Robust Object Recognition with Genetic Algorithm and Composite Saliency Map

Object detection and recognition have emerged as crucial research areas, driven by the dire need to address the challenges posed by identifying and recognizing objects in diverse domains. Accurate object identification and recognition is required for tasks such as autonomous navigation, video surveillance, precision agriculture, and medical imaging. We propose a method based on a unique combination of machine learning techniques for object recognition. The proposed approach starts by applying K-means segmentation to the input image to cluster similar regions and colors. Next, a composite saliency map is generated based on the K-means segmentation output. Subsequently, the technique of extracting objects from the saliency map is accomplished through the employment of the connected pixel object extraction method. Ultimately, the Genetic Algorithm is utilized to optimize decision tree classifier for the purpose of object recognition. The MSRC dataset was utilized to assess the proposed approach, resulting in an accuracy of 81.5%, accompanied by a precision of 83.3% and recall of 86.1%. The results depicts that the our approach is highly valuable for accurate object categorization, wherein the utilization of Genetic Algorithm significantly enhances the precision of the classification.