ROI-based image processing system that enables real-time object counting and production tracking in garment manufacturing lines.

Abstract

This invention relates to a system and method for real-time counting of products processed on sewing machines in garment production lines and monitoring production data. The system processes video streams from one or more cameras to perform object detection and counting within user-defined regions of interest (ROI). The system includes a counting mechanism that prevents the same object from being counted multiple times by analyzing the positional and temporal changes of objects. This improves counting accuracy and makes production data more reliable. The invention operates with low latency thanks to its GPU-accelerated processing infrastructure and offers high performance compatible with multi-camera systems. The acquired data is stored in a centralized structure and can be monitored in real-time via the user interface. This eliminates the need for manual counting in garment production lines, thus increasing production efficiency and accuracy.

Description

TARIFF REALITY IN GARMENT PRODUCTION LINES TIME-BASED OBJECT COUNTING AND PRODUCTION TRACKING PROVIDING ROI-BASED IMAGE PROCESSING SYSTEM Technical Area This invention is used in image processing, computer vision, AI-assisted object analysis, and industrial automation. and relates to the technical fields of production tracking systems. The invention specifically applies to products processed on sewing machines found in garment production lines. for real-time monitoring, counting, tracking and analysis of production data. It encompasses an ROI (Region of Interest) based image processing system and methodology. State of the Art In garment production lines, determining the production quantity is mostly done manually by operators. This is how it is done. These methods are prone to human error, produce delayed data, and are unrealistic. It does not allow for timely production tracking. Alternatively, sensor-based systems can only detect specific physical passages. sufficient data about the product's movement behavior, direction, position on the production line, and continuity. This is not possible. In these systems, the same product may be detected multiple times or not detected at all. Problems may arise. Existing image processing-based solutions, however, mostly do not focus on a specific production area. It also includes taking into account movements outside the production line and the temporal continuity of objects. It does not take into account. Especially in high-volume production lines: counting the same product multiple times, Misperception of non-production activities, Inconsistent performance at different camera resolutions, delayed data generation, Performance losses when using multiple cameras Problems such as these occur. • • • • • 1 TARIFF REALITY IN GARMENT PRODUCTION LINES TIME LI FLEXIBLE COUNTING AND PRODUCTION TRACKING S ĞLAYAN ROI BASED IMAGE PROCESSING SYSTEM 1 35 For these reasons, we need real-time, ROI-based analysis capabilities, object movement tracking, and dual-purpose software. We need an advanced production tracking system that prevents counting and is compatible with a multi-camera infrastructure. It is heard. Purpose of the Invention The purpose of this invention is; To enable real-time monitoring of products processed on garment production lines, To ensure that products on the production line are counted accurately and reliably. It is incorrect to exclude areas outside of user-defined ROI regions from the analysis. to reduce the numbers, by analyzing the positional changes, directions of movement, and temporal continuity of objects to prevent the same object from being counted more than once, To offer a compatible and scalable structure with multiple camera systems, Achieve low-latency data processing thanks to the GPU-powered processing infrastructure. storing the obtained production data in a central data structure and user interface to enable monitoring via, to enable analysis of production performance and improvement of production efficiency The aim is to present an image processing system and method developed for this purpose. Summary of the Invention The system developed within the scope of the invention takes video streams from one or more cameras and converts them into real-world images. Object detection within user-defined regions of interest (ROI) by processing in real-time. It performs object tracking and object counting. The system analyzes only objects located within the defined ROI zones, thereby influencing the production line. It excludes actions outside of this scope from consideration. The structure developed within the scope of the invention measures the positional changes, direction of movement, and temporal changes of objects. A counting system that prevents the same object from being counted multiple times by analyzing continuity information. It includes the mechanism. The system uses image processing and AI-powered object analysis methods to analyze objects on the production line. It can distinguish products from the background and track moving objects between frames. is able to. The invention operates with low latency thanks to its GPU-accelerated processing infrastructure and supports multiple cameras. It offers high performance in its environment. The collected census data is stored in a central data structure and can be accessed via a user interface. It can be monitored over time and used in production performance analysis. • • • • • • • • 2 2 35 40 Explanation of the Figures Figure 1: Block diagram showing the overall workflow of the system. Detailed Description of the Invention 1. Image Acquisition Structure The system uses video obtained from one or more cameras placed on the garment production line. It uses flows. The cameras are positioned to cover the products being processed on the sewing machines. The acquired images are transferred to the system in real time and the image processing module It is being analyzed by [the relevant authority]. The system is configured to be compatible with camera configurations operating at different resolutions. 2. Image Processing and Object Detection Structure Objects are identified in the captured images using image processing and artificial intelligence-assisted object detection methods. It is analyzed using [methods]. Thanks to this structure, products on the production line are distinguished from the background and appear on the image. It is being detected. The system tracks moving objects by analyzing the continuity between frames. It is carrying out. 3. ROI (Region of Interest) Definition Structure The user can define the areas on the image to be counted. These areas are referred to as the region of interest (ROI). The system only evaluates objects located within the defined ROI zones, production line It excludes areas outside of this from the analysis. ROI regions are normalized to ensure consistent performance across different resolutions. 4. Object Tracking and Counting Mechanism The system tracks the movements of objects within the region of interest; positional change, 3 3 direction of movement temporal continuity Object relationship between squares It analyzes according to its parameters. Thanks to this analysis: Counting the same object repeatedly is prevented. Only objects that meet specific motion criteria are counted. Non-production activities are being filtered out. Production counting is being made more reliable. 5. Performance and Process Structure The system operates using a graphics processing unit (GPU) to provide high processing performance. Thanks to its GPU-powered processing infrastructure, multiple camera images can be processed simultaneously, and Low latency data generation is ensured. The system is configured to operate scalably across multiple camera infrastructures. 6. Data Management and User Interface The counting data obtained by the system is stored in a central data structure. Users can access the interface via: It is possible to view the current production status. we can review past production data, It is able to conduct production performance analyses. It is possible to monitor production efficiency. The system can be configured to be accessible from different client devices. Advantages Provided by the Invention Thanks to this invention: The need for manual counting is reduced. Production data is obtained in real time. Counting the same product multiple times is prevented. ROI-based analysis helps reduce errors originating from non-production areas. Counting is carried out with high accuracy. Large-scale production lines can be monitored with multiple cameras. Thanks to its GPU-supported architecture, low latency is ensured. Production performance can be analyzed. • • • • • • • • • • • • • • • • • • • 4 4 35 40 45 50 Production efficiency is being increased.

