In visual classification, understanding among humans and objects is one of the main problems. Human object detection struggles to detect the human as well as object during complex interactions among them. In literature, some of the methods are presented to detect the human and object based on coarse spatial information and appearance features but they fail in complex situations. Hence, in this paper, Black Widow Optimization (BWO) based Deep Convolutional Neural Network (DCNN) Learning model is designed to identify the human as well as object from the video frames. The hyper parameters of the DCNN are optimally selected with the help of BWO algorithm. In the proposed methodology, pre-processing is used to enhance the image quality as well as removing noise from the images by using the gaussian filter and background subtraction. The human and objects are detected from the video frames with the help of DCNN, and performances are evaluated. The proposed method is implemented in MATLAB and statistical measurements are considered to evaluate the performance such as accuracy, sensitivity, precision, recall and F_Measure, respectively. The proposed method is compared with the existing methods such as Convolutional Neural Network (CNN), CNN-Emperor Penguin Optimization (EPO) and CNN-Particle Swarm Optimization (PSO), respectively.

在视觉分类中，人与物之间的理解是主要问题之一。人体物体检测在人类和物体之间的复杂交互中努力检测它们。在文献中，提出了一些基于粗略空间信息和外观特征来检测人和物体的方法，但它们在复杂情况下失败。因此，在本文中，基于黑寡妇优化 (BWO) 的深度卷积神经网络 (DCNN) 学习模型旨在从视频帧中识别人和物体。借助 BWO 算法优化选择 DCNN 的超参数。在所提出的方法中，预处理用于通过使用高斯滤波器和背景减法来提高图像质量以及从图像中去除噪声。借助 DCNN 从视频帧中检测人和物体，并评估性能。所提出的方法在MATLAB中实现，并考虑统计测量来分别评估准确度、灵敏度、精确度、召回率和F_Measure等性能。将所提出的方法分别与卷积神经网络（CNN）、CNN-帝企鹅优化（EPO）和CNN-粒子群优化（PSO）等现有方法进行了比较。