Compared with other non-destructive testing methods, ultrasonic infrared thermography has a larger detection range, intuitive results, and selective identification characteristics for crack damage. This article conducts experimental research on the detection of crack defects in aluminum alloy specimens using ultrasonic infrared thermal imaging technology. Firstly, a theoretical model is established, and based on the finite element simulation analysis results of the heat generation mechanism and thermal wave diffusion behavior in the crack area under ultrasonic excitation, the frequency modulated pulse ultrasonic excitation form is introduced. The principal component analysis method and dual channel orthogonal demodulation algorithm are proposed for the extraction of surface thermal wave signals. Secondly, a lightweight ultrasonic infrared thermal imaging detection system built independently is used for pulse frequency modulation ultrasonic excitation to obtain temperature sequences and perform post-processing. Finally, crack defects of different sizes were fabricated on three aluminum alloy specimens, and the correctness of the feature extraction algorithm and the feasibility of ultrasonic infrared thermal imaging method in aluminum alloy material crack detection were verified through experiments. The amplitude map of the dual channel orthogonal demodulation algorithm with high feature image resolution and good presentation of crack defects and the first principal component analysis method were selected, and the influence of different pulse modulation parameters on the detection effect was explored. It is ultimately known that the best detection effect is achieved at a starting frequency of 2 Hz, a ending frequency of 5 Hz, and a scanning period of 6 s; The detection effect of crack defects was quantified using eigenvalue contrast and contrast fluctuation percentage. For the DOD amplitude feature image, the contrast increased by 48.63%, 47.27%, and 35.22% respectively compared to the average situation; For the PCA first principal component feature images, the contrast was improved by 41.45%, 44.26%, and 42.70%, respectively.Firstly, a lightweight ultrasonic infrared thermal imaging inspection system is built based on heat generation mechanism of defects under ultrasonic excitation. The experimental platform includes an ultrasonic excitation system, an infrared thermal imaging system, and a bench. The ultrasonic power supply is controlled by computer and data acquisition card to realize the excitation of FM pulse signal, and the small infrared camera is used to capture the thermal signal of the specimen for feature extraction. Secondly, the temperature sequence acquisition and two feature extraction methods (Dual orthogonal demodulation and Principal component analysis) are given. In addition, the universal experimental machine is utilized to obtain cracks of different sizes by stretching on three aluminum alloy specimens. Lastly, a series of experiments are carried out on the constructed experimental platform, and the effects of different pulse modulation parameters on the detection effect are explored.Through a series of experiments on the constructed experimental platform, the correctness of the feature extraction algorithm and the feasibility of ultrasonic infrared thermography in crack detection of aluminum alloy materials are verified. The amplitude diagram of the dual orthogonal demodulation algorithm and the first principal component of the principal component analysis method are selected, and the feature image has a higher resolution and presents the crack defects better. In addition to the obvious temperature rise at the known crack location, tiny cracks generated at the notch edge during machining can also be observed in the characteristic cloud image. Taking specimen #3 as an example, two tiny cracks with lengths of 218.2 μm and 245.8 μm at the edge of the notch were successfully detected, which further confirms that the ultrasonic infrared thermal imaging method has a high degree of identification of fine cracks (Fig.10). The effects of different pulse modulation parameters on the detection effect are investigated, and the modulation parameter that can achieve the optimal detection effect is given, and it has the optimal detection effect in the starting frequency of 2 Hz, the termination frequency of 5 Hz, and the scanning period of 6 s (Tab.3).For non-destructive testing of cracks in aluminum alloy specimens, this paper chooses the infrared thermal imaging detection method with frequency modulated ultrasonic pulse excitation to conduct research on the detection of metal cracks in aluminum alloys. The main conclusions are as follows: 1) For the detection method that uses frequency modulated ultrasonic pulses as the excitation method for the detection device and extracts features from surface thermal wave signals, by extracting features from the simulated defect heat generation model, the amplitude characteristics of the first principal component using principal component analysis (PCA) and the dual orthogonal demodulation algorithm (DOD) can characterize the defect. 2) Three 7075 aluminum alloy sheet specimens with artificial crack defects were tested, and the crack shape and size on the thermal wave image obtained were basically the same as the actual crack. In addition to the crack defects obtained from experiments, this detection method also successfully identified two small cracks with lengths of and at the edge of known defects, further confirming the high recognition ability of ultrasonic infrared thermography for small cracks. 3) Changing the different modulation parameters of the pulse sequence will have a significant impact on the detection performance. The results show that under the given parameter combination conditions, the crack defect feature image has the highest contrast when the starting frequency, ending frequency, and compression period are used. For DOD amplitude feature images, the contrast has increased by 48.63%, 47.27%, and 35.22% respectively compared to the average situation; For the PCA first principal component feature images, the contrast was improved by 41.45%, 44.26%, and 42.70%, respectively.