Claims

REQUESTS 1. On video streams from one or more cameras, by the user It performs object detection and counting within defined regions of interest (ROI), real-time. a real-time image processing system.

2. According to claim 1, the system records the positional changes of objects within their regions of interest. It includes a system that performs counting by analyzing.

3. According to claim 1 or 2, the system shall consider the temporal continuity of objects and be the same. It includes a counting mechanism that prevents the object from being counted more than once.

4. According to any of the requirements, the system analyzes the direction of movement of the objects. a structure that allows only objects that meet specific motion criteria to be counted It includes.

5. Depending on any of the requests, the system will process the user-defined interest. a structure that normalizes its regions to work consistently at different resolutions It includes.

6. Depending on any of the requirements, the system is a graphics processing unit (GPU) supported processor. Low latency and high-performance data processing using infrastructure It is carrying out.

7. Depending on any of the requirements, the system can capture video from multiple cameras. It has support for multiple cameras that can process streams simultaneously.

8. According to any of the requirements, the system stores the collected census data in a central database. a structure that stores information internally and makes it viewable through a user interface It includes.

9. According to any of the requirements, the system is located in garment production lines. It is designed for counting the products processed on sewing machines.

10. Depending on any of the requirements, the system can analyze production data and a data processing structure that enables the evaluation of production performance It includes. 6

Images