Addressing the scarcity of monitoring products for lifts for explosive atmospheres, a comprehensive machine room environmental monitoring system has been designed. This system efficiently monitors temperature, humidity, oxygen content, and air pressure, while also detecting flames and smoke within the machine room. Initially, sensors, including DHT11 modules for temperature and humidity, SC03-O2 modules for oxygen content, and BMP180 modules for air pressure, are utilized to gather information. To mitigate environmental impacts and interference, the K210 module captures video footage, leveraging an enhanced YOLOv8 algorithm for precise visual recognition of flames and smoke. To optimize the YOLOv8 algorithm, the conventional pyramid structure is replaced by an adaptive feature pyramid structure, enhancing feature integration and minimizing information loss. Furthermore, the integration of a multi-dimensional collaborative attention module not only simplifies the model but also maintains high detection accuracy. Lastly, the adoption of the SIoU loss function mitigates issues such as false positives, false negatives, and redundant detections, resulting in an overall improvement in performance. Subsequently, the STM32 control unit processes the collected data, which is then transmitted to a remote monitoring platform via GPRS. Upon validation, the system ensures high accuracy in detecting these vital parameters. Notably, the improved algorithm achieves an of 84% and an:0.95 of 51.7%, showcasing a 3.6% and 3.2% increase, respectively, over the original algorithm. Additionally, improvements in Gflops and FPS metrics further demonstrate the system’s effectiveness.