL). Next From the root cause results and features for the overloaded and underloaded frames, the grade-recovery plot is predicted using the SkA-GRU algorithm.
Step 5: Next, The predicted grade-recovery plot for the abnormal frames and their corresponding root causes are provided as input to the optimized controller. In this method we used Gazelle Optimization Algorithm (GOA) logic.
Step 6: Finally, The performance of the proposed work is evaluated in terms of following metrics,
6.1: Accuracy
6.2: Precision
6.3: Recall
6.4: F1-Score
6.5: Grade Vs Recovery
6.6: Confusion Matrix (Actual working condition Vs Detected Working Condition)
6.7: Confusion Matrix (Actual Grade-Recovery Vs Predicted Grade-Recovery)
=========================================================================================================================================================
Software Requirement:
--------------------------------------
1. Tool: Matlab-R2020a
2. OS: Windows 10 – (64-bit)
==========================================================================================================================================================
Note: -
-----------
We perform the EXISTING process based on the REFERENCE 1 Title: - Intelligent Detection Method for Froth Flotation Based on YOLOv5
=============================================================
#MachineVision
#FrothFlotation
#MonitoringSystem
#TrackingSystem
#IndustrialAutomation
#ProcessControl
#DataAnalytics
#SmartTechnology
#EfficiencyImprovement
#InnovationInMining
--------------------------------------------------------------------------------------------------------------------------
All types of network simulation tools support for your network research ideas such as
NS3, OMNET++ , MATLAB , MININET , NS2 , OPNET , PYTHON , COOJA , CONTIKI OS , QUALNET ,
OPNET , NETSIM , GNS3 , RTOOL , OPTISYSTEM , ONESIM ….will be supported under one roof.
We would be delighted to extend our support in utilizing network simulation tools to effectively
resolve any network research challenges that you may encounter.
For further support
visit us at : https://networksimulationtools.com
Mail us at : [email protected]
call us at : 94448 47435
12 ماه پیش
در تاریخ 1402/05/23 منتشر شده
است.
65
بـار بازدید شده