Traditional text-based and conventional image-based CAPTCHA schemes are increasingly vulnerable to automated attacks leveraging modern deep learning and computer vision architectures. To address this security risk with-out compromising user experience, this paper introduces Color Boundary Object Detection (CBOD), a robust and user-friendly image-based CAPTCHA scheme designed to resist automated object detection pipelines. The CBOD framework constructs challenges through a multi-step pipeline: strategic spatial preprocessing, composite quad-image merging, the algorithmic insertion of randomly shaped and colored polygonal boundaries around target objects, and the application of a variable layer of mixed random noise. Security evaluations demonstrate that the proposed scheme achieves exceptional combinatorial robustness, generating a conservatively estimated 1.17 × 10¹² unique challenges from a minimal baseline asset library of 150 images distributed across 10 semantic categories. This structural resilience stems from defensive mechanisms such as image discontinuity, foreground occlusion, and localized noise artifacts, which intentionally exploit the structural vulnerabilities of deep learning-based object de-tection frameworks. Concurrently, an extensive usability study involving 285 participants establishes high human performance, characterized by a 96% first-attempt success rate and a low average completion time of 6.8 seconds. While the current implementation utilizes a predefined asset taxonomy and lacks specialized assistive features for visually impaired individuals, CBOD provides a highly secure, lightweight verification mechanism that effectively balances low backend resource overhead with a seamless user experience.