Trash Sorting System Using Deep Learning Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic resource recovery systems in Taiwan rely heavily on barcode identification and data bank comparisons, which are inefficient and difficult to implement on a large scale, limiting the effectiveness of recycling processes.
Innovation Solution
A smart recycling system utilizing deep learning and computer vision technology, featuring multiple cameras capturing trash objects from various angles, artificial neural networks for identification, a voting and selecting algorithm for accurate classification, and additional sensors for metal detection, enabling efficient and accurate sorting without the need for barcodes or database comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If barcode identification or data bank comparison methods are used, then identification can be performed, but implementation difficulty increases and processing time is wasted
Solution Approach 1:
The system performs preliminary actions by capturing multiple images of trash objects from different angles before identification. The multi-angle image capture and preliminary processing reduce the need for repeated database comparisons, thereby reducing processing time while maintaining identification accuracy
Solution Approach 2:
The patent replaces the mechanical database comparison system with a deep learning-based image recognition system. Instead of manually comparing barcodes against databases, the system uses neural networks to automatically identify trash types from images, significantly reducing processing time and implementation difficulty
2Measurement precision
If barcode identification is used, then identification can be performed, but massive implementation becomes difficult
Solution Approach 1:
The system replaces barcode scanning infrastructure with camera-based image capture and deep learning models. This substitution makes the system easier to implement at scale, as cameras and neural networks can process multiple objects simultaneously without requiring physical barcode labels on each item
Solution Approach 2:
The deep learning model serves multiple identification functions simultaneously, recognizing different types of trash objects (bottles, containers, packaging) from various angles and conditions. This universal identification capability eliminates the need for separate barcode scanning systems for different object types, facilitating massive implementation
3Measurement precision
If multiple cameras capture trash objects from various angles, then identification accuracy improves, but system complexity increases
Solution Approach 1:
The system segments the identification task by assigning different neural network models to process images from different cameras and angles. Each model specializes in recognizing features visible from its specific viewpoint, and their results are combined through a voting algorithm. This segmentation improves accuracy while managing complexity through modular architecture
Solution Approach 2:
The voting and selecting algorithm acts as an intermediary that integrates results from multiple neural network models. It combines the outputs of different models processing images from various angles, resolving conflicts and selecting the most reliable identification result. This intermediary layer manages the complexity of multiple cameras and models while delivering accurate results
Data Source
AI summary
A recycling system and a method based on deep-learning and computer vision technology are disclosed. The system includes a trash sorting device and a trash sorting algorithm. The trash sorting device includes a trash arraying mechanism, trash sensors, a trash transfer mechanism and a controller. The trash arraying mechanism is configured to process trash in a batch manner. The controller drives the trash arraying mechanism according to the signals of trash sensors and controls the sorting gates of the trash sorting mechanism to rotate. The trash sorting algorithm makes use of the images of trash, wherein the images are taken by cameras in different directions. The trash sorting algorithm includes a dynamic object detection algorithm, an image pre-processing algorithm, an identification module and a voting and selecting algorithm. The identification module is based on the convolutional neural networks (CNNs) and may at least identify four kinds of trash.