IMAGE PROCESSING SYSTEM AND IMAGE PROCESSING METHOD

• Main Author: MUTO TAKUMA
• Format: Patent
• Language: English; Japanese
• Published: 13.12.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION

To analyze an object with high accuracy and a smaller calculation amount.SOLUTION: An image processing system according to an embodiment comprises a reduction unit that reads out a first image with a first pixel number that is outputted from an imaging apparatus and is stored in a storage device, then generates a second image with a second pixel number smaller than the first pixel number by reducing the read first image, a first AI model that detects an object with the second image as an input, a cutout unit that cuts out a region on which the object detected by the first AI model is photographed from the first image stored in the storage device, then generates a third image with a third pixel number smaller than the first pixel number including the region, and a second AI model that analyzes the object with the third image as an input.SELECTED DRAWING: Figure 3 【課題】目標物を高い精度でかつ少ない計算量で解析する。【解決手段】本実施形態の画像処理装置は、撮像装置から出力されて記憶装置に格納された第１画素数の第１画像を読み出して、読み出した前記第１画像を縮小することによって前記第１画素数よりも小さい第２画素数の第２画像を生成する縮小部と、前記第２画像を入力として目標物の検出を行う第１ＡＩモデルと、前記記憶装置に格納された前記第１画像から、前記第１ＡＩモデルによって検出された前記目標物が写っている領域を切り抜いて、前記領域を含む前記第１画素数よりも小さい第３画素数の第３画像を生成する切り抜き部と、前記第３画像を入力として前記目標物の解析を行う第２ＡＩモデルと、を備える。【選択図】図３