An efficient and accurate machine vision detection method was proposed to detect the deformation and perforation at the bottom and edge of the cavity of chip packaging carrier tape. The registration template and standard template images are prepared offline and then detected online during production. During the detection process, the cavity image to be detected is triggered by the sensor, and then the template image and the image to be detected are registered by the template matching method, and the XOR operation is performed to detect the difference between the two images so as to locate the defect. Experiments show that the maximum error rate of edge deformation detection is 0.45%, the maximum error rate of bottom deformation detection is 0.50%, and the maximum error rate of perforation detection is 0.35%. The average detection time of each frame is 0.22 s, which meets the user's requirement that the error rate is less than 1% and the time of each frame is less than 0.5 s. This method can detect the edge deformation and perforation of the chip in real time, and effectively realize the quality monitoring in the process of chip loading processing.