Urban road traffic congestion and unauthorized vehicle intrusion into cycle lanes pose significant safety risks to cyclists. Manual enforcement methods such as monitoring and physical interventions are time- and resource-intensive, prone to human errors, and inefficient. LaneGuard, an automated real-time monitoring system driven by artificial intelligence, is suggested in this paper to automate the detection of cycle lane violations using computer vision and deep learning-based methods. The system monitors real-time feeds from traffic cameras by capturing frames at fixed time intervals and using image segmentation based on a pre-trained Roboflow model through API integration to identify cycle lanes. Vehicle detection is performed using the YOLOv8 object detection model, which uses bounding boxes on detected vehicles. The mask overlay of the bounding box is utilized to detect vehicles that violate or non-violate based on their interaction with the cycle lane. In case of a violation, the vehicle license plate number is detected using Optical Character Recognition (OCR) and stored in an SQL database for enforcement. This paper presents a theoretical overview of LaneGuard, describing its system architecture, feasibility, and potential integration with smart city infrastructure. LaneGuard offers a scalable solution to enhance urban mobility, cyclist safety, and law enforcement effectiveness by reducing the role of manual enforcement and facilitating real-time violation detection.