Abstract: The output of tomatoes in China is in the forefront of the world, and greenhouse planting is the main planting mode of tomatoes in China. At present, most of them are picked manually, but the cost of manual picking is high and the efficiency is low, which is not conducive to the development of the tomato planting industry in China. Based on the above basic situation of tomato planting industry in China, this paper proposes an image processing technology based on A-ENet model, which is used to identify tomato fruits in greenhouse. The A-ENet model introduces the attention mechanism when the Efficient Net network is running to optimize the network operation results. Efficient Net improves the classification accuracy of the network by adjusting and improving the network parameters such as the width and depth of the network. At the same time, attention mechanism is introduced to capture the weight information of the input signal features when the network extracts the input signal features, and actively ignores the interference of environmental factors on the target signal. A-ENet model can solve the problem of subtle differences between recognition targets and recognition errors. At the same time, the network can reduce the interference of random environmental factors in the recognition process, improve the recognition success rate, and play a positive role in the greenhouse tomato fruit recognition problem. Through experiments, the A-ENet model proposed in this paper based on image processing technology has a higher overall recognition rate. This method can build a more efficient and robust target recognition system.