To address the issues of difficulty and low efficiency in recognizing the status of high voltage disconnectors, this study proposes a method for status recognition based on image enhancement and an improved neural network, using disconnector images obtained from video monitoring. In response to problems such as poor quality of original image, unclear features, and a lack of sample diversity, an image feature enhancement algorithm has been designed, which includes rotation and cropping, histogram equalization, bilateral filtering, and adding noise, to achieve key feature enhancement and expansion of small sample datasets. A shallow neural network has been designed with the addition of ECA (Efficient channel attention) module to construct an improved network ECA-CNN, and the impact of different parameter conditions on model performance has been studied. The experimental results show that the designed image enhancement algorithm can effectively improve image quality, highlight key features, and provide high-quality data support for the neural network model training; the proposed ECA-CNN model can further enhance the focus on key image features based on the image enhancement algorithm, achieving a recognition accuracy rate of over 97%.