<p>Corona virus affects the lung and even causes death to human beings. The technique of metaheuristic optimization has gained more recognition in managing intricate optimization issues. Many optimization strategies inspired by natural occurrences have been developed in the last few years. The recent spread of the novel COVID-19 virus suggested that the public health system would have to deal with a number of fatalities. Vaccination, mask use, and social separation are the main measures implemented to reduce the fatal COVID-19 virus's spread. The two key control behaviors of COVID-19 are cooperation and the maintenance of a unique social distance. This paper presents a deep learning-based method for detecting COVID-19 infection from CT images, comprising two main steps: feature extraction and disease classification. In the feature extraction phase, an enhanced texture descriptor, Local Directional Pattern (LDP) based on the Robinson Compass Mask (LDP-RCM), is proposed. This feature is extracted alongside the Gray-Level Co-Occurrence Matrix (GLCM) to effectively capture image characteristics. For disease classification, an improved Deep Convolutional Neural Network (DCNN) with transfer learning is employed for COVID-19 infection detection. To further enhance the accuracy of disease detection, the Social Distancing Induced Corona Virus Optimization Algorithm (COVO) is utilized to optimize the weights of the DCNN. The proposed model obtained 96.50% of accuracy for dataset 1 and 98.18% of accuracy for dataset 2 at 90th learning percentage which is better compared to conventional methods. From the findings, it is proved that the COVO based COVID- 19 detection models are more accurate than the conventional methods for classifying CT images.</p>