Abstract: With the development of electronics technology, flip-chip technology has been used extensively in microelectronic packaging. However different disadvantages occurred in traditional detection methods. Therefore, in this paper a novel approach was investigated for solder joint inspection which was based on the active thermography. Non-contact thermal excitation was applied on the specimen, and the temperature distribution of the specimen wss obtained by the infrared camera. Eigenvalues including wavelet entropy were obtained by the wavelet transform process and self-organizing map wss used for data training to classify the signals. The experimental results prove that different defects in the solder bumps can be effectively mapped on different locations by the SOM, and they can be classified and recognized from the reference ones by this method. Therefore, the method investigated in this paper is effective in detecting defects in flip-chips.