The invention discloses a wood sawing method based on
computer vision, which sequentially comprises the following steps of: 1, acquiring wood images by using a color industrial camera and an
infrared imaging camera at multiple visual angles of which the
optical axis included angle is 30-60 degrees, obtaining an initial image
data set, and obtaining an initial image
data set of an internal structure of wood by using an ultrasonic wave and
microwave sensor system; step 2, carrying out preprocessing through algorithms such as adaptive median filtering and the like to obtain a standardized image; step 3, using a defect identification model 5000 + sample constructed by the CNN to
train, identify defects and quantify the defects; 4, in combination with the defect and size parameters, an optimal sawing path is planned through an improved A
algorithm, and sawing parameters are adapted; 5, the path is converted into a control instruction to drive a sawing execution mechanism to work, and real-time closed-loop monitoring is carried out; step 6, detecting the quality after sawing, and feeding back and adjusting if the saw
cutting is unqualified; and step 7, storing and encrypting data, and setting authority. According to the method, the defect recognition rate, the wood
utilization rate and the saw
cutting precision are improved, and quality tracing is achieved.