This paper presents a component-based person detection system that can detect frontal, rear, and side views of people, as well as partially occluded individuals in cluttered scenes. The system uses a framework of support vector machine (SVM) classifiers arranged in two layers, known as Adaptive Combination of Classifiers (ACC), to classify data. The system performs well, even when not all components of a person are detected, and outperforms a full-body person detector. The motivation for the component-based approach is to enhance performance on frontal and rear views and to address the challenge of detecting partially occluded or background-blended people. The system is based on the idea of detecting components of a person's body, such as the head, arms, and legs, rather than the whole body. This approach allows the system to use geometric information about the human body to improve detection accuracy. The system uses Haar wavelets to represent images and SVM classifiers to classify patterns. The component detectors are trained to find specific body parts, and the combination classifier integrates these results to determine if a pattern represents a person. The system's performance is evaluated using ROC curves, which show that the ACC-based system outperforms the VCC-based system and the baseline full-body detector. The ACC system, which uses a linear SVM to combine component classifiers, performs best. The system is able to detect people in various conditions, including partially occluded individuals and those with body parts blending into the background. The results indicate that the component-based approach, combined with the ACC architecture, significantly improves performance in person detection. The framework is applicable to other objects, such as faces and cars, and can be extended to other domains. The system's accuracy and robustness make it a promising solution for various applications, including surveillance and driver assistance systems.This paper presents a component-based person detection system that can detect frontal, rear, and side views of people, as well as partially occluded individuals in cluttered scenes. The system uses a framework of support vector machine (SVM) classifiers arranged in two layers, known as Adaptive Combination of Classifiers (ACC), to classify data. The system performs well, even when not all components of a person are detected, and outperforms a full-body person detector. The motivation for the component-based approach is to enhance performance on frontal and rear views and to address the challenge of detecting partially occluded or background-blended people. The system is based on the idea of detecting components of a person's body, such as the head, arms, and legs, rather than the whole body. This approach allows the system to use geometric information about the human body to improve detection accuracy. The system uses Haar wavelets to represent images and SVM classifiers to classify patterns. The component detectors are trained to find specific body parts, and the combination classifier integrates these results to determine if a pattern represents a person. The system's performance is evaluated using ROC curves, which show that the ACC-based system outperforms the VCC-based system and the baseline full-body detector. The ACC system, which uses a linear SVM to combine component classifiers, performs best. The system is able to detect people in various conditions, including partially occluded individuals and those with body parts blending into the background. The results indicate that the component-based approach, combined with the ACC architecture, significantly improves performance in person detection. The framework is applicable to other objects, such as faces and cars, and can be extended to other domains. The system's accuracy and robustness make it a promising solution for various applications, including surveillance and driver assistance systems.