Aiming at the problems that the existing automatic instrument reading algorithm occupies a large space, thereasoning speed is slow, and it cannot ellectively segment the dense and small obiects in the image, an improvecDeepLabV3+ pointer instrument segmentation algorithm is proposed, Firstly, MobileNetV2 is used to build thenetwork backbone to reduce the amount of parameters and inference weight, and improve the detection speed. Secondlythe CSP-ASPP structure is designed through the block merge strategy to reduce the amount of parameters while ensuringthe network performance, Then, the improved SKFF module is used to fuse multi-scale features in a non-linear mannerthrough the self-attention mechanism, and the two-scale feature fusion in the original network decoder is changed to fourscale feature fusion, Finally, the Dice Loss iointly weighted by cross-entropy loss is used as the total loss function of thenetwork to solve the problem of uneven distribution of pixels in each category in instrument segmentation. Finally, it isproved by experiments that the improved DeepLabV3 + average intersection ratio (mloU) and mean pixel accuracy(mPA) reached 89. 3% and 94. 8%, respectively, increased by 0.7% and 0. 6% compared with the original network, butthe amount of parameters and inference weight is only about 7% of the original network, while the inference speed onGPU and CPU reaches 91 and 16 frames/s, respectively. Meet the reguirements of realtime detection, which solves theproblem of difficult deployment of embedded devices and improves the efficiency of automatic instrument reading.