This research develops an innovative method to detect the halalness of meat by integrating Electronic Nose (E-Nose) using GM502B (C2HsCH) and GM702B (VOC) multichannel gas sensors, as well as AI Vision-based image processing technology with Edge Impulse and YOLO algorithms. E-Nose recognizes the chemical aroma profile of four categories of meat samples, namely raw beef, cooked beef, raw pork, and cooked pork, taken from 5 different markets (Jelupang Market, BSD Modern Market, Bintaro Market and Anugerah Jaya Babi Shop) and tested in a closed acrylic box. The test results show that the volatile gas levels in pork, both raw and cooked, are consistently higher than in beef, so it can be used as an indicator to distinguish halal aroma. In detail, raw beef has a C2H3CH content of 478-487 ppm and VOC of 487-500 ppm, while cooked beef reaches 623-630 ppm (C2HsCH) and 645-669 ppm (VOC). Raw pork shows values of 561-570 ppm (C2HsCH) and 580-588 ppm (VOC), while cooked pork is at 664- 678 ppm (C2HsCH) and 686-700 ppm (VOC). On the other hand, Vision technology utilizes Grove AI Vision V2 which is trained using a dataset of 2000 photos with four classes of meat, and tested at various shooting angles (45°, 90°, 135°) and lighting intensity controlled using a lux meter. The camera position at a 90° angle with lighting above 80 lux resulted in the highest accuracy of 98-99% in visually detecting and classifying all meat categories. This emphasizes the importance of adjusting the angle, lighting, and distance to improve the performance of the vision system. The integration of these two methods was tested simultaneously by monitoring the aroma and image conditions of raw beef, cooked beef, raw pork, and cooked pork. The combined system was able to detect the halal status of meat with good accuracy, indicated by an average F1 score of 88% in visual classification and separation of meat types based on the gas sensor
threshold.
Keywords: E-Nose, Halal Vision, meat detection, halal, machine learning, image
processing,