This invention relates to a
urine test strip
color recognition detection method based on
image analysis, comprising the following steps: Step 1, acquiring a
color image; Step 2, detecting black calibration blocks at the four corners; Step 3, geometric correction of the test strip area based on four-point
perspective transformation; Step 4, adaptive extraction of the effective area of the test strip; Step 5, precise segmentation of color blocks; Step 6, correction of uneven illumination and color
standardization; Step 7, multidimensional similarity calculation based on weighted
Lab color space; Step 8, outputting the detection results and marking abnormalities. The beneficial effects of this invention are: the positioning accuracy remains ≥99.8% even in complex scenarios, far exceeding existing
template matching and QR code solutions; it is completely unaffected by test strip deformation or printing offset; the CV value of the same
urine sample repeated under different mobile phones and different lighting conditions is <3%; it successfully solves the problems of batch-to-batch
color difference and test strip aging, with a clinical concordance rate ≥98%; the entire
algorithm has a small computational load, fully meeting the needs of real-time detection.