DBSCAN is a well-established density-based clustering algorithm capable of discovering clusters of arbitrary shape with numerous practical applications. Despite the significant advances achieved by optimized variants of DBSCAN, these methods still encounter challenges when handling data with uneven density distributions. Additionally, they fail to optimally distribute the computational load in parallel architectures and are constrained by the need for fixed threshold parameter settings. These limitations represent key bottlenecks in existing DBSCAN variants. To address these issues, we propose a Parallel Density peak based DBSCAN clustering algorithm, called PaD-DBSCAN. This approach dynamically detects changes in density peaks, thereby enhancing parallel processing capabilities and eliminating the drawbacks of fixed parameter settings. Extensive experiments conducted on various datasets demonstrate the effectiveness and superiority of the PaD-DBSCAN, thus justifying our design choices.