Design of a dual-camera defect detection system for CTP brackets based on FFT and feature fusion

To address the safety risks posed by surface defects such as white edges, cracks, and black streaks that frequently occur during the injection molding of CATL CTP structural supports, as well as the low efficiency and high miss rate of traditional manual visual inspection, a dual-camera online detection system based on Fourier transform and multi-feature fusion is proposed. The system employs a rotary turntable to achieve multi-station coordinated operation and uses vertically arranged dual cameras to synchronously acquire images of the front and back surfaces of the supports. A detection strategy that fuses frequency-domain and spatial-domain features is adopted: in the frequency-domain stage, global frequency features are extracted using the Fast Fourier Transform (FFT), and defect information is enhanced through Gaussian filtering and inverse transformation; in the spatial-domain stage, shape and grayscale features are combined with threshold segmentation and morphological analysis to achieve the classification and recognition of micro-defects such as black streaks and white edges. The system adopts a multi-threaded asynchronous processing architecture, enabling parallel execution of image acquisition and algorithm processing through an interleaved memory read–write mechanism, reducing the single-part inspection cycle to 1.2 s. Experimental results show that, for 4000 tested samples, the recognition accuracy for both black streak and white edge defects reaches 98.4%–98.9%, with a minimum detectable defect size of 7 mm. By integrating a PLC-controlled pneumatic sorting device, the proposed system realizes online double-sided defect detection and automatic rejection of injection-molded parts, providing an innovative solution for surface quality inspection of high-precision structural components.