Abstract: Traditional channel scenario classification is usually based on geometric and physical environmental features. However, in urban vehicle-to-vehicle (V2V) channel environments, different channel scenarios are difficult to distinguish due to overlapping environmental features. As a result, the 3rd Generation Partnership Project (3GPP) regards them as a single channel scenario in its TR37.885 report, which limits the accuracy of channel modeling. To improve the reliability and accuracy of channel scenario classification and modeling for urban V2V channels, this paper proposes a channel scenario classification and modeling mechanism based on multimodal scenario features. First, a multi-modal scenario feature dataset is constructed, including geometric and physical environmental features calculated from three-dimensional topographic maps, as well as channel characteristic features extracted from measured channel data. Then, an Adaptive Parameter Adjustment Iterative Self-organizing Data Analysis Techniques Algorithm (APA-ISODATA) is proposed to adaptively cluster the scenario feature dataset. Finally, corresponding channel models are built based on the scenario classification results. Experimental results show that compared with the traditional ISODATA algorithm, the APA-ISODATA algorithm improves clustering validity metrics such as silhouette coefficient, Calinski-Harabasz Index (CHI), and Davies-Bouldin Index (DBI) by approximately 12.4%, 11.1%, and 26%, respectively. Furthermore, the proposed scenario classification method overcomes the limitation of treating urban V2V channel scenarios as a single category, as defined in 3GPP TR37.885, and subdivides urban V2V channel scenarios into three distinct categories: urban edge, typical urban area, and urban canyon. This significantly mitigates the problem of classification ambiguity caused by overlapping environmental features in traditional methods. Based on this refined classification, three scenario-specific urban V2V channel model are established. Therefore, the proposed method can effectively enhance the reliability and accuracy of channel scenario classification and modeling.