Quality grading is usually performed to agricultural products after harvesting by sorting machines with existing problems including large size, high price, etc. To meet the demands of family farming and small scale marketing, we designed a small-scale agricultural product sorting machine based on machine vision. Agricultural products were conveyed to the multichannel conveyer belt and sorted with features including shape, color and etc. by the DSPcored machine. Quality of products in separated channels can be analyzed with single frame image. Performance tests were conducted with walnuts, red dates and chestnuts selected as testing objects to the sorting machine. The test parameters including horizontal transmission speed, product transmission interval, parallel transmission channel amount were set to 9.2cm/s, 16cm and 4, respectively. The test results indicated that the sorting machine performed reliably with selection accuracy rate of 91.66%, 94.79% and 97.39% in walnuts, red dates and chestnuts respectively, with culling speed of approximately 8800 products per hour. Hence feasibility was proved that DSP could be performed as the machine vision core component of smallscale intelligent sorting machine, which provided technical supports of miniaturization and low cost for agricultural product sorting machines.