Abstract: To effectively compensate for the measurement error of the MLX90640 infrared array sensor caused by ambient temperature variations, an improved Particle Swarm Optimization-Back Propagation (PSO-BP) neural network algorithm is proposed. Firstly, the infrared temperature measurement system is calibrated within the range of 30~100℃ using a BR125 blackbody calibration source. Test data at various temperatures under different set distances is collected by the MLX90640 sensor to compile samples for temperature compensation. Secondly, a neural network system model for temperature compensation is established. The Particle Swarm Optimization (PSO) algorithm is employed to calculate particle fitness and locate the global optimal particle. Finally, the optimal solution is assigned to the neural network to optimize temperature error correction. The experimental results show that when the temperature of the test object is between 30℃ and 100℃, the BP neural network optimized by the particle swarm algorithm reduces the average temperature error to 0.56℃ and the maximum error to 1.64℃. The proposed PSO-BP neural network temperature compensation method effectively reduces temperature measurement errors, further improving the accuracy of infrared temperature measurement.